This is the purist approach solution to the tinny contact mic sound problem if you want the lowest electrical noise, or the lowest distortion at higher levels. Piezo mics aren’t usually considered hi-fi, so distortion isn’t usually a problem of the amplifier – at high levels the mic will not be particularly linear. However, they are remarkably low noise. Although they are high impedance, they don’t need an outrageously high load impedance – not like an electret mic, where you need gigaohms to avoid losing bass. The calculation in the introduction showed a load of 330k was enough, so 1 meg is plenty for a contact mic. The first thing that sprang to mind was a TL071 – the overall schematic is shown below.
Diodes D1,D2 stop you destroying the opamp with large signals from the piezo device if you drop it. You can leave C1 and R4 out if you need less gain, though there is much to be said for the cheap version if you have lots of signal.
The results were good. a 7mV rms 1kHz signal was fed into a 10k:12ohm attenuator, and then coupled via a 15nF capacitor to approximate the piezo source, generating 8uV. A satisfying drop in noise was obtained. However, substituting the TL071 for the usual audio workhorse opamp, the bipolar NE5534 brought the lowest noise of all.
Noise comparison of 2n3819 single fet with opamp version using a NE5534 (left and right channels respectively)
freshwater hydrophone recording using this with an air cell piezo disk hydrophone. At the start of the recording you can faintly hear woodpigeons calling above the surface propagating through the water, and aircraft noise, before the chomping sound
Matching levels of the 1kHz test tone to correct for the slight difference in gain shows the NE5534 offers a noise floor 10-20dB lower than the 2N3819, in return for extra complexity. At the back of my mind was the wonder whether a piezo mic would not have more self-noise anyway, so I removed the test tone and put in the mic. This is a reasonably quiet location. The hum is because this circuit is on a breadboard with the mic laid on the shelf, and some of the hum is mechanical, from the transformer of the power supply powering the unit, which is on the same shelf unit. About 3/4 of the way in some piece of machinery is started by my neighbours – I did not hear this at the time but it was conducted through the house into the shelf with the mic. You wouldn’t normally have to put up with this level of hum – building the circuit into a metal box and using screened cable to the piezo device would fix that. However, the piezo mic has not raised the noise level significantly.
NE5534 at same gain as test circuit but with the contact mic replacing the 1k tone.
This goes to show that the piezo contact mic, for all its uneven frequency response faults, is capable of an excellent noise performance when used correctly. A more practical stereo version of this was constructed in a box, measurements and test results are available.
NE5534 is a great low noise opamp for many audio applications because of it’s low impedance nature. See its datasheet. They give input resistance min 30k and typical 300kohms.
You are trying to reach an input impedance of 1028 kohms by the schematic. However that IC makes it somewhere less than 300 kohms. That lowers the output of the piezo significantly. Especially on lower frequencies.
Never measured carefully, but I assume that you get better true acoustic S/N ratios with 071 than with 5534. (Depending on frequency of interest.)
Thanks for your comment. I have to admit that I made the same assumption initially and used a TL071, but it proved surprisingly noisy in this application, and since I’d socketed the opamp to be able to reuse it, I tried a NE5534 and was rewarded with lower noise. Only then did I look at what was going on.
I disagree with the hypothesis that the open-loop input resistance (you indicate 300k, TI seems to indicate 100k) determines the input resistance of this circuit, because of the action of negative feedback. Let’s take a look at the datasheet on http://www.ti.com/lit/ds/symlink/ne5534.pdf where on page 5 Table 7.5 they cite an open loop input resistance ri of min 30k typ 100k.
The circuit is pitched at a noninverting gain of 11. We can test what happens either using Spice or what really happens by breadboarding it; I’m a real world guy so I breadboarded it.
For laziness I used 27k instead of 56k for R2, R3 and 47u all round because this came to hand. I omitted D1 and D2, because they are there to protect against overload. C2 bypasses R2,3 at signal frequencies so in an ideal world the input impedance would be 1M. I used a 12V power supply. None of these changes should significantly change the input impedance at ~1kHz.
I applied a signal from an AF signal generator (lowish Z of about 50 ohms) of ~1kHz via a 1M resistor in series with a 150nF cap to the input, and on the breadboard it is easy to jumper across the 1M series input resistor to bypass it so it isn’t in circuit.
signal source ~ —||—-/\/\/\/\—- to pin 3 of opamp
bypassed
signal source ~ —||————– to pin 3 of opamp
I bypassed the resistor and lined up the signal to give a level as measured on the ‘scope at 7.64V p-p. I am using a 10x scope probe so levels are measured a factor of 10 low.
I then took out the jumper, so the 1M resistor was in series with the input, and observed this
corresponding to 3.78V. Theoretically the mid-point of a potential divider of two 1M resistors would give 7.64*(1,000,000/(1,000,000+1,000,000))=3.82V, so it’s possible the low reading is showing that the input impedance is slightly lower than 1M. But a scope is accurate to say 5%, and the resistors are 5% tolerance so I can’t really say due to experimental error.
Your hypothesis of the impedance being defined by a 300k open loop input resistance would suggest a reading of 7.64V*(300,000/(1,000,000+300,000) = 1.76V. If TI are right and it is 100k then the output would be ~0.7V. The hypothesis is not confirmed experimentally.
The TL071 is quite a poor choice for 9V battery operation as the spec of Vcc=+/-15V and o/p swing of min +/- 12V means you lose 3V of headroom against each rail. The 15nF source impedance of the piezo disk is also not particularly high-Z at signal frequencies – you only need a modest input resistance to stop the low-frequency rolloff making it tinny. As a result the balance between the lower en of the 5534 (4nV/root Hz (5534) as opposed to 18nV… at 1kHz) is worth more than the lower in of TL071 .6pA/root Hz (5534) and .01pA/root Hz (071), because the latter is impressed across a low capacitive reactance of ~j10k at 1kHz (assuming an effective series 15nF disk capacitance)
Hello Richard, thanks a lot for sharing this valuable information. I will build you pre soon and test it out. I was wondering if there’s any chance to get it working also with phantom power instead of a 9v battery. Do you think it’s possible?
Look at Figure 3 on this page, which is a circuit for a DI box that can be powered by P48 phantom power or battery. You’d replace the majority of that circuit with the preamp presented here, just preserve the power circuitry.
https://sound-au.com/project35.htm
Right. You can multiply the input resistance by the circuit gain.
Thanks for immediately making real life measurements. You proved me that 5534 is the right choice.
The noise stuff is surprising – I’d initially thought these were going to be like condenser mic capsules, with a low capacity at AF of a few tens of picofarads. They’re no such thing – I’ve just measured a 35mm piezo disc on an LCR bridge, it comes in at a massive 28.5nF (~j5000 @ 1kHz), and that low AF impedance is what shifts the balance away from favouring FET technology with its lower noise current/higher noise voltage to favouring bipolar with its higher noise current/lower noise voltage.
This EEV blog by the ex-designer for Rode mics talks about the interaction between transducer capacitance, noise, and the input resistance. That piezo high capacitance appears to be good for low noise, perhaps I want to up the 1M resistor to a 10M though, counterintuitively bigger is better there as Doug Ford describes.
I wouldn’t have discovered all that if I hadn’t socketed that chip and just tried the NE5534A on the off-chance. Makes me wonder if a bipolar emitter-follower wouldn’t be lower noise than the FET version. You don’t normally need the gain of that circuit with piezo discs.
I’ve been looking for a low noise piezo circuit like this for a while and yours seems great. However I’m wondering what’s the right place to put a potentiometer (and its value) to control the volume or gain of this unit.
Also, is the output suitable for connection to the mic jack of a portable digital recorder? I’d like to know what would be the best apprach to make a separate jack for that purpose.
Thanks a lot,
John
For an easy win you could change R7 for a 47k log pot and take the wiper to the output. Alternatively if you want to handle particularly loud contact mic outputs before clipping you reduce the gain by lowering R5, but you can only reduce it to 1 at the lowest with R5=0. All depends on what you are recording
I used it with the mic input of my recorder because I was interested in the faintest sounds and noise performance. Normally you’d be best off with the line in of the digital recorder if you’re tapping or bashing the thing the contact mic’s attached to, you get lots of signal out of these things!
Would I not just use a 100Ohm log pot?
And can I put in a 10K pot at R5 for variable gain?
