Small signal and high voltage

I’m building a water echosounder with a single transducer (200 KHz). All works well on the transmit side, but now I want to read the small signal returning. My transmit is in the hundreds of volts (about 600-800 vpp) and the signal I’m interested in is in the millivolts. From the end of the transmission to the start of receiver could be as low as 150 us.

My problem is that I can’t figure out a way to hook up this to a microcontroller without burning it. How could I extract the smaller signal ?

So if there is that little time between the send and receive and the send is low frequency, I assume that the tiny signal is overlaid on the big signal?
Although you can easily cancel out a signal you are generating, the magnitude of the voltage would make it hard to use an op-amp without dividing it quite a bit first. What sort of signal level would the return be?

My transmit pulse is about 100 us, this signal returns after a short period (a couple hundred microseconds) so there is no overlapping between transmit and receive, both are 200 KHz. I would like to transform the return signal up to 0 to 5v, currently it is a couple hundred millivolts on the best case.

You may copy some design elements from this circuits, essentially R1 VD1 VD2 input protection:
http://gelezo.ws.md/home_electronics/1019000/1019004/eholot_ribolova-lyubitelya.html

I would have thought if there was no overlap of the signals and you are in control of the signal generation (ie, you also have a low voltage version of it), it would be fairly easy to provide an 'off' period for the sensor. Once you have removed the DC component and low pass filtered the input, it should be fairly manageable to get into an op-amp to clean it up and make it into pulses for the arduino

Something like this (no warranties of any kind):

I'd use a similar solution to that given by pito, but instead of (or possibly as well as) the diodes to limit the signal, I would use a small signal mosfet (e.g. 2N7000) to short the op-amp input to ground while the signal is being transmitted.

100K is too large a series resistor between the transducer and the op amp, because in conjunction with the capacitance of the diodes and the input capacitance of the op amp, it will cause too much attenuation @ 200kHz. Also, you need to use an op amp with a greater bandwidth than LM324, because that one will give you very little gain @ 200kHz.

Thanks for all the replies!

The dual diodes trick certainly got me started! I tried that spice circuit and as dc42 mentions, I used the 102 capacitor, replaced the 100k resistor with a much smaller one and got better results, also used a 1mH inductor to dissipate the DC build up, the end result so far is a really short blanking area of <100us (way better than I expected).

As dc42 also mentions, the bandwidth limitations of the LM324 would not allow me to even make a unity gain filter.
I now realize I need much better bandwidth op-amps. What would you guys suggest?

100K is too large a series resistor between the transducer and the op amp, because in conjunction with the capacitance of the diodes and the input capacitance of the op amp, it will cause too much attenuation @ 200kHz. Also, you need to use an op amp with a greater bandwidth than LM324, because that one will give you very little gain @ 200kHz.

The 4 diodes contribute with 4pf in total, plus 10pf opamp and stray - let say 14pf total. With 100k it makes 115kHz corner. Mind 800V input, and diode’s peak current - therefore 100k (8mA), but it could be smaller (you are loading the transducer with this resistor - the higher value the better). The op-amp is just a working example, higher BW would help, indeed.
PS: as the R1 - I would use few in series = R1+R5+R6 (mind high voltage). C2 is for 2kV at least!
PS1: the Q is what you are going to do with the output signal - you cannot sample it with arduino 8bit (maybe DUE), so you need a peak detector there… :wink:
PS2: the op-amp - maybe with >3MHz BW, a “beyond-the-rails” op-amp (mind the op-amp’s input voltage, could be negative)
No warranties of any kind…

pito,

Good points on resistive loading on my transducer, the datasheet on the transducer says 60 +j0(t) (uses internal transformer) 500 watts, so it does not seem to hurt it so far.

I'm currently using a 10nf in series with a 2k resistor. I also use a 1mH inductor in parallel with the diodes, which gives me a great looking signal atm. (I'll soon post a screenshot of my scope)

You are correct about the 328p, the fastest I can sample with the Uno is around 16-20us, way slower than what I need. I will however try and use the Due tomorrow (I got two with me).

The current application for this is an echosounder, just for depth which I feel I can accomplish with just timing but I rather digitize the signal and do some DSP with it for future projects of underwater acoustics. Seems like 4us sampling (250KHz) is the best I can accomplish with the Due, from reading this:

Seems I'm close now to getting this into the Due, if it could do 500KHz (apparently stated in the datasheet) I would be golden...

I'll also look into that higher BW op-amp, Thanks!

bastukee:
Thanks for all the replies!

The dual diodes trick certainly got me started! I tried that spice circuit and as dc42 mentions, I used the 102 capacitor, replaced the 100k resistor with a much smaller one and got better results, also used a 1mH inductor to dissipate the DC build up, the end result so far is a really short blanking area of <100us (way better than I expected).

Reducing the 100k resistor is fraught with issues - if the incoming signal during TX is hundreds of volts a low value resistor will absorb all
the power and burn out… You need to switch the signal, which means using 1000V rated transistors or similar. Basically an RX-TX switch is
a requirement.

As dc42 also mentions, the bandwidth limitations of the LM324 would not allow me to even make a unity gain filter.
I now realize I need much better bandwidth op-amps. What would you guys suggest?

