Need Help - Ultrasonic AFE

[mzdenkov]

Greetings, im working on this project which is part of masters degree.
Link here:
http://ww1.microchip.com/downloads/en/DeviceDoc/40001902A.pdf

I made everything by their schematic, ok i did use MCP6291R which is 1 OP-AMP in a chip.

My problem is that the filter is oscillating and it gives around  37khz at its output.

Here are my .sch and .brd files

Any suggestions what should i do next?

Thank You

EDIT:
When the filters aren't connected together there are no oscillations on both of them.

[Audioguru again]

It will oscillate if you made it on a solderless breadboard with its stray capacitance between all its rows of contacts and its long wires all over the place. The breadboard and its wires will also pickup interference (compact fluorescent light bulbs operate at around 40kHz). Make it properly on a pcb with some power supply decoupling capacitors and thick ground traces.

[mzdenkov]

The circuit is soldered on one sided pcb, not on solderless bredboard.. but thanks for advice

[StillTrying]

I didn't notice anything obvious between the app note, schem, and PCB layout, it's hard work with the schem. being a mirror image.

I'd try a decoupling cap 0.1uF to 1uF on the Vref connected to the op amps +ve inputs.

[SiliconWizard]

Yeah, I don't even know how the OP managed to get the schematic mirrored.
Maybe fix that first if you want more help.

[ogden]

Well, yes. That mirrored schematic definitely did not help. I immediately noticed few crucial things: 1) mentioned opamp is quite fast - 10MHz GBW 2) PCB do not have ground plane 3) opamps do not have decoupling VCC capacitors. it all may sum-up into parasitic feedback and cause unwanted oscillations. First thing to try - bodge 100nF decoupling ceramic capacitors on top of each opamp, see it  helps or not. If possible - find 1MHz GBW opamps with same footprint or redesign PCB, straighten-up layout meaning do not tangle filter traces.

[mzdenkov]

Hi,

Im sorry for mirrored board in PDF.

Ill try to do that as you all said, to put caps near OPamps. 

Well im limited in space for pcbs, so i cant put ground poligons on board as i make pcb-s at home.
Ill give my best.

here is eagle .brd file. Just remove .txt from the end of the file

[ogden]

Well im limited in space for pcbs, so i cant put ground poligons on board as i make pcb-s at home.

I don't make PCBs often, but when I do - they all have ground plane because need for ground plane is main reason why I make PCBs. It's simple: use two-sided UV presensitized PCB, leave bottom layer with protective film till end - that will be solid ground plane. There are not that much signal traces on bottom of your PCB as well (as usually on mine) - run them using wire on component side, do not even drill associated vias. Drill only ground vias and solder them.

here is eagle .brd file. Just remove .txt from the end of the file

This format is not appreciated here. Many use tablet or even smartphone to read this forum, they can't open application-specific files. Please show some respect to people who spend their time answering your questions by using .png format for drawings (circuit & PCB) and .jpg for photographs.

[coppercone2]

yep solid ground is my method. I really need to try to glue two etched pcbs together

need better method of photoexposure because I think the problem is not alignment, but deformation of the transparency. need to try tracing paper

[mzdenkov]

This format is not appreciated here. Many use tablet or even smartphone to read this forum, they can't open application-specific files. Please show some respect to people who spend their time answering your questions by using .png format for drawings (circuit & PCB) and .jpg for photographs.

Im so sorry i didnt know that.

I made a new PCB and used different op, that is much slower than MCP6292, the filter side still oscillates,
ive added caps between VCC and GND and the problem still stick.


Best Regards

[ogden]

I really need to try to glue two etched pcbs together

Bad idea. You really want to use double-sided PCB's. I like those with photoresist on both sides, but you may apply photoresist film on component side yourself and cover bottom with tape or paint. Perhaps you missed what I said - that I do not etch traces that goes on the bottom, I trace them on the top using wires so bottom is nothing else but plain copper GND plane you never expose to the etch.

I made a new PCB and used different op, that is much slower than MCP6292, the filter side still oscillates,
ive added caps between VCC and GND and the problem still stick.

How unfortunate. I blame layout where output of 2nd opamp is close to input of first. When you short input of the first filter (r11 to the gnd) - does it still generate? Your PCB must clean, w/o any residue.

[mzdenkov]

How unfortunate. I blame layout where output of 2nd opamp is close to input of first. When you short input of the first filter (r11 to the gnd) - does it still generate? Your PCB must clean, w/o any residue.

When inverting input of the first opamp is shorted to the grond the whole circuit stops oscillating...
On the R11 there is small oscillation before shorting, and i think that amplifies a lot at the end.

Ive attached the new version, it aint much, but i thought that it will work

[ogden]

Try to decrease gain of one or both filter stages. It may fix feedback problem and at least let you proceed further with project. You definitely shall straighten-up component placement such a way that input is on the one end of PCB and output - on the other.

Useful tool: https://www.analog.com/designtools/en/filterwizard/

[StillTrying]

You definitely shall straighten-up component placement such a way that input is on the one end of PCB and output - on the other.

I think that half the time, but in listening mode the US transducer is the actual input.

When inverting input of the first opamp is shorted to the ground the whole circuit stops oscillating...

Not too surprising, but if it never oscillates with just T1's base-emitter shorted I think that would be good sign.

If you''re using the transmitting signal are you sure it switches off after a few cycles.

[SiliconWizard]

And what happens if you short TRND+ and TRND-?

[mzdenkov]

Trnd+ and Trnd- is the transciever connection...
As i can see the noise is generated at filter part.
Basically nothing comes from  the  transducer driver and A cass amp