Thanks in advance
You can make R5 a pot, but remember the NE5534 isn’t unity gain stable. can’t recall if it was min gain=3 or 5, look it up in the data sheet, and pad out the pot with a series 2k2 if it’s 3 or a 4k7 if it’s 5. You don’t usually need more gain than 11 😉
10k is a better choice than 100k for modern loads. If you are going into a mic input they’re usually ~6k8 to 10k if PiP mics, so with a 100k pot all the action will happen at the max vol end regardless of the law of the pot, because the load impedance will dominate until you get to within 10k of the end. You could always use a lin pot 4x the load impedance and take advantage of this effect to pseudo-logarithmicize the effect. If you’re going into a line input they are often ~50k so you’d get away with 100k. On a mic input you are much better off controlling the mic level with the recorder, else you will get excess noise by padding down the output and boosting it again.
The 100k R7 is there to reduce the thunk of C4 discharging Vcc/2 into the mic input if you plug it in after switch on. It’s not necessary to make it work, but it’s more gracious living, bringing the output to 0V a few seconds after switch on. You can dispense with it if there is a pot there.
Thanks so much for your advice. I’ll be running it into Hi-Z instrument/line level inputs. Basically things like instrument peddles and amps.
Basically I wanted variable gain as I’m not sure exactly how much gain is needed for the pickup and how it’s installed and don’t want to clip, nor do I really want to play swap the component.
I want variable volume just because not all live environments have good control over their levels. (Some cheap desks have fixed gain, sure it will get all noisy but these are the joys of cheap music equipment that some venues buy)
Basically I’m replacing a very cheep and nasty buffer/amp on a silent cello. Just to clarify my goals and elicit any extra insight you might have.
For R5 and after reading the spec sheet it looks like 5-6 is the min stable, so I was going to use a 5k pot and a 4.7k resistor. Giving a min of 5.7K and a max of 10.7k (11 seems too high based on my rough testing)
Right, I wasn’t 100% clear on the purpose of R7. So Just dropping a 47K log pot in and using the wiper for the out, should get me both of the things I need.
I’ll probably add a standard tone (Bass/Treble) adjustment after the output but before R7. It’s all passive so it should be pretty straight forward, I believe.
Sorry 100K ohm log pot.
Sorry for the misunderstanding
Hello Richard, I want to build this version by myself, and I wondered, if it works with ‘usual’ piezos and line-ins? So for example, do you think it would work for example with a piezo type 717770, like I have one? I am not sure because I thought that normally piezos have pretty high impedancies, while this one has a impedancy of only 600 ohms. Also, would like to plug in the preamp into normal line-in inputs, for example a ax-390 amplifier or a usual pc-mic-input. Do you know if the circuit works with my equipment and if no, with which modifications?
Something expecting a load impedance of 600 ohms is either not a piezo pickup or it has its own signal conditioning circuitry inside. I’m not familiar with the device number you cite. If it is this conrad device https://www.conrad.nl/p/miniatuurluidspreker-spanning-30-v-717770-1-stuks-717770 then yes, you can use it backwards as a piezo pickup . The specs indicate it is a 50nF capacitance, though resonant at 1kHz so the mechanical impedance of the unit will distort this. You may not perform well at high frequencies above 1k because of that mechanical resonance, but loading the element with a high impedance will preserve low frequencies as described here. Simply substitute 50nF for the values I used here
http://www.richardmudhar.com/using-piezo-contact-mics-right/
you will start to roll off at about 500Hz with a typical 7k plug-in-power mic input impedance
Do you have better resolution schematic for this? I can’t read values here 🙁
It’s readable on a desktop but probably not on a mobile phone
From left to right:
R1 = 1Meg
R2,R3 = 56k
c2=10uF 16V
D1,D2=1N4148
R4=1k
C1,C3=47uF 16V
U1=NE5534, pin 2 is wired to junction R4,R5, pin 3 to junction of D1,D2 and R1, pin 6 to jct R5 and R6
R5=10k
R6=470
C2=10u 16V
R7=100k
,Hi Richard,
I read your article about the amplifier. Very interesting and good explained.
However, it is very hard for me to read the values of the parts because it is so small.
In your post dated May 4, 2018 at 8:09 am you named the values; however, I cannot find C4 in the diagram because I cannot read it.
Can you help me out here.
Thank you
Best regards
Harald
It’s a fair cop, I’ve replaced the old Orcad schematic with a Kicad schematic since in my comment of May I called it C2 again 😉 So here you go – everybody should be able to read this OK now
I’ll appreciate this community’s help with this challenging project…..aka, masochistic exercise in futility.
Given this blog focuses on piezo elements I thought it wouldn’t be too far fetched to ask whether I can apply this eq approach to a vintage crystal mic.
I intend to use this dual crystal mic on AM.
Yes, the principle is the same. Vintage crystal mikes were designed to run into a vacuum tube input which were of the order of 1Meg to 5Meg. Their high output was an advantage because you don’t need so much gain from the rig which was expensive in tube days. You really don’t need the opamp version for a close-talked crystal mic, use the FET version instead. You can dispense with the diodes which could give you RF feedback issues in ham radio service.
Some modern rigs provide bias for electret mics which is basically the drain resistor of the low-cost FET version already present but in the rig. Something you should note is that some vintage mikes used Rochelle Salts in the piezo element, this substance is hygroscopic and can deliquesce in extreme cases, which will result in extreme loss of sensitivity.
Thx for this page Richard,
i have a Menorm Piezo-Ceramic Pickup and was wondering if you knew if this would be a suitable circuit for it? I plan on using it on my double bass.
Andy
Yep, provided the pickup is piezo-ceramic. Nearly everything the Big G shows me about Menorm is magnetic pickups, so probably worth a double-check.
Thx Richard, i didnt give you much info to work with but my wife has had it since new in its original packaging. Appreciate the reply.
I was wondering, as i waited for my parts to arrive, and as i attempt to teach myself about this IC whether it might have any advantages to use two 9volts batts to give a 9/0/-9 volts supply? If this might change anything or just be overkill?
Thx
Hi Richard,
I’d like to know which kind of capacitor I should use :
– Tantalum electrolytic capacitors
– Aluminum electrolytic capacitors
Does it matter anyway ?
Sorry I newbee in electronics.
Thx for youe answer.
You may use either. You will tend to find aluminium electrolytics cheaper. You need to make sure you get them the right way round, and you need to make sure the voltage is as specified or greater. For 16V electrolytics you can use 25V or 50V, but not 6.3V, other than C2. There’s nothing wrong in using 16 or 25V capacitors for C2, they will just be physically bigger
Thx you so much for your quit answer.
I can start now building my amplifier.
I’d like to use it with my home made piezo hydrophone to listen for humpback whale vocalizations in New Caledonia where I live. It’s still mating season.
Hi Can this circuit be run with 5V?
No. Not a hope. You tend to lose about 3V off each rail for max output swing. There’s not much left of 5V after that.
Analogue is different from digital, supplies tend to be higher for performance. You can get 5V single supply opamps, but everything is a compromise, they tend to be noisy. For example OPA341 en is 25nV/root Hz compared to 3.5nV/root Hz for the NE5534.
Most consumer 5V supplies are noisy as hell, as they’re designed to run digital gizmos that don’t care. There’s some protection against power supply noise in this, but probably not enough for a cheap Chinese wall wart
Hi i was wondering if you had any pictures of a complete build of this – would you recommend housing it with the actual piezo like you did with the FET buffer? Planning on building one of these in the coming weeks.
also just want to confirm – the diode is a 1n414B not a 1n4148 right?
thank you so much!
Diode is a 1N4148 though any general-purpose signal diode will do.
The wire from the piezo to the amplifier is a high impedance so will tend to pick up hum, and probably trash like mobile phone rat-a-tat-tat interference. In practice if you use audio screened cable up to 50cm it’s OK. Too long and the cable capacitance will start to shunt the signal and you will lose gain. But running power is a pain, though I used twin overall screened cable with power on the red core and return signal on the blue, with overall screen as ground.
Would this circuit be capable of producing any sort of musical sounding distortion? I was thinking of making a four channel mixer to blend the signals of multiple piezo contacts into a stereo output. the idea is to have four of these 5534 preamps as inputs going into some filtering then a summing amp. I wanted the option of having some clipping in the circuit and wasnt sure if I would need another gain stage or if i could simply alter some resistor values.
If you want clipping shunt a couple of back to back diodes in series with 1k across R5. If you want more even-harmonic distortion then make one diode in series with 1k and the other in series with 4k7. That will give you a gain reduction to about 2 above 0.6V on one half-cycle and about 4 on the other half-cycle. Even-harmonic distortion (asymmetric waveforms) is said to be less unmusical than odd-harmonic distortion. How much distortion you get is a function of how sensitive your piezos are.
