Piezoelectric hydrophone low noise amplifier design questions

[Marco]

The problem with your diode based protection is that, maybe I didn't mention it good enough, the antena works both in TX and RX modes, so when we are transmitting the impulse, about 50...100 volts will be applied to the protection circuit, thus it should have some kind of series protection.

Just put something like a VO1400AEF in there to block the receiver during transmission. As for MOSFET current limiting, no resistor with low noise (ie. ~10 Ohm) is going to turn off MOSFETs at reasonable currents. You'll need something slightly more complex if you go that way.

You simply can't have it both ways, 1k range series input resistance and low noise don't mix.

The opamp doesn't do the biasing in the schematic, you should see the circuit as a composite opamp where the gate is the positive input and the source is the negative input (with a very high offset voltage). The JFET sets the noise, the opamp provides the loop gain and low output impedance. You can do much the same thing with a PNP transistor, but the opamp allows more flexibility (the PDF is good reading, his cheap circuit and hydrophone beats the much more expensive ones on the market ... probably for the reason mentioned above, input series resistance is poison to low noise).

[David Hess]

Just put something like a VO1400AEF in there to block the receiver during transmission. As for MOSFET current limiting, no resistor with low noise (ie. ~10 Ohm) is going to turn off MOSFETs at reasonable currents. You'll need something slightly more complex if you go that way.

I was thinking something like that but wondered if it would be too slow.  500 microseconds comes out to a range of about 1 foot so it would actually be quick enough.

[Sparker]

Do I understand your depletion MOSFET idea correctly, David? The picture is in the attachment.

As for MOSFET current limiting, no resistor with low noise (ie. ~10 Ohm) is going to turn off MOSFETs at reasonable currents.

It looks like I can only buy a BSP129 depleted MOSFET here. With Rs=20 Ohms and Rds=7 Ohms (typical), the total equivalent resistance of the whole protection (2x20 Ohm resistors and 4x transistors, all in series) is 50 Ohms, or ~0.9 nV/sqrt(Hz) of noise, not so bad compared to 1k resistors.

500us for an externally controlled switch(like the optical relay suggested) is probably not bad. The pulse length will be about 600us, but extra 500us of 'off time' won't matter much for a side scan sonar if it's located several meters above the ground.

I'm quite constrained in space so I can't afford a nice & expensive toroid 40mm in diameter, so I stopped at this one: 10x6x4 mm toroid, T38 material. Probably I can fit about 60 turns total one layer. Is it a good choise?

I am thinking if I could just move the protection to the "op-amp" side of the transformer, so its noise voltage is not amplified by the transformer, but probably it won't work well with the 10mm transformer mentioned above.

[Marco]

For a 100V RMS signal you're passing 100 mA and burning 10 Watt in the MOSFET (which it can't take for very long).

The problem with the transformer is getting sufficient self inductance, the primary impedance does need to be significantly higher than the piezo to not make a mess.

[David Hess]

Do I understand your depletion MOSFET idea correctly, David? The picture is in the attachment.

As for MOSFET current limiting, no resistor with low noise (ie. ~10 Ohm) is going to turn off MOSFETs at reasonable currents.

That is exact it.

It looks like I can only buy a BSP129 depleted MOSFET here. With Rs=20 Ohms and Rds=7 Ohms (typical), the total equivalent resistance of the whole protection (2x20 Ohm resistors and 4x transistors, all in series) is 50 Ohms, or ~0.9 nV/sqrt(Hz) of noise, not so bad compared to 1k resistors.

Depletion mode MOSFETs are kind of expensive also.

500us for an externally controlled switch(like the optical relay suggested) is probably not bad. The pulse length will be about 600us, but extra 500us of 'off time' won't matter much for a side scan sonar if it's located several meters above the ground.

I was thinking that it might be a problem if the sonar was up against something or if there was another sonar in the water.

I'm quite constrained in space so I can't afford a nice & expensive toroid 40mm in diameter, so I stopped at this one: 10x6x4 mm toroid, T38 material. Probably I can fit about 60 turns total one layer. Is it a good choise?

Inductor for what?  For the transformer?  I would probably use a pot core for easier winding.  What transformer are you using for the transmit side?  I might use the same transformer or at least the same core for both.

I am thinking if I could just move the protection to the "op-amp" side of the transformer, so its noise voltage is not amplified by the transformer, but probably it won't work well with the 10mm transformer mentioned above.

