Transimpedance amplifier oscillates with long input coax

Thread Starter

apg2011

Joined Jul 26, 2011
6
Hi,

I need to convert the output current (up to 2uA) of a micro-channel plate, MCP, (with 32 anode strips) into a suitable voltage (5V max) for our data acquisition system. The required gain for the transimpedance amplifier is around 2megahoms and the bandwidth 300kHz, in order to see the signal shape (not the single pulses of the MCP, but the whole bunch of particles arriving the detector). For that purpose, I have designed a transimpedance amplifier based on a 50ohms termination resistor and a two stages voltage amplifier (see attached images). In the lab tests, the amplifier doesn't show a brilliant signal/noise ratio but it allows me to do good enough measurements of a test signal. When I connect it to the MCP (through 45 metre of URM-43 50ohm cable, as the detector is within a high ionization radiation area), the amplifier oscillates at about 33kHz. Does anybody know how to stop this oscillation? Is it caused by the cable capacitance? Any ground loop perhaps?

Many thanks
 

Attachments

ErnieM

Joined Apr 24, 2011
8,377
First step: add bypass caps around the power pins of all op amps at the amp itself. Shortest leads as possible.
 

Hi-Z

Joined Jul 31, 2011
158
I'm guessing that you're still not having much luck with your circuit. Well, if I can add my thoughts:

The OPA847 seems to be a pretty clever device (I've not used one), but it may be rather tricky to use, owing to its high bandwidth. Actually, it may be overkill for your application. However, one thing does strike me - it's being asked to provide an awful lot of gain. Now, usually you get stability problems in designs which operate at low gains (i.e. with a lot of feedback), so that won't be a problem here.

But the high closed-loop gain will mean that the circuit is very prone to imperfections in the layout. I would suggest the following:

Split the high degree of gain over several stages, using several op-amps. I don't think there's any great need for the first gain-stage to be very high gain.

Make sure you're using a ground plane and surface-mount decoupling/reservoir capacitors. Keep signal paths short, and keep output paths away from input paths.

The fact that you're seeing oscillation at 33kHz may be the result of "squegging", caused by high-frequency instability (caused by bad layout). On the other hand, could the 33kHz be pickup from some nearby source (there is an awful lot of gain there!)?

Anyway, I wish you luck!
 

Thread Starter

apg2011

Joined Jul 26, 2011
6
Hi,

Many thanks for the replies and I apologize for not including the power supply and decoupling circuits in my original post. The circuit I've posted is the one for simulation but not the actual prototype one. I attach the power supply, decoupling and board layout for the version I'm testing. You will find there are extra components (TVS protection diode and a series zero ohms resistor at the input) and some resistors on the decoupling circuit that were replaced by ferrite beads later on and a DC blocking capacitor between the stages that was replaced by a zero ohm resistor as well. All of them adding potential problems (probably).
I've chosen a high gain input stage in order to increase the S/N ratio regarding the thermal noise. I've tried to reduce this gain before and increasing the second stage gain but it reduced the S/N ratio. With previous versions, I had three stages and it created high frequency oscillations above the low pass filter cutoff (Rf and Cf). Can you see any potential problems in the PCB layout?

Many thanks
 

Attachments

Hi-Z

Joined Jul 31, 2011
158
Hi, apg2011,

Well, it's been a long time since I last looked at a PCB layout, but I don't see any really glaring problems. However, if it had been me, I would have:

Concentrated on ground-plane integrity - especially under the op-amps.

Given A1 priority to the signal paths and local supply decoupling - and making extensive use of the underside for these, where appropriate.

Relegated power distribution and biasing etc. to lowest priority (they're REALLY unimportant) - basically kept them from affecting how I laid out the signal paths, and from interfering with ground plane integrity around the signal paths.

I wouldn't have reserved the underside for power distribution - it's a waste.

Sorry if I'm sounding a bit critical - hopefully you'll find it useful info for future reference.

Anyway, this isn't solving your problem. I can't help feeling that it's all to do with the huge gain at the first stage. You might like to try the effect of progressively reducing the value of the feedback resistor - down to a minimum of 1k, say. It will be interesting to see how the 33kHz responds.

Incidentally, if you've got access to a decent spectrum analyser (up to a few GHz), it would probably be much more useful than viewing via a 'scope. Then we'd see (I'm assuming) oscillations at high frequencies. (If there aren't any, and 33kHz is the only artefact, then we'll have to think again!)

Keep in touch!
 

Thread Starter

apg2011

Joined Jul 26, 2011
6
Hi Hi-Z,

Many thanks for your suggestions. It looks that the input stage high gain is messing it all. I'm working now in a reduced gain version with a slower op amp and another with a simple feedback resistor for the transimpedance gain and a 50R resistor, in series with the inverting input, in order to set a 50ohm input impedance (through the virtual ground, as the non inverting input is grounded). In this second configuration, the input impedance stays around 50ohms only for low gains and it depends a lot on the op amp I use (all using PSpice). I'll try it later on in the lab.

