7

What are the effects we need to consider when transmitting extra low voltage (0.1 uV) signal over about 1 meter of cable, either coaxial or shielded twisted pair ?

The source is low impedance (~5 ohm), and the receiver would be high impedance amplifier input. The signal is about 10kHz in bandwidth, in the 10 kHz to 50 kHz range.

Cable impedance matching is not an issue at these frequencies, for transmission line effects the cable is probably too short.

I would guess that shielding efficiency would be a primary concern ? would triboelectric noise be significant in these conditions ? Anything else ?

EDIT: The context for the question is working out the noise budget for a complete signal path (physical medium -> sensor -> preamp -> cable -> main amp -> ADC -> digital signal processing gain), to aid in exploring the trade-off of placing the cable after preamp vs after the ADC (in the digital domain).

vadimus
  • 83
  • 5

2 Answers2

8

Shielding efficiency (unless balanced) and triboelectricity are certainly concerns. Balancing mitigates shielding efficiency, and there are cables developed for microphone connections designed to reduce triboelectric effects.

But so is simple noise - Johnson noise, the thermal noise in any resistance.

(Shot noise, the statistical variation in current would be significant at high impedances, but can be ignored here).

Assuming the bandwidth is 10 kHz to 50 kHz (40 kHz BW) and the receiving amplifier is moderately low noise, like 1 nV/sqrt(Hz), its input noise contribution would be 200 nV or twice the signal amplitude.

The 5 ohm source impedance itself contributes 0.28 nV/sqrt(Hz) or 56nV, more than half the signal amplitude.

If the lower frequency limit is 10Hz (not kHs) you'll also have 1/F noise (aka flicker noise) to contend with.

Do you have any signal/noise ratio requirements?

And is there any way you could place a step up transformer (impedance converter) at the source end?

Say, 1:4 in voltage, 1:16 in impedance? That would give an 80 ohm source impedance or 1.1 nV/sqrt(Hz) noise floor, giving you a fighting chance of amplifying it with a noise figure of 2-3 dB.

A transformer will increase both the signal and the source impedance's own noise (56 nV. giving nearly 6 dB SNR). What you gain is that both are now larger than the amplifier's own noise contribution.

A really good amplifier (using discrete PNP transistors) can approach 0.5 nV/sqrt(Hz) which is still about 6dB above your inherent noise level without the transformer, giving 0.1 uV rms noise in 40 kHz, which would degrade your SNR to 0dB. (I've never seen ICs better than the 0.7 to 0.8 nV/sqrt(Hz) range)

But after a 4:1 step-up transformer, this noise is added to 0.4 uV signal and 0.224 uV (56 nV * 4) noise.

Sqrt(0.1 ^2 + 0.224 ^2) = 0.245 uV or only about 1 dB worse; you can say the amplifier has a 1 dB noise figure with the transformer, or about 6dB without it.

(Side note : with transformers for source impedance conversion, vacuum tube based mic amps can still approach the state of the art)

EDIT following question edit : you can then (slightly) improve S/N ratio using a priori knowledge of the signal frequency (per Andy's leading questions).

If you can accurately (or fairly accurately) know or predict the signal frequency, and its amplitude is teh quantity of interest, there are signal processing techniques you can use to extract it from broad band noise. (Useful search terms : PSD or phase-sensitive detector, or lock-in amplifier, for the traditionalist. Now just digitise the lot; FFT it, and analyse the frequency bins of interest).

Absent that, you can filter the 10kHz band of interest - after the low noise amplifier, so the filter's own noise is insignificant. By selecting 1/4 of the original spectrum you can hope to improve SNR by 6 dB.

  • I'd hold off on accepting this : I can see where Andy was going, he may want to answer too... –  May 09 '20 at 14:57
  • thanks for pointing out the source impedance as the noise source. Tried to work through the numbers with the step up, and i don't see how it actually improves the noise figure: seems that both signal and noise floor are amplified at the same rate. – vadimus May 09 '20 at 15:01
  • Right : you stil have the source impedance's own noise ( 56 nV. giving nearly 6 dB SNR). But by choosing a step-up ratio, you can make the receive amplifier's own noise contribution as small as you like. A really good amplifier may approach 0.5 nV/sqrt(Hz) which is still about 6dB above your inherent noise level. Will expnad the answer. –  May 09 '20 at 15:09
  • Thanks, that clarifies it. It seems most of this would still be relevant when simply connecting the amplifier without the cable, say with short PCB trace. So I take it there should be no significant cable-specific noise sources as long as it is balanced ? – vadimus May 09 '20 at 15:27
  • Balanced, screened, and suitable cable per answer. We still don't know if you have a SNR target to meet, or if whatever physics permits is good enough. –  May 09 '20 at 15:28
  • I also don't know this specific SNR target yet :) I'm working out the noise budget for a complete signal path (physical medium -> sensor -> preamp -> cable -> main amp -> ADC -> digital signal processing gain). Also exploring the trade-off of placing the cable after preamp vs after the ADC (in the digital domain). Extra thanks for mentioning discrete BJT vs IC – vadimus May 09 '20 at 15:44
  • No, I don't want to answer @Brian +1 – Andy aka May 09 '20 at 17:17
  • re: PNP transistors. Keep an eye open for Hitachi 2SA1085. –  May 09 '20 at 18:51
2

Not having worked (professionally) at such low voltages, tho I have with experiments without good equipment to characterize the results, I'm concerned about

(1) trash entering the cable thru weaves in the copper braid ---- so use SOLID SHIELD coax cable.

(2) trash entering the cable, because either the shield or the center conductor are useful return paths for external charge pulses/sines that will of course explore all possible paths proportional to susceptance ( 1 / impedance).

  • With a 5_ohm Rout, the center conductor may be a fine path

  • With nominal DC impedance of zero, the cable's shield will certainly be a fine path

  • Thus use heavy braid between the signal_source GND and the destination circuit GND.

(3) having carefully read many posts on diyAudio.com, the best systems have the power supply located at a DISTANCE, with only DC power cables entering the system. I think TOPAZ still makes multi-shielded power transformers, to further reduce primary-secondary coupling. I recall TOPAZ transformers being specified for use in equipment to be sited inside Nevada tunnels during testing of the "package".

The diyAudio.com people were amplifying 100 uV signals from Moving Coil cartridge/needle sensors atop vinyl audio records, seeking 100dB SNR. With 100dB below 100uV being 100uV/100,000 the floor is 1 nanoVolt.

Thus the single-tone (60Hz, 120Hz) leakage into the audio system was below 1nanoVolt.

Notice such systems have a FIFTH WIRE (in addition to left/right coaxes with 2 wires in each), that FITH WIRE used to strap the turntable chassis to the preamp chassis. I personally have seen that FIFTH WIRE (bare copper multi-strand wire 1/16th diameter) reduce the 60Hz from 30uV to less than 1uV, in a workshop near an airconditioner, with various imbalanced power cables all over the workbench.

analogsystemsrf
  • 33,703
  • 2
  • 18
  • 46