Charge Amplifier Instability

[philwinder]

charge amplifier

Hi again,
I'm having trouble with my AD8066 charge amplifier circuit (See attached). On the input is a hydrophone with a capacitance of 2.3nF, so I should be getting a gain of 230 ish.

Whatever I seem to do, the output oscillates at a very high frequency (MHz possibly?) and I cant figure out why. I have tried separate power supplies and I am fairly sure its not due to them, and it is made on a PCB with all SMD's.

I have altered the value of the feedback resistor which only changes the sensitivity of the amplifier. I have also tried adding a resistor in line with the input to create a low pass filter, but that had a wierd DC gain effect and flew off the rails, not sure why.

Sometimes the oscillation saturates and sometimes it stays within the rails. This seems to be a function of the feedback capacitor, C1, which also alters the oscillation frequency slightly (still very high).

So, does anyone have any tips on how to save this amplifier?
I am a bit worried about the AD8066 itself because I also used it in a voltage gain mode and that was also slightly unstable. Could it be my PCB design? Oh, and adding a metal can makes the oscillation worse! I'm sure its something to do with the capacitor, or stray capacitances, but I don't know how to test or fix.

Thanks,
Phil

 

[FvM]

charge amplifier circuit

I suspect capacitive output load as source of instability. You should add a series resistor, e.g. 50 ohms at CON2. Unsuitable PCB design (no effective ground plane) can be another issue. However, assuming that C1 enforces a feedback factor k=-1 at high frequencies, bypassing isn't that critical - if capacitive loads are isolated.

 

[philwinder]

charge amplifier design

I tried that and it didnt work I'm afraid. Its one of those really annoying bugs that seems to change all the time. When I alter something, it could work. So then I alter the other amps (4 amplifiers in 2 dual packages) it breaks again!

I think an image of the noise could help...

This image is when the transducer is connected,

And this one is with nothing connected.
Timebase is 5ms/square, 5V/square, power +10/0V. Biased at about 3.5V, but the image is ac coupled.

Thanks,
Phil

 

[philwinder]

charge amp capacitor

Just as another update:
I have gone back to an older design that used an AD8066, not a new AD8666, which is basically just a slightly worse amp, and it works fine.

So, I email AD tech support and hopefully they might be able help. It must be something to do with the AD8666 high bandwidth...

Thanks,
Phil

 

[FvM]

charge amplifier circuits

Although AD8666 is more different from AD8066 than being "just a slightly worse amp", I don't see a reason for the observed behaviour. It simply looks defective.

 

[philwinder]

charge amp

It simply looks defective.

In my experience, its usually my fault

 

[philwinder]

charge amplifier input resistor

Another update and another question.

I've redone the pcb with pretty much exactly the same circuit (only difference was a series resistor on the input) but with a different layout. Before, I had a guard trace around and underneath the amp driven by the reference voltage (created by a voltage regulator and filtered), but this time I made no guard trace and kept the ground far away to minimise input/ground capacitance.

And now, there is no instability. So this leads me to believe that the instability was due to one of three things.
1) The first is that the guard trace was adding some capacitance to the inputs, adding a pole, and making the amp unstable at some frequency.
2) The guard trace made some sort of feedback loop from the output, though the guard capacitance and back into the driven input (unlikely)
3) Or thirdly I am using a large LC filter on the output of the voltage reference (regulator). By itself, with such a large capacitor, it is stable, but the addition of another capacitance due to the guard trace tipped it over the edge into instability.

Which do you think is most likely so I can mitigate in further designs?

Cheers,
Phil

 

[FvM]

log charge amplifier

As far as I remember, you've using a slow AD8666. Thus I would almost exclude the first two points. I'm curious, where a voltage regulator would be used, for the Vcc/2 node? It could actually open a lot of interference options. But instabilities from this side could be easily detected.

I didn't think yet about the 2.2n capacitor, but it should be better connected to ground. Another problem may arise, if the input node including the transducer isn't perfectly isolating at DC.

 

[philwinder]

filter high frequencies charge amp

8666: Yes, that and the 8066 which is faster, and this is the one that oscillates.

VCC/2: well the problem I had was that if I used a proper series voltage reference it would introduce far too much noise and I couldnt filter it because it couldnt handle a large capacitance, so I get better noise performance from a normal regulator with a LC filter. And noise performance is of the most importance in this case.

2.2nF cap: Oh right, rather than connected to the Vcc/2 bias? I haven't tried that. I'll give it a go at the next opportunity.

Transducer: I would expect there to be a leakage, just like there is in a cap, but I have no control over this. And even if this was the case, then the newer circuit should still oscillate.

Thanks,
Phil

 

[FvM]

how to test a charge amplifier

8666: Yes, that and the 8066 which is faster, and this is the one that oscillates.

I understand your previous post regarding AD8666 just the other way around. So if the oscillation issue is with AD8066, a parasitic capacitance, bad circuit layout, inappropriate load and similar would be a more likely cause.

Generally regarding the nature of oscillations: A lot can be seen from the oscillation frequency. Your screenshots neither have sufficient resolution nor a time scale.

 

[philwinder]

charge amplifier ofset

Yep, and I think if I leave the guard trace out, that should solve the parasitic capacitance issues, and hence the oscillations.
Yeah, the oscillation freq was too high for my poor 10MHz CRT oscilloscope!

Thanks,
Phil

 

[E-design]

how charge amplifer works

Are those 100Meg Ohm resistors??

 

[FvM]

ad8066

Are those 100Meg Ohm resistors

Apparently, yes. It's not that unsual for a charge amplifier. You possibly may doubt, if the bias current compensation resistor at non-inverting input is reasonable, but it actually is considering the AD8066 specification (if offset voltage is an issue at all).

 

[philwinder]

charge amplifier bias compensation

Yes, 100Meg resistors, Offset voltage isnt a big issue. Just want a nice high input impedance.