High Impedance Piezo Hydrophone Amplification - General Hints?

Tobias;

I think that you have presented your choices very well. One parameter that you did not mention is the upper frequency that you need from your hydrophone/amplifier.

I think that your best choice would be a charge amplifier using a low input bias current, low noise op amp. Depending on your bandwidth requirements, you might consider an OPA129 or an OPA827. Generally,a JFET op amp will be better than a CMOS op amp due to the higher 1/f noise of a CMOS input. Autozero CMOS op amps are available that eliminate the i/f noise but be sure to check their current noise specification before getting too excited about them. High resistance surface-mount resistors are available from a few manufacturers but you need to be careful of their parasitic capacitance and excess noise ratio.

Another possibility, again depending on bandwidth, might be an IVC102 switched integrator. This topology eliminates the high value feedback resistor and its attendent thermal noise contribution.

Thank you for your answers. The upper frequency i need to amplify ist roundabout 200KHz. So i should take the opa827 as a starting point for my charge amplifier.

Again thank you.

Hello again,

i deciced to pick the OPA320 for my charge amplifier (see attached image file). This preamp has a gain of ~ 10 and a band-pass response from ~1Hz to 100KHz.

But now i am unsure about noise of my reference voltage ("V_VGround" in the image). I need this Bias DC Voltage to provide some sort of virtual ground allowing pos and neg swing around it. Best Voltage References i found have about 50uV-PP noise at their output (10Hz-10KHz).  My input signal voltage is also in the range of 10uV to 50uV.

- Should i better take a dual supply design to benefit of common-mode rejection?

- Should i add a charp low pass at the output of any voltage reference?

What do you suggest?

Thank you very much!

What's going to be fun will be the guarding arrangements. The 5G might be a chain of 1G resistors, and you'll need to guard each end of each resistor, with a second resistor chain (much lower impedance, say, 5x10K) from the output back to Vground.

Hey thank you for your answer. I thought of using one piece of resistor instead of a chain. Whats the reason for your suggest using a chain? Price? Capacitance?

Tobias;

I've attached a TINA circuit for a charge amplifier using a CMOS OPA320 and a FET OPA827 for comparison. You can see the results in the curves that I've included.

The noise results show that the OPA320 has lower noise up to about 50kHz and the OPA827 has lower noise above that. The lower current noise of the OPA320 (0.6fA/sq rt Hz) is an advantage over the higher current noise FET (2.2fA/sq rt Hz). You can low-pass filter the charge amplifier output as well as provide additional voltage gain with a good op amp (OPA320) in an active LP filter circuit. See FilterPro for filter design help. Note that an OPA827 cannot operate on a single +5V supplu.

I've shown the additional noise that is contributed by the 12k resistor in series with your input. It is advantageous to eliminate this if possible but I am guessing that this may be a current-limiting device for the input protection diodes (I omitted them in my circuit since, if they are low capacitance, the effect on BW & noise is small). Calculating the charge from your piezo sensor that might be dissipated in the OPA320 input ESD diodes might tell you if the additional diodes are necessary.

The Vs/2 voltage divider will contribute little noise if it is heavily filtered; the only disadvantage to this is that it takes a few seconds to establish its DC operating point when the circuit is powered up.

I would select the feedback resistor with care; most high value resistors are not very low noise. I have had good experience using G ohm resistors from  Ohmcraft  ;  http://www.ohmcraft.com/?gclid=CP_Xle-R2qsCFQNggwodKg9BUg   they offer values up to 50G ohms in surface-mount types.

Hello,

thank you Neil for your great help. I worked throw your files and read additional literature about this topic. I think i will use the OPA320 because i think the efford to realize the neg. (or the >5V Vcc) supply is not worth the benefit in my case.

The initial (primary beside the current limiting aspect) idea for using the 12K was to set the upper frequency limit (of the formed bandpass) by me and not by the op-amp itself. I thought the noise gain (1+C_s/C)  could get the opamp into ocillation, but now i think the gain of ~10 is not too high. I decided now to take a 1.2K instead auf 12K just working as input current limitation/protection.