I was thinking the same thing but the transformer will increase the applied voltage from the transmit chirp making protection more difficult.  Applying the pulse to the transformer itself could be a problem if it saturates and the protection is on the secondary side so a larger than necessary transformer is needed for this.

For a 100V RMS signal you're passing 100 mA and burning 10 Watt in the MOSFET (which it can't take for very long).

This needs to be considered but the chirps are pretty short.

The problem with the transformer is getting sufficient self inductance, the primary impedance does need to be significantly higher than the piezo to not make a mess.

I was going to say getting a sufficiently high frequency response but obviously this is not a large problem given the existence of the transmit transformer which must be larger to prevent saturation.  I wonder if a tuned transformer would be better in this case.

[Sparker]

I'm quite constrained in space so I can't afford a nice & expensive toroid 40mm in diameter, so I stopped at this one: 10x6x4 mm toroid, T38 material. Probably I can fit about 60 turns total one layer. Is it a good choise?

Inductor for what?  For the transformer?  I would probably use a pot core for easier winding.  What transformer are you using for the transmit side?  I might use the same transformer or at least the same core for both.

I am thinking if I could just move the protection to the "op-amp" side of the transformer, so its noise voltage is not amplified by the transformer, but probably it won't work well with the 10mm transformer mentioned above.

I was thinking the same thing but the transformer will increase the applied voltage from the transmit chirp making protection more difficult.  Applying the pulse to the transformer itself could be a problem if it saturates and the protection is on the secondary side so a larger than necessary transformer is needed for this.

The problem with the transformer is getting sufficient self inductance, the primary impedance does need to be significantly higher than the piezo to not make a mess.

I was going to say getting a sufficiently high frequency response but obviously this is not a large problem given the existence of the transmit transformer which must be larger to prevent saturation.  I wonder if a tuned transformer would be better in this case.

I also thought of winding the RX winding at the TX transformer in the beginning, but I am planning to use one TX amplifier to drive two antenas through diodes like in the picture. So I can't use this approach.

For a 100V RMS signal you're passing 100 mA and burning 10 Watt in the MOSFET (which it can't take for very long).

This needs to be considered but the chirps are pretty short.

Chirps will be transmitted about every 100 milliseconds, take chirp length 1 ms, and we get about 0.1 Watts average in this case.

[David Hess]

Below are the examples I mentioned of diode bridges being used as series protection elements for low impedance inputs and outputs.  The current fed into the diode bridge limits the current between input to output.  The control inputs to the diode bridge could also be reversed to completely disconnect the input from the output.

[Sparker]

Thank you! It's actually a very smart design. Do you think it will make a lot of noise?
It seems to me that it's not noisy at all, at least in my rough simulation a pair of 4148 diodes with 10mA running through them produce around 0.3 nV/sqrt(Hz). I'll try to simulate it in detail later.
Meanwhile I've assembled a rototype of our amplifier on a PCB, seems to be working fine. When I figure out how to communicate with the digital potentiometer, I'll try to analyze the self-noise of the circuit and share it here. 

[David Hess]

Thank you! It's actually a very smart design.

The Tektronix implementations were for protecting 50 ohm oscilloscope inputs with a 350MHz (485) and 1GHz (7A29) bandwidth.  I have never seen this used anywhere else.

Another interesting thing is how Tektronix used the 741 operational amplifiers in the circuit from the 7A29; they are configured as current feedback amplifiers with 2 high impedance differential inputs, 1 low impedance differential input (the output pin), and 2 current outputs from the supply pins.

Do you think it will make a lot of noise?

It seems to me that it's not noisy at all, at least in my rough simulation a pair of 4148 diodes with 10mA running through them produce around 0.3 nV/sqrt(Hz). I'll try to simulate it in detail later.

Noise is where it gets interesting.  The noise is attenuated by the ratio between the impedance of the current source which is high and the impedance of the signal which is low.  Tektronix was protecting a 50 ohm signal, required the highest possible bandwidth, and had to operate down to DC so their current source had only a moderate impedance.  In your case active current sources could be used with cascodes to get incredibly high impedance at 240kHz and associated low noise.  The diodes and current sources have to be able to survive the maximum input voltage though.  I might also disable the current sources while the transmitter is actually active to limit power dissipation in the current sources and loading on the transmitter.

