1

I am designing a PCB based on the ADS1299 chip from TI, which is essentially an ADC with amps and some other functionalities.

I have seen a lot of emphasis on differential inputs into an amplifier to be symmetrical, as in their path being symmetrical. I assume that this will improve the common mode noise rejection ratio.

As a connector, I was thinking of having a 2-row, 8-pin header to save some space, for the analog input into the PCB, however this might cause slight asymmetry. Is it better to go with a 1-row header?

Further info: I am trying to measure EEG, which depending on the rhythm, would be in the range of 10's of microvolts, and from 0 to 100 Hz.

ocrdu
  • 8,705
  • 21
  • 30
  • 42
NeuroEng
  • 462
  • 2
  • 13
  • 2
    symmetrical won't be nearly as important as having the differential inputs live together in twisted pairs, so that most of the EM noise is common mode. – Scott Seidman Oct 05 '22 at 17:30
  • I don't understand the "2 rows might be less symmetric than 1 row" argument: intuitively, the opposite should be true, right? – Marcus Müller Oct 05 '22 at 17:30
  • If the signals are coming in on "near" and "far" pairs, then one trace would be longer, but keeping pairs on the same row would mean routing one pair in between the pins of another which is probably worse. There are really fine pitch single row headers out there; I think you can have your cake and eat it too. – vir Oct 05 '22 at 17:35
  • 2
    @vir length mismatch below 10m will not make much of a difference here; consider the wavelengths of a 16 kHz signal. Noise immunity is given by keeping the conductors together, not necessarily closely at the same length – Marcus Müller Oct 05 '22 at 17:44
  • @vir re connector: I'd probably just, for cost reasons, recommend modular 8P8C connectors. Two of them next to each other is really not much space, and there's readily available dual-RJ45 connectors for most board mounting methods. – Marcus Müller Oct 05 '22 at 17:45
  • 1
    True, I think mechanical fit and ease of routing are the main factors here. I would probably not recommend 8p8c connectors since there's always the possibility of someone plugging a 48V passive PoE cable into a "10s of mV" circuit. – vir Oct 05 '22 at 17:48
  • @vir also a good point, though I assumed this was more internal – if this really goes out to a patient's head, pin headers, RJ45 … won't do, for reliability reasons already. – Marcus Müller Oct 05 '22 at 18:08

1 Answers1

3

I am trying to measure EEG, which depending on the rhythm, would be in the range of 10's of microvolts, and from 0 to 100 Hz.

I have seen a lot of emphasis on differential inputs into an amplifier to be symmetrical, as in their path being symmetrical.

Is it better to go with a 1-row header [compared to a 2-row header] ?

1-row header or 2-row header shouldn't make a lot of difference.

The pick-up of the differential noise is proportional to the loop area. That's the only thing worthy of a concern. The loop area between the header pins should be small enough that you don't have to worry about it. Compare it to the untwist and asymmetry between the electrodes on the patient end of the cable. [The author of this post got away with a lot worse.]

The impact of the "geometrical" asymmetry on the signal integrity caused by the connector pins is negligible. It's only a 100Hz signal. [Maybe an 8kHz signal, if you consider the 16ksps max sampling rate. But it's not a 100MHz signal.]

Nick Alexeev
  • 37,739
  • 17
  • 97
  • 230