I am very happy for your idea with the resistor divider + shunting cap. Because of lower space & price i prefer this to using a voltage reference now.  Two questions:

- Am i right, that the gain for the voltage divider noise would be ~1 (for frequencies >0) without shunting it?

- Was there a special reason to pic 1G voltage divider beside of current consumption?

I will now start to make a first prototyp of this design and hope that there a not too much pitfalls with guardrings etc...

Again thank you!

Tobias

Hello again,

So far i have designed  the amplifier stages & anti aliasing filter on paper. I will use single supply OPA2320 Amplifiers in chargeamp and Sallen-Key Configuration. I am quite unsure if my thoughts are right. It would be nice if anyone with expertise in such things could give me some feedback.

The system facts:

The smallest sensor (unfortunatly it has a poor sensitivity) signal i want to "hear" is:  6.6uVrms -> (*2*sqrt(2)) -> 0.019 mVpp

The biggest sensor signal i want to "hear" is: 47uVrms -> (*2*sqrt(2)) -> 0.134 mVpp

For dynamic range follows: DNR = 20log(47/6.6) = 17dB

The total needed gain for 3.3V supply is: Gain = 3.3Vpp / 0.134mVpp = 25000 = 10*50*50

For total Noise (from calculations/Spice) from 1Hz to 100kHz follows Vnoise_rms = 5uVrms*25000 = 0.1175Vnoise_rms (5uVrms = input referred Noise of 1st Stage)

SNR = 20log(Vrms_big/Vrms_noise)=20log((47uVrms*25k) / 0.1175) = 20dB (biggest Signal)

SNR2 = 20log(Vrms_small/Vrms_noise)=20log((6.6uVrms*25k) / 0.1175) = 3dB (smallest Signal)

Max. frequceny of interest: 100kHz

I ask myself now which order of anti-aliasing filter, how may ADC Bits and which sample rate would be sufficient. This questions i  have:

1)

I will use a 4. Order Lowpass with f_corner at 100kHz. This filter is made (by filterPro) concurrent with the two gain of 50 stages. This filter will damp my signals with 80dB/dec so the damping would be 20dB at a frequency of 275kHz. If i take 0.5fs=275kHz any aliasing effect would be damped in the region below my noise floor. So for a 4th order low pass i should take 550kHz as sampling rate. Is this point of view ok?

2)

The SNR is just 20dB implies this that i theoretical need only a 4Bit ADC!? I don't see the benefit of more Bits if the uncertainty between noise and signal is so big. Or even worse i need only a 2Bit ADC because of SNR2?

3)

Am i right that i could neglect the quantisation noise of any adc because my system noise is so big and it's a root squared sum addition?

Thank you for clarify my thoughts!

Tobias

Tobias;

It sounds as if you are making great progress on your hydrophone amplifier. The calculations that you have made are useful but it would be a good idea to run a noise and AC simulation in TINA-TI to verify them. Noise spectral density is not flat so the simulation will solve that for you.

Be careful about designing high gain high order filters-- this puts a severe requirement on the gain-bandwidth required of the op amps used in the filter. An alternative is to add a gain stage ahead of a lower gain LPF. As far as the order of the filter is concerned, a 4th order LPF sounds about right; it will give good anti-aliasing and can be accomplished with two op amps or one dual op amp. The type of filter characteristic required will be dependent on the type of signal you will be dealing with. An impulse signal will need a Bessel Function LPF, but a Gaussian LPF is also not bad, somewhat like a cross between a Bessel and Butterworth filter. A Butterworth is a good general-purpose filter but if you need very sharp skirts, a 1dB Chebyschev LPF will do it at the expense of lots of ringing.

Hello again & thank you for your answer.

I did AC Simulations and Noise Simulation with Tina and they verified my calculated Numerbers :)

My biggest problem is the number of ADC Bits. I am unsure how many i have to take - what would be a good dicision if SNR is best 20 dB and worst 3 dB? 4/2 Bits only?

