Comments on Changing PCB trace width once signal-to-noise ratio is high
Parent
Changing PCB trace width once signal-to-noise ratio is high
I'm designing a PCB to filter and amplify a differential body signal with amplitude 5 mV up to 5 V. The first part of the circuit consists of a first order highpass filter to remove any DC components from the signal. Then it moves into an instrumentation amplifier for amplification of the differential signal and attenuation of the common mode voltage. The instrumentation amplifier has a common-mode rejection ratio of 120 dB. The first stage of my circuit is seen below.
The signal-to-noise ratio before the instrumentation amplifier is
I want to maintain the signal integrity of my differential signal. To do this, there should be as little voltage drop of my signal across any impedances on the path to the amplifier as possible. A
My question is: After the amplification stage, the signal now has a 5 V amplitude, and the SNR is much greater. To what trace width am I allowed to go down to? Routing with 1 mm trace width on the entire board is not doable. Would a trace width of 0.2 mm be appropriate? Is there anything I should be aware of when changing the trace width? The board is entirely analog. Not digital circuitry.
Post
None of these ideas optimize the SNR that you need.
You need to design with error correction or (suppression);
- 50V/m grid E-field (Right Led Drive RLD) >100 dB suppression
- AM radio noise (LPF > -40dB @ 1MHz)
- Galvanic skin response (motion artifacts) (firm electrode strain relief)
If you search for front-ends on EKG/ECG/EMG solutions you will find the best use the common-mode signal fed back to create a CM gnd somewhere on the body not moving much, away from the sensors. This becomes your signal ground but using good medical* isolated supplies or batteries will not conduct much (uA) current. *=low leakage rated/approved.
This can be done with a single supply using an INA ensuring that Vref is near V+/2.
You will never get 120 dB CMRR if you use 1% passives in the front-end and you net a 1% DM/CM error that's a 20 dB CMRR. This is why INA's are used with laser-trimmed errors.
1 comment thread
1 comment thread