This is RF circuitry really - the normal approach is to use several stages of transistor or FET amplification with
tuned transformers to couple between stages. This filters out all but the 200kHz and the LC circuits tune out
all the stray capacitance. Failing that find a decent high frequency opamp - there are many, but don’t expect them
to also be 5V and rail-to-rail…

MarkT:
Reducing the 100k resistor is fraught with issues - if the incoming signal during TX is hundreds of volts a low value resistor will absorb all
the power and burn out... You need to switch the signal, which means using 1000V rated transistors or similar. Basically an RX-TX switch is
a requirement.

The resistor certainly mustn't be reduced too much. The OP said the tx signal level was 600-800V p-p, so 300V or less rms. If he reduces the 100K resistor to 30K then the resistor will absorb 3W of power from the transmitter. Depending on the peak power fed into the transducer (which is probably at least 50W), this may be acceptable. In an echo sounder application, the mark-space ration of the pulse is low, so heating in the resistor is unlikely to be a problem (OTOH voltage rating may be an issue). A 1W metal film resistor should be more than adequate, or three 10K 0.5W resistors in series.

So I don't think it is necessary to switch the signal path using high voltage transistors, provided a small power loss in the resistor is acceptable. Alternatively, the OP can stick with using 100K (giving about 1W max instantaneous power dissipation), minimize the circuit capacitance and accept some loss of signal.

I still think it is a good idea to use a mosfet to short the amplifier input to ground, to prevent the amplifier going into saturation, from which it may take a little while to recover.

bastukee:
I also use a 1mH inductor in parallel with the diodes, which gives me a great looking signal atm.

Check the self-capacitance and/or self-resonant frequency (normally given on the datasheet). You want this to be appreciably high than 200kHz. Alternatively, tune the inductor so that in conjunction with the circuit capacitance, it resonates at 200kHz.

dc42,

Here is my current circuit...

I some calculations and came up with:

fc= 1/(2pisqrt(1mH * 670pf)) = 194.438KHz

Not sure how to apply the 4pf capacitance from the diodes, which may or not make a difference.

Here are some results...

R1 = 30k

R1 = 5.6k

R1 = 2K

I have a small assortment of transistors (I'm not as good on transistors) my experience has being that they were too slow on RF level, likely because of the saturation you mentioned. I'm looking for a signal mosfet for that clamping of the Tx but in the meantime, how would I go about the transistor amplification? Given that I'm short on a high BW op-amp.

First, you should ground pin 1 of the transducer SG1. Otherwise, you are relying on capacitance of the wiring and the transformer to provide a return current path for the received signal.

Second, even if the diode capacitance is only 4pF, stray capacitance and input capacitance of the amplifier is likely to be much more than that.

As for the ampllifier, I would use something like the attached. The differential-pair configuration gives low input capacitance, wide bandwidth, and with suitable component values is non-saturating. I once has to build a 1 to 500KHz amplifier to count the beat frequency between two lasers, and I used a chain of amplifier stages something like this. The best component values depend on how many volts positive supply you are running it from.

PS:

I some calculations and came up with:

fc= 1/(2pisqrt(1mH * 670pf)) = 194.438KHz

Not sure how to apply the 4pf capacitance from the diodes, which may or not make a difference.

You've misunderstood me. I was meaning to choose an inductor so that it and the amplifier/diode/stray capacitance resonates at 200kHz. For example, use a inductance of around 22mH, then bring the capacitance up to 28.8pF, and aim for a Q factor of at least 10.

btw 1mH has a reactance of only 1.25K ohms @ 200kHz, so in conjunction with the 30K resistor, it will attenuate your signal by a factor of ore than 20.

Indeed I was confused, Thank you very much for the simplification of the concept. I noticed that the reactance of the capacitor and inductor values was both close to +30k and -30k, which makes sense now why you asked for that value.

http://www.st-andrews.ac.uk/~www_pa/Scots_Guide/info/signals/complex/react.html

I changed C1 to 22pf and inductor to 22.2, the circuit is so much better tuned now, I get a good signal with a 30k resistor and I use much less current (15 times less?) My transducer voltage went up significantly too, given the smaller current drawn.

My next step....

I like to try and clamp that source (TX) down to ground. I have a 100 transistor pack and both a NTE2361 (NPN) SI NPN HI speed switch and a NTE2362 (PNP). As I mentioned I'm not as good on transistors, re-learning my way though. Any help on figuring out those values?

To clamp the signal to ground, you need a small-signal mosfet, not a BJT. The 2N7000 is not difficult to obtain, it’s available from distributors, hobby shops, and eBay.

The NTE2361 would be OK in that amplifier circuit I posted yesterday; or you could turn that circuit upside down and use the NTE2362. As I mentioned before, the values of the resistors in that circuit depend on the supply voltage you are going to use. Alternatively, use a wideband op amp such as OPA835 or OPA836 (unfortunately only available in SMD packaged, but you can get adapter boards from Sparkfun, Futurelec, eBay etc.).

..I would even consider the AD8307 logarithmic amplifier:
a) 92dB range with 0-2.5V output
b) up to 500MHz
c) 2.7-5V Vcc