It’s a pretty rum way of achieving sonic colouration IMO, but each to their own 😉
Hi! I bought a cheap (10$) throat microphone that comes with a builtin preamp for mobile phone (meaning that you connect it to 1/8″ and I think the phone is sending around 2V to power to preamp).
I want to use this microphone with a soundcard (1/4″ audio connector). The piezo inside is small (around 2 cm).
My question is: is your circuit a good candidate to replace the builtin preamp (very basic transistor circuit) and drive a line-level soundcard. I guess I need a lot of gain.
You can get more gain by raising R5. Gain is (R5+R4)/R4
Piezo elements have a higher output level than many other transducers as long as they are terminated in a high enough impedance. You may not need as much gain as you expect.
What is the function of C1 in this? I built this on a breadboard and haven’t been able to get it working with C1 in place – It completely cuts the output signal
also, can you recommend a particular type of piezo element to work with this circuit? I’m about to place a digikey order
(the circuit seems to be functioning fine otherwise – I’m taking a 20 mvpp signal up to about 210mvpp)
C1 AC grounds the lower end of R1. If you leave it out gain drops. If it is short (or possibly the wrong way round) the U1 will try and amplify 1/2 VCC by 11. Since the power rail doesn’t go up to 5.5*VCC the opamp will saturate at the positive rail and you will get no output.
Hi Richard,
That’s amazing work. Thank you. I wonder how this circuit would work with a ring piezo, like these ones http://m.piezoelements.com/piezo-ceramic/piezo-ring/piezo-rings-for-sale.html? These rings would be encased in a metallic casing and used as contact microphones.
Best Regards,
Mar
Hi,
I build few versions of this schematics and noticed I can receive radio stations with it. In next version I added 1 nF MKT capacitor parallel to feedback resistor making low pass filter at around 15 kHz and the problem is not noticeable.
Thank You for great job!
I am intrigued. Possibly you are in an exceptionally high RF environment, but normally the massive capacitance of the piezo and using shielded cable on the input should reduce ingress.
The NE5534 will oscillate at low gains but is stable at the gain of about 11 here. C3 should be close to the opamp as per usual good practice.
Thank you for your response.
It might have been low quality cable (I had used cheapest microphone cable available), but similarly I receive radio when no cable is attached (I did not use capacitor between inputs nor the inputs were shorted so my implementations were poor)
I wonder how to make gain variable. Won’t the noise of potentiometer placed in feedback loop instead of R5 be amplified to unacceptable level? Is it better to place the potentiometer in place of R4 resistor?
In a theoretical way making R4 lower down to about a hundred ohm will reduce your high frequency noise a tad (roughly at frequencies where the combined impedance of the input capacitance across the main input is down to a similar level). In practice it’s not going to make any difference. You are not usually short of level with a piezo 😉
Making R4 lower means you have to increase C1 so you don’t start rolling off low frequencies. It’s easier to put the pot in R5, but if you want to tweak R4 instead go right ahead.
Hello, just wondering is metal box really better for this? I have thought wooden box is best so it does not take connection from the circuit?
Naturally you have to insulate the board from the box, typically done with plastic M3 pillars, though I have seen bubble wrap and double-sided foam tape used. In the immortal words of Horowitz and Hill’s The Art of Electronics that sort of thing is ‘bush league’.
Then ground the box at one point only, to the ground of the input socket. You get this for free if you use a phono socket. Ideally you’d use an isolated socket for the output, but this is not a critical circuit and levels aren’t dreadfully low, so a normal grounded phono output socket will be just fine too.
Hi,
Will rail to rail power supply [-4.5, +4.5] make a difference?
If you have mains power available go for +/- 12 – the 5534 is happy with up to +/- 18 😉 But yes, it would work with +/- 5V. You would be wise to redesign for balanced supplies as you could get rid of C1 and R3 and R2 taking advantage of the ground 0V.
Would it be possible to swap the opamp for a THS4521 to receive a balanced output signal?
(sorry if this is a stupid question, I’m very new to this)
> Would it be possible to swap the opamp for a THS4521 to receive a balanced output signal?
No. If you wire the THS4521 device as recommended in the TI datasheet you’ll have an input impedance of 2k differential, the piezo is going to really hate that and your ears will bleed with the tinniness. You could raise Zin by raising the resistor values but you’re pushing your luck raising it to 100k all round – I’ve never used that part but you’ll probably get odd HF stability issues and the noise will go up.
Why do you want a balanced input? Is it just that you have balanced ins on your mixer? This is battery powered and not grounded*, you won’t get a ground loop. RANE Audio are your friend with wiring scheme #17 to wire to a balanced input.
If you have a really long cable going through a lighting grid and dimmer packs you could use a DI box at the send end to get a balanced output, in which case first try putting your piezo straight into the high impedance unbalanced guitar input of the DI box. This Behringer DI100 has an input impedance of 250k f’rinstance. That’ll go down to about 45Hz (with the type of piezo I measured at 14nF source impedance, YMMV) which is good enough for most people. But if you want to go lower then you need to put the output of this amp into your DI box.
Most people will be fine using RANE hookup #17 if the piezo and mixer are in the same room.
* as long as you don’t ground the piezo on some lump of metal already wired to grond
Wow, thank you so much for your quick and thorough reply. I am learning a lot and it seems that balanced transmission is not neccessary in this case. My thinking was balanced = better and I wasn’t sure about the output gain of this circuit. The rane page is very helpful as well. Thanks again!
Hi, thank you so much for this schematic, I built the PCB (5cm x 5cm), it works BUT I have a lot of hum (no radio but constant hum). When touching the ground (for example the 1/4″ audio out cable) the hum is still there but much lower. Any idea how to get rid of the hum? I can send an audio extract.
Final design:
https://imgur.com/a/IUPjBS8
I think I found an error in the schematic, the gndref (4.5v) is not connected at the voltage divider (56k) & the op-amp ground should be connected at the battery’s ground. Something like:
https://imgur.com/a/qj8DWO2
@Pat: You are not correct in the redrafted version. Do not connect the 4.5V point to ground. For AC signals C2 decouples the 4.5V bias to the op-amp -ve rail that is used as system ground. C1 reduced the gain of the opamps at DC to 1, which is why the 4.5V bias will bias the output midway at 4.5V too (DC gain of 1 applied to the input of 4.5V =1). TI have some good stuff on single supply operation of opamps, of which this is an example.
If you connect a source like a dynamic microphone to the input that passes DC it will screw up the biasing, you would solve that with a 0.1uF capacitor in series where Y1 joins R1 and the + opamp input. I have taken advantage of the fact that the piezo disks look like a voltage source in series with a capacitor and pass no DC. Your picture shows what looks to me like a piezo throat mic, I have one of these and this works fine with that, inasmuch as a throat mic ever works right 😉
If you have hum that reduces when you touch the ground of the circuit I would test for ground continuity from circuit ground (the -ve pole of the battery, opamp -ve) through the cable to the recorder. Look with a multimeter between the cable ground where it goes into your recorder and the -ve pole of the battery. It should read less than 5 ohms.
I have built this circuit as drawn, and so have others. I think enough people have made these to verify the design and the transcription of the circuit by now.
Chase cable and connector continuity. HTH
Hello pat
I was wondering if i could get the pcb design if possible?
Thanks for sharing your work. I’m in the process of working out how I’ll put it together on perfboard, and I’m a bit confused by the diodes. It looks like the anode of D2 is connected to the negative rail, and the cathode of D1 to the positive, and then I’m not at all sure about what’s happening in between. I am going to proceed under the assumption that the schematic is correct, but I would like to understand it better if you have the time to explain.
> It looks like the anode of D2 is connected to the negative rail, and the cathode of D1 to the positive, and then I’m not at all sure about what’s happening in between.
Yep. The job of these diodes is to short out any high voltages from the piezo unit away from the opamp pin 3 into the power supply if the input goes below ground or higher than the positive rail. These sensors are piezoelectric and the input is a high impedance. You get quite high voltages from the sensor if you drop it or let the magnet snap it onto a lump of steel. The diodes are not necessary for circuit operation, but the service life of the opamp will be longer with these rather than relying on the internal ESD protection diodes.
Hello Richard, I really appreciate the detailed schematic overview!
Do you think you could explain to me the purpose of R6 and R7? Is it that op-amp expects a load at the output? Is this common practice for preamplifiers?
