noises and frequencies

[Ogitek]

If the source resistance increased from 100 ohm to say 15,000 ohm. How do you compute what the noise of 8uV would become? And how is noise related to the input resistance/impedance.

Its not going to get simpler than this:

Suggested starting points:
https://www.ti.com/lit/SBOA345

I read it. But it's not mentioned how to compute it.

If 8uV noise is produced from 100 Ohm input resistance (input impedance).
What will 15,000 Ohm input resistance produce? 15,000/100 = 150uV noise?

[magic]

Use this for Johnson noise. Use manufacturer's specs () for voltage noise and current noise of the amplifier, multiply current noise by source resistance for equivalent voltage noise contribution. RMS sum all three as instructed before. Vpp is ~7x more than Vrms.

In absence of detailed specs from the manufacturer, if you know total noise in 100Hz bandwidth with 10kΩ source, going to 1000Hz with 10kΩ source should be a matter of multiplying by sqrt(1000/100) if noise is mostly "white". Johnson noise is white, and amplifier's noise hopefully is too over almost all its bandwidth (except very low frequencies, but you seem to be mostly concerned with varying the BW at the top end). Increasing source resistance typically increases total noise anywhere between sqrt(X) and X, where X is the ratio of increase (e.g. 15k/10k = 1.5). Decreasing source resistance decreases similarly, until you reach a limit imposed by voltage noise of the amplifier. In your case this appears to already happen near 10kΩ, so going lower is unlikely to make things better.

And of course read the basics. You're getting nowhere without understanding what those words mean.

[Ogitek]

Use this for Johnson noise. Use manufacturer's specs () for voltage noise and current noise of the amplifier, multiply current noise by source resistance for equivalent voltage noise contribution. RMS sum all three as instructed before. Vpp is ~7x more than Vrms.

In absence of detailed specs from the manufacturer, if you know total noise in 100Hz bandwidth with 10kΩ source, going to 1000Hz with 10kΩ source should be a matter of multiplying by sqrt(1000/100) if noise is mostly "white". Johnson noise is white, and amplifier's noise hopefully is too over almost all its bandwidth (except very low frequencies, but you seem to be mostly concerned with varying the BW at the top end). Increasing source resistance typically increases total noise anywhere between sqrt(X) and X, where X is the ratio of increase (e.g. 15k/10k = 1.5). Decreasing source resistance decreases similarly, until you reach a limit imposed by voltage noise of the amplifier. In your case this appears to already happen near 10kΩ, so going lower is unlikely to make things better.

And of course read the basics. You're getting nowhere without understanding what those words mean.

Ok. So increasing source resistance really increase total noise anywhere. If the spot noise at 100 Ohm input impedance is 8uV (as actually measured by the manufacturer). The noise at 10,000 Ohm is really (10,000/100) = 100X and multiply it by 8uV = 800uV or 0.800mV or almost 1 mV! This can totally drown the 1mV signal!

But something puzzling. In the unit (see pdf spec I shared before). Their known specs is 2 uV peak-to-peak, DC to 100 Hz with 10k Ohm source.

Yet the real noise at 1000 Hz with 10k source impedance appears to be 800uV or 0.8mV (see first sentence), and it becomes 8uV at 100 Ohm only source impedance (instead of 10kOhm). Something is not right.

Note projecting the original 2uV at 100 Hz in the spec sheet to 1000 Hz (with similar 10k Ohm input impedance as your described) would produce sqrt (1000)/sqrt 100= 31.6/10 = 3.16 times or noise of 2uV x 3.16 = 6.32 uV only. Yet the real noise appears to be 800uV in actual. What happened?

Are the computations correct, magic?

Ohs. David Hess. Just help build one for me, $300   It's ok if no fancy knobs and only potentiometers and no dc offset adjustments. The latter is only moving the scale up 0V and not affecting the waveforms.

[donlisms]

Yes.  Yes you do measure audio or signal over a bandwidth.

I don't think you are trying to understand the things everyone has taken the time to answer for you. That's not fair.  If you are a student, you have to work to understand it for yourself.  You're not doing that.