Tobias

Tobias;

These days it is hard to find an ADC with fewer than 12 bits; 16 bit and higher resolution converters are so popular that they are very inexpensive... so it makes no econimic sense to use a low resolution ADC. There is really no drawback to using the high resolution ADC. If you are looking for a periodic signal, you can take digital records and "stack" them together to get signal averaging. This reduces the random noise by the square root of the number of records that are averaged.

Hello and again thank you for your answer,

attached you see the complete Amplifier as a TINA-TI Simulation. I think i will give this Design "a try" if it does not raise any concerns here. It's the chargeamp & two stages with gain 50 + Unity Gain 4th order Sallen-Key Filter with fc at 100kHz. I'am a bit unsure if i should decouple (as in the tina-file) or not the stages. Decoupling raises the lower cut frequency a bit...

Btw. the 2nd use of oversampling (reduce noise by averaging) was very helpfull. I will use a 8Bit ADC with sample rate of about 500kHz.

Regards

T.Rohde

Tobias;

Nicely done. I have only one suggestion-- you only need one voltage divider to generate your 1.65V. Since the input bias current of these CMOS op amps is so low-- even with many tied in parallel-- that it isn't necessary to use a voltage divider for each op amp non-inverting input.

I usually model a piezoelectric transducer as a current generator in parallel with a capacitor but it is only a matter of personal preference.

BTW, are you related to Ulrich Rohde or the famous Rohde und Schwarz?

Hello,

thank you for your advises. You helped me a lot. I finisched now a first PCB-Design and i'll see if it works like the simulations.

I am not related to these two persons, but you guessed correctly the country :) I do this amplifier in relation with my diploma at the Technical University of Hamburg. Very likely it will be used in commercial device soon.

Tobias;

I think that you will find pretty good agreement with your simulation. Good simulations rely on having good device models; the OPA320 is a decent one.

Yes, I "recognized the country", Tobias. I spent three years in Germany in high school-- one year in Heidelberg and two years in Frankfurt. Later, I spent two more years in Germany-- Babenhausen and Hanau, this time as a GI. Those days in Germany were wonderful.

Hi Neil,

I read this post with much interest. I have a piezo hydrophone with its capacitance of 30 nF, so I wonder if the schematic you proposed in that time for a charage amplifier using OPA320 can be applied as such. Our aim is to get a flat response on the band 1Hz-50KHz (not less that 30KHz). It can be of low amplification (10-20) as it will be followed by a PGA. Is there now a better amplifier suitable for this application than the OPA320?

Thank you

Kind regards

Traian

Traian;

Yes, this basic circuit is suitable if the feedback capacitance is increased to 100pF; now the resistor can be reduced to 2G. The OPA320's advantage is that it has low noise in the region where most op amps exhibit high 1/f noise.

I hope this helps.

Regards, Neil P. Albaugh

Ex-Burr-Brown

Thank you Neil.

Neil, is there any other OA with similar characteristics for such application but could work to higer supply voltage (up to 15V)? I want to avoid voltage regulators. Thanks

The OPA192 is the closest that I could find. Watch its common-mode input voltage-- the voltage noise is higher at low CMV.

Actually a 5V zener diode and a resistor can be used in place of a voltage regulator if you need to operate an OPA320 on a higher supply voltage.

Regards, Neil P. Albaugh

ex- Burr-Brown

Thank you Neil again,

I just want yet to ask, we tested some voltage amplifiers with our 30nF piezoelement and we got a "ringing" response to pulse excitation. Can you advice what might be the reason for this behaviour? I attach you the photo, so you see the frequencies and amplitudes for both, pulse excitation and response. Thank you. Traian

Hi Neil, did you receive my mail below?

"Thank you Neil again,

I just want yet to ask, we tested some voltage amplifiers with our 30nF piezoelement and we got a "ringing" response to pulse excitation. Can you advice what might be the reason for this behaviour? I attach you the photo, so you see the frequencies and amplitudes for both, pulse excitation and response. Thank you. Traian"