R6 stands off the capacitive load of a long cable to the recorder. Most of the time it’s not necessary, particularly with an opamp that’s not set to a low gain – this one is set to a reasonable gain level of ~11. But some combinations of cable capacitance and opamp can lead to oscillation. Having some stand-off of opamp outputs is good practice. See Analog Devices on the subject.
R7 is there to bleed away any charge on C4 when you power it up, to reduce the level of any DC thunk you get as you plug it into a low-Z input, which would otherwise need to sink the 4.5V charge on that cap on plugging in. Again, you can live without it, but life is a little nicer without needlessly large bangs.
Thank you so much for the explanation Richard! and much appreciative for the linked article
Hello Richard, Thanks for sharing your work.
Is this schematic work at 1.8V~3.0V, a low-voltage power supply? My project needs this piezo contact mic but the power budget is only 3V and 1.8V.
Not a hope. Use a battery. See previous response to the fellow trying to use 5V
3V is not enough to make decent audio anything. There’s a reason analog mixers thrashed their opamps on +/- 18V supplies. It’s called headroom. If you’re going to try to get away with digital supplies you’re not going to get decent audio performance. Are you aware that you could easily get 5V p-p out of a piezo disk? How’s your 3V power rail-to-rail opamp going to reproduce that?
You could think laterally. Your puny 3V PSU isn’t ever going to handle max output from one of these bad guys. You probably have an input impedance of ~10k if it’s designed for an electret mic. Whack a series resistance of 330k in front. Not only will this pad down your signal by about 30 times but it will raise your input impedance to enough to keep your bass.
Sure, it will knock your noise figure for six, but if it’s for something you bash like a drum pickup then you’ll still have plenty of signal. For the cost of a 330k resistor it has to be worth a go 😉
Update: I tried this. It works surprisingly well IMO. Certainly if you’re interfacing this to a computer to trigger something rather than listening to the output, the series resistor is a low-cost easy win. If your contact mic is attached to something someone is bashing, you’ll have enough SNR for listening to the audio.
Headphone amplifier.
Hey Richard. This is a seriously awesome blog mate. I am going to have a go at building the low noise version for the piezo hydrophone head I built. I came across your preamp after endless hours of surfing. I am new to the world of electronics and built a headphone amp from another internet source using the LM386 IC.’s in a plastic box. I found that the circuit I built had a lot of hum and hiss. I dealt with the hum by shortening the connecting wires and twisting pairs but the hiss is another issue entirely. I don’t have an oscilloscope or any other test / analysis equipment other than a basic multi-meter. I was hoping that you could provide a couple of links to some low noise headphone amp build projects that you have had a look at and thought… hmmm. that looks pretty good, or even better that you have built yourself. On another subject, what do you think causes the chomping sound, I have read musings on territorial signalling of fish, as well as crushing food prior to ingestion? I have also read about signalling using noises generated in the swim bladder. I have yet to find any specific A.V. recordings though.
Hi Tony,
Listen to this from soundscapes of Stirling – it has the same sort of thing. Daan Hendrinks says some of the sounds are plants offgassing
Personally I tend to use my recorders as a headphone amp, flip the LS-10/LS-14 whatever into record-pause and I am monitoring, ready to record. The LM386 is noted for noise, if you can live with losing some treble – either because you are recording a piezo disc which has no useful response above the self-resonance of about 4k or it’s a voice comms application then you can get an easy win with this mod – explained in the comms usage here
If you have adequate volume but the resting hiss at low volume is the issue then look at the gain distribution – lower the gain of the LM386, compensate in the earlier parts of your system. If you are using personal audio headphones which are 32 ohm then you can consider 33 up to 100 ohm resistors is series with the output. With 33R you lose 6dB of hiss, but the downside is you also lose 6dB of signal out before clipping.
Failing that, you can get a decent amount of grunt out of the good old NE5534, it’s good for about 15mA output (38mA s/c limit) TI’s Amp up your cans indicates you need 2Vrms at 47mA for 100mW for a 110dB SPL. That’s loud, if you’re prepared to live with a tenth of that a 5534/5532 will do, just stand off the output with a series 22 ohm resistor so it doesn’t hoot at about 1MHz with the headphone inductance. Downside is you need a 9V battery or more.
But I wouldn’t muck around with that. Use your recorder – the manufacturer has a SMD production line, access to whatever the latest low voltage rail to rail output power ICs are and you’ve already got it.
You can even use the recorder to power your hydrophone. I really can’t emphasise enough that for most real-life field recording applications the FET version is more than good enough, easy to optimise/debug with just a multimeter and because it’s basically what’s inside an electret microphone capsule, just with a piezo disk instead of an electret capsule, your recorder will be able to power it like an electret mic with plug in power.
Maybe I need to get out and get some hydrophone recordings using the FET version to convince people you don’t need the complication of this version IRL. There’s nothing wrong with using this one, but it’s more work. I’d be surprised if you could actually hear the difference between this and the FET version once you’re out there in the real noisy aquatic field recording environment.
Thanks for all the valuable information Richard,
I am about to order from a local retailer and wanted to ask you about the piezo selection before i do:
What is the ideal resonant frequency and capacitance of the piezo should be? Also is going with the generic piezo (which is a buzzer piezo, ceramic ones i guess) alright or should I pick a different type ?
Thanks in advance,
Ekin
> What is the ideal resonant frequency and capacitance of the piezo should be? Also is going with the generic piezo
You’re overthinking this IMO. Theoretically the resonant frequency should be above the audio bad, but that’s not what you are going to find. The sensitivity is increased by the resonance in the audio band – back in the day telephone mics used this, and you can see the principle in this 1982 paper on making a telephone mic from PVDF film to boost the sensitivity. Then cheap electret capsule manufacture came along and ate their lunch for that application.
These things are cheap. Get a few and play around. Anecdotally, some people say large diameter piezos give you better low frequencies, I have not personally experienced this, but perhaps it is a input impedance issue, since larger piezos have greater capacitance as a general rule. They tend to have a lower resonant frequency too, which is not necessarily a good thing.
The biggest issue with low frequencies once you have solved the high input impedance is getting a close enough coupling between the solid piezo and the solid lump of something you are trying to contact. Any air between lowers LF response – put a contact mic on a computer and listen as you press harder on the back to improve contact and you will hear the LF come up as you improve the contact area through a solid medium.
Hi Richard,
Thanks for posting this, I am trying to simulate this circuit in LTspice to better understand it but am having some trouble simulating the piezo (currently using a voltage source in series with a capacitor but getting some weird results when doing a transient simulation). Do you have any insights on how to model the piezo or is this one better done at the bench?
Thanks!
Simulating the motional components of the piezo at resonance can be theoretically done in the same way as for a crystal.
https://www.pspice.com/resources/application-notes/modelling-quartz-crystals
but you and I haven’t got the specialised test gear to characterise the motional components. For the bass rolloff I measured the capacitance of the element directly, using a frequency below resonance. Modelling the piezo as a voltage source in series with that capacitance is a rule of thumb, but it works well enough below resonance.
I’m a bench guy for this sort of thing. You can simulate the amplifier on its own easy enough, leave out D1 and D2 which are protection diodes and shouldn’t be doing anything under normal operation. It’s a pretty standard noninverting config with a gain (R5+R4)/R4 where 1k << Z C1.
Thanks Richard for the insight and the link. I was able to get the amplifier simulating but the capacitor in series with the voltage source continues to cause some weird outputs during transient simulations (AC analysis and Noise look great). Seems like the bench is the way to go from here.
Thanks again!
> Seems like the bench is the way to go from here.
hehe – a couple of days on SPICE can save you an hour on the bench, to paraphrase Frank Westheimer.
There’s nothing that special going on here, this is an audio frequency amplifier with a modest gain, using an opamp that every mixing desk in the world for the last 30 years has been rammed full of.
The NE5534 can get stroppy if you try and run it at unity gain, but the datasheet says it’s stable above gains of 5 ISTR and R6 is there to stand off the cable capacitance on the output to reduce any tendency for it to hoot and take off. As show the AF gain is 11, safely away from a gain of < 5.
Hi Richard,
Thank you so much for this helpful post.
I’m building 6 of these units for a piezoelectric piano.
I was wondering what I should do to couple the 6 units to get a single output, suitable for a guitar amp. Can I just connect all outputs together or should I change R7 value?
Thank you, Cheers
> what I should do to couple the 6 units to get a single output, suitable for a guitar amp. Can I just connect all outputs together or should I change R7 value?