[magic]

Note projecting the original 2uV at 100 Hz in the spec sheet to 1000 Hz (with similar 10k Ohm input impedance as your described) would produce sqrt (1000)/sqrt 100= 31.6/10 = 3.16 times or noise of 2uV x 3.16 = 6.32 uV only. Yet the real noise appears to be 800uV in actual. What happened?

No, the only real 1kHz measurement that you have is 8μVpp (with 100Ω impedance). Which is close enough to 6.3μVpp, I guess.

Again, it appears that anything less than 10kΩ isn't making much difference for this device, so extrapolating from a few μV at 100Ω to hundreds μV at 10kΩ is invalid. Most likely 100Ω is just as bad as 10kΩ, which is just as good as 100Ω, if that makes sense.

[Ogitek]

Note projecting the original 2uV at 100 Hz in the spec sheet to 1000 Hz (with similar 10k Ohm input impedance as your described) would produce sqrt (1000)/sqrt 100= 31.6/10 = 3.16 times or noise of 2uV x 3.16 = 6.32 uV only. Yet the real noise appears to be 800uV in actual. What happened?

No, the only real 1kHz measurement that you have is 8μVpp (with 100Ω impedance). Which is close enough to 6.3μVpp, I guess.

Again, it appears that anything less than 10kΩ isn't making much difference for this device, so extrapolating from a few μV at 100Ω to hundreds μV at 10kΩ is invalid. Most likely 100Ω is just as bad as 10kΩ, which is just as good as 100Ω, if that makes sense.

Ok, so you are saying the amplifier has noises with values akin to 10k Ohm Johnson noises so anywhere below 10k Ohm won't improve it. Yes, I'm taking the ideas of donlisms into account where you have to measure over a bandwidth.

Whatever, the device official noise spec is 2 uV peak-to-peak, DC to 100 Hz with 10k Ohm source.  I have a Brainmaster EEG and also an EEG signal generator in my AI with emotions lab I'm setting up. I set the generator to 10 microVolt and it is what Brainmaster is displaying in the software. Now if the noise is say 2 microVolt. How much would it make the 10 uV waveform wiggle? What is the rule of the percentage of noises before it is negligible. 2uV noise/10uV signal is 20% Is it bad?

Also when  you mentioned  "if noise is mostly "white". Johnson noise is white, and amplifier's noise hopefully is too over almost all its bandwidth (except very low frequencies)". Did you mean the "very low frequencies" as 100 Hz? and when you said "except". Did you mean that at DC to 100 Hz with 2uV noise and 10K ohm source. If I lower the source impedance to 1k, the noise would go down from 2uV to nV? Something that doesn't occur for higher frequencies? 

[David Hess]

I used an impedance checker. My electrode and wire to skin impedance is about 10k to 20k. The test at 8uV noise was at 100 Ohm. If the source resistance increased from 100 ohm to say 15,000 ohm. How do you compute what the noise of 8uV would become?

The increase in resistance, 14,900 ohms in your example, creates two new sources of noise, the Johnson noise of the resistance and added voltage noise from the input current noise multiplied by the resistance.  In the example I gave, the voltage noise from the input current noise through the resistance can be ignored because the current noise is so low, which should always be the case with a 1 or 10 megohm input.  The Johnson noise dominates in this case.

That leaves the Johnson noise which is equal to 0.13*Sqrt(resistance*bandwidth) nVrms, which would be added to the 8uVrms.  Over 10kHz, that will be 1600nVrms or 1.6uVrms, and 8uVrms+1.6uVrms=8.16uVrms.

And how is noise related to the input resistance/impedance. The unit has input impedance of 10 megaohms (see last message for the full spec sheet). If you'd use the voltage divider concept. The 100 ohm vs 15,000 ohm would not have much significant against the 10 megaohm input impedance. But how about noise? Why is noise more affected by increase in the input impedance.

The input impedance of the amplifier does not really matter, except insofar as an amplifier with a high input impedance must have a low input current, which requires using parts and a design which has higher input voltage noise.  When modeled, the input impedance is a resistance across the input.  This resistor contributes its own noise, however since the source impedance is across the input also, it effectively shorts that resistance and noise out.