Regarding the layout, I'll try to follow your advice. I have the bad habit of making fast prototypes with two layers PCBs (as we can get them in 48h from the manufacturer and we don't have to pay any tooling) and probably I should go for a four layer one in order to get a dedicated ground plane and a power plane.

I'll try to use the spectrum analyzer instead (as you suggested) and see the "raw" spectrum from the installed amplifier. It's proving quite difficult to recreate the impedance of a micro-channel plate + anode strip in the lab as I don't have any spare one or a vacuum tank. I'm using the VNA plus a 70dB/50ohms attenuator and a high output impedance AC current source.

I'll let you know how things are going.

Cheers
 

Hi-Z

Joined Jul 31, 2011
158
Hi again,

Sounds like a good idea. There's just one possible fly in the ointment though - it occurs to me that you might be better off with the old, non-inverting configuration for the first op-amp. The reason I say this is that you've got a long piece of coax hanging on the input, and at low/medium frequencies this will look like a lumped capacitor of, say, 500pF. So the feedback path would look like: R(feedback) connected to the op-amp input, which is shunted by 50ohms + 500pF in series.

This will cause phase shift which might cause instability and oscillation. Something to look out for. As I say, I don't think you'll lose noise-wise if you go to a non-inverting configuration (where the coax is well away from the feedback path). But I would try both, if it's convenient.

But I like the idea of a less demanding op-amp. Also, don't worry about 2-sided pcb for simple circuitry like this - MUCH less bother, though you do have to have the right layout strategy.

I'll be interested to see how you get on!
 

Thread Starter

apg2011

Joined Jul 26, 2011
6
Hi,

I've tried the inverting amplifier configuration. In the lab. I measured an input impedance, with the network analyzer, of 1kohm! (not the 50 ohms expected from the input series, resistor the virtual ground and a 22kohms feedback resistor) Despite of this value, I manage to get a bandwidth of 300kHz and no signal reflections with 1uA pulse and 45 meter of URM43 cable. When leaving the cable connected to the amplifier at one end and disconnected at the other, the amplifier became quite noisy but, just by grounding the connector to any ground point of the bench equipments, the noise disappears. I've connected the amplifier into the real detector (through 45 meter of URM43 routed through different walls and machinery) and the amplifier oscillates quite badly at around 330KHz, with uneven amplitude and offset. I'll try to connect the spectrum analyzer to the amp in order to characterize the noise and take some screenshots of the measurement and of the analogue signal. I'll post them as soon as I manage to take them. We are going to check if we have any ground loop issues, as the detector is grounded on its side and the data acquisition system + amplifier are grounded at the other side.

Cheers
 

MrChips

Joined Oct 2, 2009
30,708
45 meter cable seems very long to me. No way you can reduce that or move the preamp closer to the MCP?

I am working on a similar project consisting of four 16 x 24 semiconductor detectors



Each detector has 40 signals for a total of 160 channels.
Oscillation is a definite problem and cable lengths must be kept to a minimum.
 
Last edited:

Hi-Z

Joined Jul 31, 2011
158
Hi apg2011,

Nice to hear from you again, but I'm sorry to hear that you're still having problems.

I have to say I would have preferred the non-inverting configuration, as it keeps the op-amp in a more controlled environment vis-a-vis feedback - also, you would have had a better high-frequency 50-ohm line termination. Does the pulse generator include a termination (i.e. at the originating end)? If so, that would be the reason you're not getting reflections; in a real-life situation I presume you wouldn't have a 50 ohm source, and you'll end up with unwanted reflections. (You should have termination at either the sending end or receiving end - or both.)

As for the grounding situation, it appears that you're picking up interference when you have the free end of the coax un-grounded (I assume you mean that the "noise" disappears when you ground the cable sheath - as opposed to the inner conductor).

I wonder what's so magic about the detector that it causes oscillation? Is it very capacitive, for instance? If so, then switching to non-inverting ought to fix it, I should imagine. Or perhaps it's a very noisy environment?

I think the spectrum analyser should provide the answers - by far the most useful instrument in these situations. It probably would have revealed the true nature of the oscillations - and "noise" - in your original circuit.

Anyway, I'll be very interested in your findings (and just hope I can offer some help!).
 

Thread Starter

apg2011

Joined Jul 26, 2011
6
Hi all,

It's been a loooong time since your replies and I haven't got back to you yet, so I owe you a big apology. Sorry!

Finally, the configuration that worked better is the inverting one (don't ask me why) but with the amplifier at the detector end with "only" 4 metre of URM43 cable between them. I've managed to take fairly reasonable data from the detector, with not too much background noise (in comparison with previous configurations) and good time resolution (see attached picture). I had to modify the detector biasing in order to have a better efficiency, thus a larger signal.
upload_2015-6-5_12-54-8.png

Thank you again for all your comments and advice.
 

Bordodynov

Joined May 20, 2015
3,177
As an amplifier, you can use the op amp inverting inclusion. Thus take input resistor of 50 ohms. The input impedance of the amplifier is equal to the resistor (50 Ohms). The output voltage will be like in a conventional transimpedance amplifier. Unfortunately, I can not now draw a to draw the circuit (only on Monday).
 
Top