Personally I’d change R6 for a 10K and put the outputs (the non op-amp ends) in parallel, though I haven’t tried it 😉 You can get away with only one o/p cap. You’ll have a Z out of 10k/6 which is OK, you will lose about 80% of any individual output in what is effectively a passive mixing bus but I imagine you won’t be short of signal.
If only one input is active at any time you may start to find noise builds up from pickup on the unused channels. There you’d take the individual outputs to mixer channel strips and EQ/noise gate them.
R7 is only there to bleed away to charge on the output cap, so if you spark it up and only then plug it into your amp you don’t get a disturbing thunk as your amp discharges 4.5V to ground through its input resistance. Good practice is don’t plug things together live, but R7 is there to be kinder to those of a nervous disposition if you forget.
Thank you so much for your answer! I will definitely try it that way.
Hello Richard, thanks a lot for sharing this valuable information. I’m going to test your circuit as soon as I receive the piezos.
I was wondering if there’s any chance to use phantom power to supply the circuit. I see the NE accepts max 20v so there’s the need to reduce it I guess. Any suggestions would be more than welcome.
Thanks in advance,
Emanuele
> there’s any chance to use phantom power to supply the circuit.
Not easily. You’re better off with the Alex Rice FET amplifier – Zach Poff has the details if that’s important to you. You take some hit on ultimate noise performance but hey, this is a piezo disk, it’s not high fidelity in the first place and it’s prepared to chuck out masses of signal. I measured 60V coming out of one of these bad guys with only minor provocation so you need a really good reason to think you need great low-noise performance.
Note that piezo disks are not generally balanced devices. Every one that I have seen really wants the big brass disk part to be grounded else you will pick up mahoosive amounts of hum. I’ve never tried the AR circuit though I did buy a nice little board from somewhere on ebay to try it, I haven’t got round to it yet. Trying to make a physical rig to balance a piezo disk and get it in contact with the surface does my head in. Clearly Zach Poff got it to work, but as the man says
Personally if I wanted to avoid the battery I figure plug-in-power with the FET version is really hard to beat if you’re prepared to select FETs on test to work with the low PiP voltage. It’s usually 3 to 5V or thereabouts fed via 6k8, though if you don’t need stereo then paralleling L and R gives you about a 3k drain resistor and double the current. That way your big brass disk part stays grounded.
Hello Richard,
Im looking for a very compact version of your low-noise circuit with 3V supply and smd parts.
Do you have any recommendations which amplifier I can choose?
Bests
Maximilian
See above. There’s a very good reason analogue mixers use +/- 18V rails…
The FET version can run off 3V, if you select the devices. An emitter follower can give you the impedance transformation and you can bias to work off 3V. Just make sure you have a good answer to where the 60V from the sensor is going if you drop the damn thing, because the base-emitter junction of the transistor won’t appreciate that very much at all.
You can try using a 3V capable rail to rail opamp. A parametric search at your favourite supplier should give you a start. Just make sure your supply is a low enough impedance to sink that 60V pulse, and that your opamp doesn’t mind the input going 0.7V above the + supply rail and -0.7 below the -ve rail.
Hello Richard,
Piezo Elektr. capaciteit ≈ 20000pf Very quiet ，Piezo Elektr. capaciteit ≈ 500pf high noise ，why？（sorry if this is a stupid question, I’m very new to this）
Hard to say with the information you’ve given. 500pF sounds exceptionally low for a piezo disk, I would first search for a broken wire. I have never come across one that measures this low.
What are your piezo components? How are you measuring the capacitance? In particular if your meter is measuring at a frequency above the motional resonance of the piezo (typ 2kHz to 4kHz) you may get odd results.
If you run the amplifier with an open input then you do get a higher noise level, because the noise voltage at the amplifier input is to some extent shunted by the input impedance, which is of the order of a few k at mid-audio frequencies. The noise voltage is ~ SQRT(4kTBR) where k is Boltzmann’s constant, T is absolute temperature and B is the system bandwidth.
Hence the common experience of an open audio input being noisier than when you plug a mic in. And why combined with an observed exceptionally low input capacitance that you could easily pick up with a couple of meters of audio cable I would suspect a break in the connection at the far end.
In the stereo version of this if I am using it with a single piezo disk I have to parallel the inputs to stop the racket in the RH side from the open input rather than just wire the disk tip and sleeve. I hadn’t spotted that when I first tried it in the field but I quickly learned the error of my ways.
Hi Richard,
thanks for that awesome little circuit! I set up a PCB layout (about 5 cm * 5 cm or 2 in. * 2 in.) for a dual/stereo version and added decoupling capacitors for the ICs while removing the battery (BT1) part and using a wall wart PSU instead. I even skipped the switch, but that that could easily be added with flying wires if desired.
Also, you can optionally replace R7 with a 100k pot (wiper to audio out jack) for volume control. I did that for my DIY Kalimba equipped with two piezo contact mics. Works 😉
If you’re interested I will gladly send you the gerber files – and yes, your name is on the PCB, and it’s strictly non-commercial. I do not intend to make one single penny or cent from your work.
Cheers, Bodo
Would you be willing to send me those gerber files?
Sure, I just need to know where to send them!
Richard,
This is very helpful! I build your preamp per your specs for a DIY contact mic for underground outdoor pipe leak detection. I feed the output of your preamp into a headphone amplifier and use std audio headphones for listening. It works fine but I believe I need more gain… What is the ~max gain (and value of R5) that this circuit can manage?
I’m using a 35mm disk piezo with a backing of hot melt glue ~3/8″ thick applied directly to the piezo side of the disk to increase the native disk resonance freq. The hot melt seems to greatly increase the damping of the piezo but frequencies above ~2kHz do come through fine. Specs for piezo claimed are Resonant Frequency 1.8±0.3kHz, Resonant Impedance 300 ohms max, Capacitance 50,000pF ±30%.
I put in a 100k resistor I had on hand in for R5 and the amp seemed to work very well at much higher gain (leak can be heard that was not noticeable with R5 = 10k) for a minute or two but then changes to sounding as though the gain had decreased a lot. This seems to be repeatable, in that a day later the same behavior was experienced. The input noise remains very low at this gain while it is working. I have no signal generator to test the circuit and have not pulled out my scope to look at the behavior yet. I may try to use a phone based audio tone generator app with air coupling to the piezo as an input to get some type of characterization.
Did I get too gain greedy with 100k for R5 and throw something else out of whack? I believe the NE5534 can handle a gain of up to ~200+ across the audio freq band easily.
Thanks,
HardwareJ
I can believe hot melt will improve the damping that could well improve the flatness of frequency response at the cost of sensitivity. But extra mass loading usually lowers resonance. I’m not a materials scientist though, perhaps the difference in compliance between the softish HMG and the extremely stiff ceramic gives an impedance mismatch that means the ceramic resonance isn’t lowered so much. Some people use plasti-dip for waterproofing and damping, along the same lines of thinking. Be that as it may-
> but then changes to sounding as though the gain had decreased a lot.
Get your scope on it. I suspect you have oscillation. Remember the input is high impedance noninverting, so any leakage output to input will give you positive feedback, with about 15nF to ground to give you a nice phase shift so there should be some frequency it can hoot at.
The 5534 will do a gain of 200 OK but most audio sources are low-z – hundreds of ohms to about 20k tops. You’re pushing your luck to get that much gain in a single stage at hi-Z and that capacitive load. Lower the gain and push your headphone amp a bit harder – a bog-standard LM386 will give you a gain of 200. Alternatively if you don’t care about low frequencies drop the input impedance, which will make it less sensitive to stray capacitive coupling in to out.
Dunno if your pipe leak is air – if so then use ultrasonic piezo detectors. I use a bat detector to locate punctures in a bicycle inner tube, works a treat. If it’s in liquid then that’s not so good 😉 These things are tuned to 40kHz, so you have to mix them with about 35kHz to get them to easy human hearing.
If you want liquid sensing and higher frequencies than 2-4kHz, then China will sell you piezos designed for ultrasonic fog machine replacement elements in water which might be worth a go. The sensor sensitivity craters above the resonance, so having a higher resonance will give you a lot of win. These are designed for about 40k in water. If your pipe is carrying a gas then run with the 40kHz ultrasonic sensors designed for air – they are used in liquid level sensors and robotics object detection, ebay and robotics places carry them.
Richard,
Got the scope out, took apart sensor and probed but saw no oscillation. Still not clear on what was going on but could be additional capacitive coupling of circuit to its metal container when assembled. I lowered the gain resistor R5 with a pot set to 50k and circuit seems to work as expected now. The leak is an underground water leak =( of ~0.1 Gallon a minute. I can now hear water flow in the main input underground when I start and stop a flow of that magnitude with a valve so will begin the hunt.