Also stand alone EEG, ECG, EMG units have fixed frequency. In the UFI unit, it is adjustable from 20, 50, 100, 1000, 5000, 10000 Hz. What parts of the circuits need adjustments for these switches? And would they introduce more noises? Why. Thanks!

They probably use a mechanically switched lowpass filter for that, which requires switching various combinations of resistance and capacitance into the circuit after amplification.  All resistances and active elements add noise, however this can be ignored because the noise level in the circuit after amplification is much higher so later contributions are negligible.

[David Hess]

Ohs. David Hess. Just help build one for me, $300   It's ok if no fancy knobs and only potentiometers and no dc offset adjustments. The latter is only moving the scale up 0V and not affecting the waveforms.

I think you would be better off finding a working Tektronix AM502 amplifier.  I do not think I could do justice to a project like you require for a reasonable price.  The best I could do is to test and sell you a Tektronix AM502 and TM500 power supply mainframe, but I could not replace mine for that price right now.  AM502s do not show up at bargain prices on Ebay very often.  There is this one currently which is exactly what you need, but who knows if it is working, and it has some physical damage.

As an aside, several years ago I spent more than a year looking for an AM502 on Ebay and by the time I got one, I had received two or three 7A22s as part of 7000 plug-in lots.  The 7A22 is the same thing as an AM502 but lacks an external output for use with other oscilloscopes.  The AM502 that I got just needed to have its switches cleaned and lubricated.

[Ogitek]

Ohs. David Hess. Just help build one for me, $300   It's ok if no fancy knobs and only potentiometers and no dc offset adjustments. The latter is only moving the scale up 0V and not affecting the waveforms.

I think you would be better off finding a working Tektronix AM502 amplifier.  I do not think I could do justice to a project like you require for a reasonable price.  The best I could do is to test and sell you a Tektronix AM502 and TM500 power supply mainframe, but I could not replace mine for that price right now.  AM502s do not show up at bargain prices on Ebay very often.  There is this one currently which is exactly what you need, but who knows if it is working, and it has some physical damage.

As an aside, several years ago I spent more than a year looking for an AM502 on Ebay and by the time I got one, I had received two or three 7A22s as part of 7000 plug-in lots.  The 7A22 is the same thing as an AM502 but lacks an external output for use with other oscilloscopes.  The AM502 that I got just needed to have its switches cleaned and lubricated.

Some Bioamplifiers have Isolators built in. The spec like:

Maximum isolation voltage    700 volts, AC
Typical leakage current         <5 uA at 240 VAC

If you would just use a battery like 12 Volts on Bioamplifier without isolators. It can't shock or electrocute you when the electrodes are all over your body. Isn't it. So I wonder what is the advantage of models with Isolators. Could they help with noises?

Since my oscilloscope has 10mV vertical scale. And since I owe one of the most noise free ADC around (the E1DA Cosmos ADC). I asked if I could use REV software to check for the microVolt noises. They said it was possible (someone shared the image below). Have you got any experiences on this? The ADC may have some lingering noises. But at least you can see the signals than using oscilloscope with 10mV resolution. I want to check noises of some cheap $20 modules I got.

[David Hess]

Some Bioamplifiers have Isolators built in. The spec like:

Maximum isolation voltage    700 volts, AC
Typical leakage current         <5 uA at 240 VAC

If you would just use a battery like 12 Volts on Bioamplifier without isolators. It can't shock or electrocute you when the electrodes are all over your body. Isn't it. So I wonder what is the advantage of models with Isolators. Could they help with noises?

Medical stuff uses galvanic isolation for safety, but there can be some noise advantages, and it makes connecting different pieces of equipment together without ground loops and common mode problems easier.

Since my oscilloscope has 10mV vertical scale. And since I owe one of the most noise free ADC around (the E1DA Cosmos ADC). I asked if I could use REV software to check for the microVolt noises. They said it was possible (someone shared the image below). Have you got any experiences on this? The ADC may have some lingering noises. But at least you can see the signals than using oscilloscope with 10mV resolution. I want to check noises of some cheap $20 modules I got.

The usual method is to make a standard deviation calculation with the input shorted, or connected to a suitable resistance, which returns the RMS noise.