EMG amplifier circuit acting highly unstable and often lying at rails

Thread Starter

Michael Adventure Hopkins

Joined May 14, 2019
2
I have a design for an EMG (electro-myograph) that needs some stability assistance. The core components are an LT6370 in-amp, and an OPA2192 dual op-amp package for amplification and generating a stable reference. The reference is set by a capacitance multiplier that seems solid as a rock when probed. The in-amp is set to ~1.5 X and the op-amp (not as ref) is set to ~ 30 X in a Sallen-key configuration to snuff high f noise. The reference is pushed to the patients skin as well in a driven right-leg configuration. This is of course just on a bench top right now so no confounding lead attachment issues.
The circuit occasionally acts properly, but mostly finds itself sitting at the ground rail. In shorting the input leads of the in-amp I expect 2.485V out of my in-amp and thus of course 0 out of my 2nd amplifier.
Am I missing something easy? LTSpice shows this operating great (of course it would!) Any ideas on what to check? I'm flummoxed.
EMG.png
 
Last edited by a moderator:

TeeKay6

Joined Apr 20, 2019
572
I have a design for an EMG (electro-myograph) that needs some stability assistance. The core components are an LT6370 in-amp, and an OPA2192 dual op-amp package for amplification and generating a stable reference. The reference is set by a capacitance multiplier that seems solid as a rock when probed. The in-amp is set to ~1.5 X and the op-amp (not as ref) is set to ~ 30 X in a Sallen-key configuration to snuff high f noise. The reference is pushed to the patients skin as well in a driven right-leg configuration. This is of course just on a bench top right now so no confounding lead attachment issues.
The circuit occasionally acts properly, but mostly finds itself sitting at the ground rail. In shorting the input leads of the in-amp I expect 2.485V out of my in-amp and thus of course 0 out of my 2nd amplifier.
Am I missing something easy? LTSpice shows this operating great (of course it would!) Any ideas on what to check? I'm flummoxed. View attachment 198084View attachment 198084
The MMBT3904 is a general purpose NPN switching transistor, not a MOSFET as you have shown; it serves no useful purpose here. I have attached a circuit/simulation-file, copied/modified from yours, with the MMBT3904 removed. I don't have a model for the OPA2192 opamp you used, so I have arbitrarily substituted an LTC2050. I have added 100k resistors to simulate skin resistance. I agree that I see no problem in the LTspice simulation.

I do suggest that a more stable reference than the +5V power supply may be desirable. There are numerous low cost devices available to provide a 2.5V reference (e.g. TL431, TL431LI).

Just in case it may be relevant, I offer this link (found accidentally) wherein someone complains about the OPA2192 input offset: Prob with OPA2192
 

Attachments

Last edited:

Alec_t

Joined Sep 17, 2013
11,646
You are running the OPA from a 5V supply , which is very close to the minimum rated one. That may affect its performance.
 

Thread Starter

Michael Adventure Hopkins

Joined May 14, 2019
2
Thank you to you both. I'll up the power on the rails and see if I get what I'm looking for, otherwise.. I'll try another chip. Admittedly, I went overboard on these parts for signal to noise and neglected to account for the edge I was walking on. I'll update this post as I try further solutions.
 

DonQ

Joined May 6, 2009
321
Read the section in the LT6370 spec sheet about "Input Bias Current Return Path". The results you are seeing are exactly what this problem would create.

(When in doubt, read the instructions.)
 

TeeKay6

Joined Apr 20, 2019
572
Read the section in the LT6370 spec sheet about "Input Bias Current Return Path". The results you are seeing are exactly what this problem would create.

(When in doubt, read the instructions.)
Assuming the T.S. is using the circuit as generally used, there is a well-defined path for input bias currents. Each input is attached to (in contact with) test-subject's skin at test locations and the Vref potential for the LT6370 is connected to test-subject's skin at "left leg". This does raise the question of what T.S. means by "This is of course just on a bench top right now so no confounding lead attachment issues." If by that statement, T.S. means the inputs are simply not connected, then you are undoubtedly correct. Also, as Alec_t noted, 5V is very close to the lower limit for spec'd operation. If the 5V is a bit low, then malfunction could be expected.
 
Top