I believe the resonance spec refers to the fundamental mode of oscillation when simply supported of the circular plate/piezo when a piezo disk is used as a speaker or buzzer. By massively increasing the bending moment of inertia of the disk with additional thickness of glue in the bending beam it is stiffened and the fundamental resonance frequency is increased, at least that is what I am going with.
Simply, the stiffness is a function of the cube of the thickness of the plate, oscillation is a function of the stiffness and the mass. The system is a spring and a weight, but the spring is a metal plate and I have stiffened (and damped) the spring significantly more than I have increased the mass. Any time you glue the plate to any stiffening element it will shift the fundamental oscillation freq. higher. (I don’t know why EE’s get all confused about this aspect of piezo’s. They are a mechanical system and the properties change if you change the mechanical layout. This includes how the disk is mounted. )
I will have to check out ultrasonic piezos and downshifting the frequency. Could make a bat listening device…
Thanks for your Blog!
HardwareJ
So is your piezo in the water, or is it glued to the wall of the metal container in the water? I’ve had good results with these as a contact mic gluing them to magnets. People make hydrophones out of bog-standard piezo discs by dipping the piezo in plasti-dip. You get pretty good coupling to the water doing that, though you also get a lot of LF signal from the piezo flexing on the cable in the water flow. But since you are after high frequencies you could get away with that – shunt the input to make a lower impedance.
Hi Richard,
Thank you very much for such detailed article. I’m a composer who just started learning circuits like this to create my own tools, and even though there are many examples on the internet for audio circuits, this is the most detailed and professional I’ve seen.
I ‘m trying to make this one on my breadboard, and I’m pretty sure I got everything in the place where they’re supposed to be, but for some reason when I hook this up to my audio interface, I can only record “pop” sound, and that’s only when I tap the piezo itself. If I attach this to things like a trash can (which is how I’d use for actual recording) and tap on the surface, nothing will happen.
Would you be able to what the problem is with mine? I apologize for a newbie question and sorry if my explanation isn’t detailed enough to locate the exact problem.
If you want to build electronics then you absolutely need a digital multimeter (DVM or DMM) which for example are cheap these days. With the DVM measure the voltage on pin 7 and pin 6 on the IC relative to pin 4. Put the negative lead of the DVM on Pin 4
Remove connections to the Piezo, and any output cable.
Pin 7 to pin 4 should be +8 to +9V – the same as battery voltage. If not check your wiring. And battery.
Pin 6 to pin 4 should be half the battery voltage measured above, plus or minus 1V
If this is not the case, take out D1 and D2, which are diodes and easy for newbies to get the wrong way round. They are not necessary for the functioning of the circuit, they protect the IC against damage from large signals. Recheck Pin 6 to Pin 4.
If good then fit one diode back, recheck. If you get it the wrong way round then pin 6 to pin4 will not be half battery voltage. Do the same with the other diode. If taking the diodes out doesn’t bring pin 6 to pin 4 to half battery voltage leave them out for now.
If p6-p4 it’s still not half the voltage, take out C1 and C2 next. These need to be the right way round without them you will get low gain and a higher noise, but first see if pin 6 to pin 4 measures half the battery voltage.
If it’s still not half the battery voltage, measure the junction of R2 and R3 which should also be half the battery voltage, if it isn’t then check that R2=R3.
After that you are out of options, other than carefully double checking, or taking all parts off the breadboard if it is this sort of thing and starting over on a different part of it. You can try swapping the IC but the NE5534 is reasonably robust.
Note that the x on pins 1,5,and 8 means not connected, these pins must NOT be connected to anything else
Wow, thank you so much for telling me these! It’s very useful and this process is a great learning for me. Also I apologize for late reply, I was watching for an email notification that I thought I’d get when I get a reply, but I didn’t.
I measured the voltage difference between pin 7-4 and 6-4 like you told me, and with both diodes on it was 48mv for former and 380mv for latter, which didn’t match with the number you gave me at all.
I double checked the placement and polarity of the diodes and it was all fine, but I realized taking both of them out fixes the issue, so I decided to replace both the diodes and it works perfectly now!
Apparently both the diodes were faulty, since replacing only one of them didn’t work, and only replacing both of them gave me the correct measurements and right amount of signal.
I recorded bowed iron bar fixed on a trash can with this, and it sounds beautiful. No noise, lots of bass. Thank you so much for your help again!
Nao
Hi Richard,
Thank you for your incredibly clear and detailed instructions. It made me understand a huge deal about piezo microphones.
I am trying to build a low-noise preamp for DIY soil microphones – it seems all working in theory, but it doesn’t. I can’t make the circuit work.
I face a similar issue : there is only ‘clicking’ sounds when I hit the piezo itself and only a buzzing high-pitched noise the rest of the time.
With a lot of hope, I tried all your troubleshooting steps, but it didn’t change anything. The volage is correct at R2=R3 (approx. 4V) but it stays around 0.7V between pins 6 and 4.
I tried to change the NE5534 and strangely, the volage between pins 6 and 4 starts around 7V when I connect the new chip, but then decreases continuously to reach 1V within a few minutes. Sounds very strange to me.
I already made the circuit from scratch twice, but will try a third time.
Do you have any clue what might be going on here?
Also, should GND ref be connected to 9V- (i am using a 9VDC power supply)? I did, and the noise intensifies when I disconnect it.
Hi Richard, Thanks for sharing your work. I succeeded with a Piezo contact microphone .But there is no sound when I switch to the Under-Saddle Piezo Pickup . What should I modify?
Thanks for your Blog!
John
Hello Richard,
What is the lowest frequency this circuit can deal with? For my intended application the circuit+piezo needs to be as flat as possible to almost DC. A -3dB point of say 10Hz would already be too high.
Your circuit is interesting, because it does not use outrageous R values ( 100-400Meg) for the input circuit I see in piezo charge-amps.
Thanks,
Eelco
That was a very good post! I’ve been toying with the idea of making an electric cello, and using polymer film piezos in it, because they are noted for flat frequency response. They have more voltage output than ceramics, but have very low capacitance, so I’ll need a very high input impedance, around 10-20 meg. I think I’ll use your schematic as a starting point, although I’d have to change the 1 meg resistor, of course.
The handbook from the manufacturer notes that I might want a «low leakage» opamp with such a high input impedance. I see you are very happy with the NE5334 which has very high input bias current, which would probably therefore be classified as high leakage. Would you happen to know anything about that? I posted a more detailed question on an electronics board about it: https://www.eevblog.com/forum/projects/how-much-opamp-leakage-is-acceptable-for-polymer-film-piezo-preamp/
Anyway, thanks or the schematic and valuable info!
> I see you are very happy with the NE5334 which has very high input bias current, which would probably therefore be classified as high leakage.
The bias current isn’t a big deal in audio applications because they are AC coupled, but if you are going to lift R1 by an order of magnitude then bias current will start to matter, because to a first approximation there’s an error voltage due to the input bias current through R1. If the DC resistance to the opamp inputs is the same this tends to cancel out but that’s not true here, R1 >> R5 (you can ignore R4 for DC conditions because of the series cap) TI says it’s 2uA worst case, so 2uA into 1Mohm is a 2V offset. Into 10Mohm that’s 20V so your opamp will be stuck at one rail.
While I disagree with Kleinstein that the NE5534 is a bad choice for 1Mohm, measurements in the comment above I would agree with him that for your use and impedance levels go to a JFET input opamp, to get the input bias current down. Jfet opamps have a much lower current noise in though a higher voltage noise en. You can see that – Kleinstein’s suggested more modern JFET OPA 172 has an en (called vn in the datasheet) of 7nV/rtHz at 1kHz, for the ancient NE5534A that’s listed at 4.5nV/rtHz at 1kHz. However in is developed across the input resistor, that’s 0.4pA/rtHz in the NE5534A. In this circuit that would give you an apparent 400uV/rtHz (!) across 1Mohm but in practice it is heavily shunted by the high sensor capacitance, which is why in this application the NE5534 is lower noise with the sensor connected.
The lower in of JEFT opamps comes into their own at higher impedances, you have a higher R1 and a lower capacitance sensor by more than an order of magnitude. Go with the OPA172, though it is SMD only, or the OPA134 which is available in DIP which is easier to solder by hand.