I just thought his was a neat idea which might apply to your application.  I have only used it for protecting low voltage signals like function generator outputs and trigger inputs and outputs.

[Sparker]

So here are the results with the real circuit!
With the input of the first opamp completely shorted I am getting about 8 nV/sqrt(Hz) which totally satisfies me for now, until I shield the whole circuit completely.  I made the whole setup: I connected the MCU i want to use in this application, recorded 100 ms of data, calculated its RMS, divided it by gain and divided it by sqrt(noise bandwidth of the filter). I think the circuit is mostly a success.

But then I have a problem: I use the LM2594 DC-DC converter running at 140 kHz to generate the negative -2.5V rail for the op amps. When I connect the transformer to the input of the opamp (and remove the short between opamp's input and GND of course), somehow the 140kHz and 280kHz from the converter get coupled in so much that the output of the circuit gets saturated. I've provided much filtering for the -2.5V rail with an additional LC filter, which should result to about 2uV of ripple at the -2.5V rail. From all this I conclude that most likely the transformer is responsible for this. What do you think?
 I should have probably used the pot core for the transformer because it's better shielded, but honestly I thought that toroids don't leak electromagnetic field and thus don't pick it up.
I've tried putting a tiny tin box over the toroid but it didn't change anything much.

[David Hess]

With the input of the first opamp completely shorted I am getting about 8 nV/sqrt(Hz) which totally satisfies me for now, until I shield the whole circuit completely.

I just happened to run across this yesterday and it applies to your application.  Check out figure 52 on page 17 of the AD797 datasheet which shows the use of a low input voltage noise high input current noise bipolar operational amplifier with a large area photodiode.  It makes my point about low noise with a high capacitance transducer at medium frequencies.

But then I have a problem: I use the LM2594 DC-DC converter running at 140 kHz to generate the negative -2.5V rail for the op amps. When I connect the transformer to the input of the opamp (and remove the short between opamp's input and GND of course), somehow the 140kHz and 280kHz from the converter get coupled in so much that the output of the circuit gets saturated. I've provided much filtering for the -2.5V rail with an additional LC filter, which should result to about 2uV of ripple at the -2.5V rail. From all this I conclude that most likely the transformer is responsible for this. What do you think?

I should have probably used the pot core for the transformer because it's better shielded, but honestly I thought that toroids don't leak electromagnetic field and thus don't pick it up.
I've tried putting a tiny tin box over the toroid but it didn't change anything much.

Bobbin type inductors like you used are not self shielding like toroids and pot cores so they spray magnetic flux everywhere which gets into everything.

Operating the switching regulator at the frequency of interest is just asking for trouble whether magnetic flux gets into everything or not.  Filtering will need some serious attention.

[Sparker]

This thing is so weird. I've replaced the converter's main inductor (the one which gets pumped with energy) with a toroid of the same type I used for the amplifier input. Generally I was getting almost same level of coupling, bug by tuning its rotation very fine, I could find the position which produced almost no 140kHz(and its harmonics) leakage. It's of course impractical because the tiniest movement will result to 140kHz noise increase.
Pot cores are very hard to find here, but I think I've found a few, going to try them next week.
As for converter's frequency, would you suggest getting a converter with higher or lower frequency of operation? The majority of converters operate in this frequency band. My guess is, the higher it is the better, because it will be easier to filter it.

[David Hess]

Toroidal and e-core inductors do not leak as much flux.  Distance between the power supply and sensitive circuits helps; magnetic flux falls with the cube of the distance.

Some switching regulators allow their oscillator to be phase locked to an external reference; this would allow a non-interfering frequency to be used.

Another alternative is to disable the switching regulator during the receive gate.

Higher frequencies can be filtered with smaller components but electrostatic coupling increases at higher frequencies so parasitic capacitance becomes more important.

Keeping interference from the switching regulator out of your 4nV/sqrtHz noise level will be tricky.

[Sparker]

I've learned that there is no such thing as a perfectly shielded transformer.
So, I've tried the pot core transformer for the input of the circuit, but I was still getting a lot of coupling from the switching power supply. So I tried the last option: a different frequency DC-DC converter. ST1S10 runs at 900 kHz and I am getting no troubles with it. 

[Sparker]

I think I need some help with this device again...
I can't decide which of the protection schemes to use. In the attachment there are two schematics.