You may need to think about low-leakage diodes, or perhaps leave out D1 and D2 altogether. Piezo film normally has a much lower output than piezo ceramic disks and you aren’t likely to drop the cello on a hard surface, so you may not need that protection. A 1N4148 has a reverse leakage of ~25nA @ RTmp
The FET version is worth a go for an easy win, again, perhaps without the diodes. It won’t mind a 22M input resistance in the slightest. Shielded audio cable can easily be 450pF a meter or more, so two meters of cable will shunt half your wanted signal because of the lower sensor capacitance – the FET version can be built in at the send end next to the sensor which gets rid of that problem. I don’t know why everyone reaches for the opamp version – with instrument sensors you are usually really not short of signal 😉 That said, I don’t know about film, when I messed around with PVDF sensors they were a lot lower output. But the FET version isn’t terrible with noise at all.
Good luck!
Thank you.
The reason I went straight to the opamp version is that I thought everything needed an opamp! But after reading this I’ll probably try the FET version first. In either case, the opamp will be build into the cello, a few cm from the sensor.
Regarding the output, I have been doing some back-of-envelope calculations and see that the PVDFs output far more if you stretch them in the length direction or bend them, than if you compress them in the thickness direction. They have gotten 1600 V out of it by stretching, and you can easily, according to the papers, get 70 V by bending it. So output shouldn’t be lacking, when used that way. So I figured I might build it into the bridge and have it be stretched as the bridge rocks back and forth during playing. Experimentation is needed to see if it works. I’ll order a bunch of ceramics as well, in case it doesn’t.
Hi. I’m looking for a good pickup for my 34 string bar.
I’ve noticed there’s two type of piezos. The ceramic disc types like K&K twin and dusty strings 4 piezo, and the. There’s “contact mics” like the c-ducer and ischell, which claim to be better quality than standard piezo discs.
Is this the difference between your two articles or high quality version of the pickup and the cheaper version?
This one being similiar to a contact microphone c-ducer type of pickup, while the cheaper ceramic disc type in your other article is similiar to a K&K twin or dusty strings piezo disc pickup?
It’s purely about the electronics of a high-impedance amp, not particularly about the transducer.
I’m a beginner and I try to understand how exactly the amplifiers work, power amplifiers, preamps. I’m a little confused and the videos I watched so far haven’t helped me much. They say you need a preamp for a microphone but the power amp already have a preamp. What’s the difference? Is it just impedance matching or do the power amps have a minimum level that they work with and the microphones have a low level?
Then there are some who say that piezo can be connected directly as a microphone to a sound card, android phone and now I understand from your article that they do not have the same impedance and that’s why you need a buffer or this circuit. If I want to connect this microphone to a power amplifier ,computer speakers is possible? Can you guide me how I could amplify this microphone to be able to listen in real time and use it with maximum gain? I understand that I can increase the gain by increasing r5 value.
Sorry for too many questions but I see that you have a lot of knowledge in the field and you really know what you are talking about. Thank you for your time and for the generosity with which you help others.
Wikipedia on power amps is a good start 😉
> They say you need a preamp for a microphone but the power amp already have a preamp. What’s the difference?
An all in one karaoke rig or busker’s PA often has both preamp and power amp in one box for convenience. If you are trying to fill a venue your power amps will be separate, because you need a lot more power and take more care to avoid feedback.
A preamplifier normally lets you control volume and can do signal conditioning for oddball sources – the RIAA eq for record decks was a classic case, a microphone which is usually quite low output is another oddball source. If you were to connect a mic to the line input of your hifi you would have to yell very loud to be audible at all. A mixer is a special case of a preamplifier allowing you to combine many sources, often microphones.
This article is about how to match a specific oddball source, a piezo disk that requires an unusually high input impedance to give of its best.
The output of most preamps and mixers is line level of the order of 1V p-p and a input impedance of 10-50kohms, most power amps can take that in and get it up to the nominal output level. The input to most hifi preamps is also line level – the preamp is there to let you choose which source you want to hear, to control volume and in some cases to control tone. You could connect a CD player to a power amp but it would be very loud and you couldn’t turn it down.
If you want to specifically use a piezo disk then find a musician and borrow a DI box and a all-in-one PA to feed the signal into. A DI box will feed amplified computer speakers (ie ones with a mains wall wart and a volume control) OK if you can adapt the connectors. Or try borrowing a small practice guitar amplifier, which should have a reasonably high impedance.
I would not suggest you try and build this, which may end up an exercise in frustration given your current audio experience.
You can connect a piezo disk directly to the mic input of an android phone*, but it will be tinny as hell as described and you’ll hate it. I did develop a circuit for using directly with electret microphone inputs but while it is simpler, it is more fiddly and you would need some basic test gear.
* if you connect it raw, ie without that circuit, may have to shunt the piezo with 1.5kohm or similar to make the phone activate the input. That will, of course, make it even more horribly tinny. In general, phones are jack-of-all-trades and master of none, so avoid unless you want more pain.
Hello Richard,
Very nice contact microphone amp works good.
What would i have to change to get rid of low frequencies all below 250Hz or so.
Thanks for your time.
Mario
Change the value of C1 such that 1/(2*pi*f*C1) = 1k would be a good place to start. Alternatively you could shunt the piezo with a series combination of 47k and a 0.1uF – in the original article that gives that sort of 200Hz-ish rolloff response (the 0.1uF cap is only so the shunt doesn’t wreck the DC biasing). The second method depends on the characteristics of your source piezo disk, experimentation is necessary.
Hi, Richard.
I have been struggling with gain control and amplification.
I am trying to implement this piezo circuit to use it for my ant project, placing piezo mics inside of ant nests and listening to them. So it should have as much gain amplification as possible, with low noise. Also it would be nice to have some sort of lowpass filter effect bc I don’t need low end for getting audio signals of ants.
I would love to have a gain control (for higher amplification) so I replaced R4 and R5 with 10k pot and hooked the wiper to pin2 of NE5534. But it doesn’t really seem amplify the signal but rather making weird distortion on the high end while significantly decreasing the low end, which sounds tinny and crashed. (I don’t know much about electronics, I just ask around ppl online and follow their advice.)
I have read the comments, you and the other guy recommended replacing R7 with a pot. But with 47k or 100k? What would be the difference?
Thank you.
> I replaced R4 and R5 with 10k pot and hooked the wiper to pin2 of NE5534.
You are asking for trouble doing that, you don’t want any more capacitance to ground from pin 2 than you have to. And since the gain is (R4+R5)/R4 when the wiper is at one end and R4=0 then you will demand infinite gain and you have connected a 47u cap C1 from pin 2 to ground, and it’s gonna hate that 😉 The NE 5534 is only specified stable for gains over 5 ISTR, so R5 is gonna have to be at least 4k to keep it within spec
Keep R4 fixed. You get a 3dB rolloff when 1/(2*pi*f*C1)=R4, so sure, lower C1 if you wish to raise that rolloff
> But with 47k or 100k? What would be the difference?
gain = (47k+1k)/1k or it’s (100k+1k)/1k. 100k is about as far as you want to push it. If you don’t want oscillation at the low gain end of the pot then put a 5k6 in series with it, pref at the pin 2 end on the board to minimise stray capacitance seen by pin 2
Follow this with an audio recorder set to manual gain, mic input. You don’t generally want to go mad with the gain of the first stage, because once that overloads you can’t do anything. Use only enough gain to overcome the noise of your audio recorder mic amps, and to do the signal conditioning to match the oddball piezo impedance demands. Control system gain with the audio recorder. The specified system gain of 11 is about right IMO, but you can push it up if you really feel the need.
Unless you have something really weird going on with your piezos, you aren’t usually short of gain. Due to the high impedance, you need to keep water and damp away from them, or encapsulate them with Plasti-Dip. Else you will lose signal, and low frequencies.
Hi, Richard.
First of all, thank you for your reply. It really helped me fixing the circuit.
But the overall volume decrease instead of increase whenever I hook it up with the circuit in any condition.
Also, some piezo elements that has been working fine alone, doesn’t work all of a sudden when I connect it to the circuit which is weird.
I have built 6 of them and they pretty much work the same way.
1) they lower the volume
2) they don’t work with some piezo that has been working fine alone.
Dear Richard,
What does R1 (1M) exactly do?
Does that piezo like DC voltage?
Thanks,
Joe
R1 provides bias to the opamp from the R2/R3 junction at Vbat/2, and mainly defines the AC input impedance. The piezo can be modelled as a voltage source in series with about a 10nF capacitor so no current flows through the piezo.
These things are designed to be driven with 5-10V. They don’t seem to mind the DC. You could perhaps get electromigration if you ran them 24/7/365 in a damp environment, but for a couple of hours of field recording every so often, not so much of a worry.
Hello Richard,
Thanks a lot for your work and the explanation you give. I have to admit I don’t really understand half of it but hey, that’s on me 🙂
I’ve drawn a version of your preamp with a gain adjustment pot in Kicad and designed it to be as small as possible (SMD components) and for direct assembly by JLCPCB.
I’ve received the amps and starting to test them. I must have borked the gain resistor / pot values because the opamp shuts down when I turn it too much. I also have a bit much of background noise, but I need to test with shielding the piezo in conductive foil and use better cables.
Let me know if you’d like me to send you the files, I can also send a couple of prebuilt amps if you want to take them out for a spin.
> I must have borked the gain resistor / pot values because the opamp shuts down when I turn it too much.
You really shouldn’t have problems with this, it is not a particularly troublesome circuit. My original one was on veroboard. To see how bad I could make it I printed off the circuit diagram and constructed this in a particularly messy way on a solderless breadboard
I got away with that dreadful standard of construction, I normally try and do better on one of those breadboards 😉 I built it exactly as drawn other than I uses 47u for C4 as I was out of 10u caps by then. I connected 1.5m of low grade audio coax to the output, open circuit at the other end, to capacitively load the o/p. I used two piezos weighted down together, one as my sensor and the other driven with an audio oscillator to give me a mechanical signal into the sensor.
I was unable to make anything oscillate by putting hands over the wires. I then replaced R5 with a 6k8 in series with a 47k pot. Apart from convincing myself you really don’t need that much gain – the scope trace responded to touching the bench and walking on the floor, it worked and did not oscillate, though as you turned the pot and it made and broke contact it was scratchy, but that pot is probably decades old. So once set it was good.
Mine is prepared to put out 4.7V before clipping (I have a 10x scope probe), the battery is 8.7V under load, it is not new.
I didn’t shield the piezo. Make sure the brass disc is to ground. The crazy nutters that supplied my disk wired red to the brass disc, which gave me hum until I looked closer. You really don’t need to shield the disc in a domestic environment, the brass disc is shield enough.
Hi Richard,
Weird, I subscribed to the email response but got nothing… Thanks for your fast reply though !
I’ll have to investigate more why I lose all signal when turning the pot too far then (the ‘borking’), even though my values for the res (4.7k) and pot (5k) look alright, but looking at my schematic and at your picture, I think I made a tiny mistake : I wired the second leg of the pot to ground…
I have a hard time making it out on your picture, but it looks like you only wired one leg and the wiper (white/blue and white/green wires) on your pot, and left the second leg floating. Can you confirm ? I can kind of see a tiny bit of purple wire on the pot, but it looks like it’s on the second “stage” and not important.
More tests to do !
I tried to shield everything using aluminium foil, even wrapped the disc in some (after protecting the backside with electrical tape) and it got rid of most of the noise, but I’ll try again after removing the path to ground on the pot.
In the meantime here are some pics (not sure I can embed them in a comment)
Yup, in the case of
schematic
reflect for a moment what you are doing wehn the pot is fully clockwise. That connects pin 2 to pin 3, effectively. Pin 2 (also the opamp output) is connected to pin 3 which is connected to…
ground
the opamp ain’t gonna like that one little bit, you’re basically shorting the output to ground (and the DC pedestal the output is on). Cut the wire on the pot pin 3 to ground and you’re off to the races.
I’d also suggest you raise the resistor to the LED, because you’ll be thrashing that battery. Say the LED drops 2V you got 7V across 300R which is 23mA. 10mA is good for an LED, you can usually get away with less.
I don’t know why I wired that pin to ground… It’s such a stupid mistake. Works much much better with the leg cut, fortunately is a pretty easy rework to do 🙂
For the led, I went with a pretty bright white one and used the values in its datasheet to calc the resistor, but I can switch it for a lower value one on a potential second run. Not feeling brave enough to unsolder 0402 components right now 🙂
Thanks Richard,
I just did not understand the exact purpose ot that 1M beyond the biasing… 🙂
No waterproof construction is needed, I’m going to use it for eavesdropping through a wall.
According to my intentions, if needed, I will replace the piezo with a traditional electret microphone (or pickup coil, around holes and audio cables… ).
For this, I build a low-noise input preamp, which can be good for both types of microphones.
This phase I make some “pre-boxing” tests with different circuits.
The task is to select the optimal input impedance and avoid self oscillation with favorable noise level and switchable gain at 5V power supply (power bank).
Do you have any advise for me to choose a circuit?
This one (BJT version) works fine with elektret Q:
http://www.janascard.cz/PDF/Design%20of%20ultra%20low%20noise%20amplifiers.pdf
And this one is seems to be very interesting, I have high hopes for this:
https://www.cryptomuseum.com/covert/bugs/opec/index.htm
This is not suitable for an electret. It will not work. It also won’t work from 5V
Use a standard audio recorder with plug in power set on. You do not need an amplifier for an electret mic capsule with an average audio recorder. The noise figure is set by the first stage which is the FET in the capsule. All halfway decently designed analogue signal processing paths have the noise floor set mainly by the first stage
> I’m going to use it for eavesdropping through a wall.
What is wrong with using a glass in the time honoured tradition 😉
I suggest you’re overthinking this. If you want to bug a room then spending a little time with Google will show you many commercial off-the-shelf products at low prices from the Far East.
>What is wrong with using a glass in the time honoured tradition?
This device is more complex.
Various audio input devices, and audio out also goes to modulate a small built-in VHF TX PA and finally a remote wake-up also needed (by 433 TX-RX set).
Yes, … the goal is not simple. But… that TV transmitter what I built a year ago is already / still works. But this is another story…;)
I can’t buy such a complexity from Far East (and usually exact technical specs. are missing and those devices are not robust.) 🙂
The special audio op amps (NE5534, OP07,OP27,…) that have already arrived will also remain in the drawer until I find an acceptable solution, since HIFI sound quality is not the expectation now.
I just tested a piezo membrane microphone built in 10 minutes, but it would be good to know something about the average output signal expected from the piezo in a through-the-wall contact application and what the gain should be…
Gain = 200 plus an LM386 with another 200x gain seems to be less than enough now.
Unbelievable.
But how much is the maximum that can be handled in practice…?
Something is wrong. Is the piezo still good, or did I overheat that.. .?
So many tiny questions…
I have to walk my own path to find out the answers and I need to continue the development and tests. 🙂
Your blog is excellent to summarize knowledge and derive ideas on this topic. Thank you for your work. I saved these important pages for the future into my collection. 🙂
But maybe I will ask you again. 🙂
Hello Richard, thanks a lot for sharing this valuable information.
I’m amateur guitar pedal builder from japan.
I like the sound of the ne5534 and often use it for my own work.
My knowledge is that in non-inverting amplification, setting a bias resistor (1 MΩ in this case) limits the input impedance to be too large impedance.
According to Richard’s analysis results, it seems to expand the low input impedance of the ne5534.
Is this view correct? I apologize for my poor reading comprehension.
Hi Akinori,
The input impedance is increased with NFB compared to the open-loop case. this link gives a more detailed reasoning. And it was confirmed experimentally
A DC effect you have to watch is the effect of input bias current on mismatched resistors to pins 2 and 3, but it’s not normally an issue for audio applications. TI list the input bias current at 1500000 pA (=1.5uA). On the input side with a 1M resistor R1 that will develop 1.5V offset, against 10k R5 that offset is 0.015V. R4 doesn’t matter due to the series cap. However, the DC gain of the amp is 1. So the output will be shifted by ~ 1.5V from the centre. You could take some of that out by changing R3 a little, but I figure it’s OK. You wouldn’t want to make that 10M, however. TI’s listing is for the maximum input bias current, so this offset is likely to be less in practice.
hello richard
Thank you for your clear answer.
Sorry for the elementary question, but how did you decide on the bias resistor values (two 56k) this time? Would 10k not be a problem?
Hi Richard,
Thank you so much for sharing this circuit. I’ve successfully built it on a breadboard and would like to create a board for it so I can use it for field recordings. My electronics skill level is hobbyist at best. If you could share a layout for it, I would greatly appreciate it.
Cheers,
Emanuel
I built mine on veroboard 😉 There’s nothing that critical in it. I do this from the schematic placing parts as I go along, but you can use fritzing or similar if you want to plot it out beforehand.