First question is, which way to connect the diodes. In the starfish sonar they used the top variant, with two double-diode assemblies connected to the ground. In other designs I can see only one pair of diodes, like in the bottom picture. The top variant produces more voltage when the input is saturated but It's not a problem for me.

Second question I can't resolve is: do I need to connect the central tap of the primary of the transformer to the ground? I see no reason to do that here, but I might be mistaken.

[dmills]

Either is fine.

The traditional approach to a single transducer sonar Tx/RX switching is to just stick a pair of back to back diodes in series with the secondary of the transmit matching transformer and take the receive signal from across them, simple, robust and no need for active TR switching.

I always like the BF862 as a preamp input stage for low frequency sonar, not as nice as the Linear Systems Siliconix copies, but way cheaper, and above about sea state 3 or so you are usually noise limited by surface noise anyway.

Regards, Dan. 

[Sparker]

Hi, Dan!
Thanks for your response!
Good to know that both designs are OK.
I understand what classic solution you are talking about. Just for reference, I've attached a schematic I took from some old soviet book.
Am I right if I say that this solution is worse than the previous ones because It's not symmetric and thus will provide less suppression for common mode noise? 
I'm afraid I can't use it anyway because we have a single TX matching transformer driving two transducers. The transducers are isolated from each other with back-to-back diodes, there's also a schematic at page one of this thread.

[dmills]

So you have some sort of push pull driven side scan affair? Weird, but whatever.

I would be looking at something with a pair of back to back diodes in each leg and some small transformers.

By 250KHz you are getting into the place where minicircuits have some interesting offerings for ready made small RF transformers, might be worth a look at something like a 16:1 impedance ratio jobbie (4:1 voltage gain).

Are you building the transmit transformer as a tuned job (to tune out the fixed capacitance of the transducer)?

On the variable gain amp front, there is something to be said for just using a log amp, 100dB of range is not that hard, AD have the good log amps, but it depends on what your waveform looks like. 

Regards, Dan.

[Sparker]

The circuit you've attached looks like a nice solution. Previously we were discussing how to make a protection without noisy series resistors. I think this one is great, looks very symmetric  and without serial resistors! 

Yes, I was planning to use a push-pull driver for the transducers. Why do you think it's weird? We use this driver for several sonars here. Push-pull configuration seems like a traditional transformer configuration. See the attached schematic below to see how I'm going to attach the transducers.

No I didn't consider tuning out the capacitance of the transducer. But we have already a pair of transducers with a TX amplifier driving them, so I was going to reuse it.

I am going to use a chirp signal sythesized with an STM32 timer, so a demodulating logarithmic amplifier is not suitable here. Or are you talking about a non-demodulating one? I'm not really familiar with these devices. At the signal processign side, I am digitizing the signal right after the amplifier and do the rest in digital domain, the STM32 is very capable of doing that.

[dmills]

I worked mainly on the transducer side of things, and typically our transducers did not have a centre tap on the ones with several series elements.
The output was typically unbalanced so we would just connect the screen to system ground and pick off single ended on the other side of the diode pair.

Primary side power stages are generally push pull into the transformer, we found that tuning things so the transducer ran on the inductive side of resonance was easier then running on the capacitive side as it makes for a simple active clamp to real with the reactive power flow.

Agree that a log amp is no use if you are chirping for range to target, but they are great for things like interferometric sidescan where relative phase is what you really care about.

Regards, Dan.

[Sparker]

Hi again, guys!
I've finally tested the side scan sonar I was making and I'm excited that it worked flawlessly! I wanted to thank you for you valuable help with the design of the analog part. I'm not an analog expert at all and it meant a lot for me.
Here's one of the pictures we've made with the sonar. It's some crack at the floor of one of our arctic seas.  There were other interesting scans of different radioactive garbage buried there but I don't think I can share them here.

[Sparker]

Just came across a (what seems to be) nice series of T/R switches for ultrasound applications from microchip:
https://www.microchipdirect.com/Chart.aspx?branchId=9066
Here's datasheet of one of the devices: http://ww1.microchip.com/downloads/en/DeviceDoc/MD0100-Single-and-Dual-Channel%20High-Voltage-Protection-TR-Switch-Data-Sheet-20005738A.pdf

Sorry to bump that old thread, but let it be there in case someone comes across it while searching for ultrasound rx amp design.