preamp noise

[green]

Could someone calculate peak to peak noise I should expect to see with a scope(1ms/div)at TP3?

[Polymer]

Could someone calculate peak to peak noise I should expect to see with a scope(1ms/div)at TP3?

Are you having noise problems ... again or still ... we definitely share that problem if I remember right.
I am following through on all this noise stuff, it takes time and a lot to learn.
Do you live in a generally humid environment? Alabama does sound kinda swampy, at least from what I've seen in non-documentary films. But that just might be bollocks.

Are you using smd resistors?

Here a chart that might be helpful to you as I have been studying resistors in depth in my efforts to mitigate my noise problems.

[green]

Are you having noise problems ... again or still ... we definitely share that problem if I remember right.
I am following through on all this noise stuff, it takes time and a lot to learn.
Do you live in a generally humid environment? Alabama does sound kinda swampy, at least from what I've seen in non-documentary films. But that just might be bollocks.

Are you using smd resistors?


Here a chart that might be helpful to you as I have been studying resistors in depth in my efforts to mitigate my noise problems.

[ATTACH]45084[/ATTACH]

I calculated what I thought the p-p noise should be. Not sure I did it right. I live in northern Alabama, not swampy where I live. Noise from coil pickup is normally higher than shorted input but noise can't be less than shorted input. Trying to learn something if I'm doing the calculation wrong.

[Carl-NC]

I get about 4.7mVpp.

[green]

I get about 4.7mVpp.

Thanks for the reply. I got 21mv so I must be doing something wrong. 1k input resistor(4nv/rt-Hz), NE5532(5nv/rt-Hz) BW10MHz. (4E-9^2+5E-9^2)^.5=6.4nv/rt-Hz. 6.4nv/rt-Hz*6=38.4p-p. 38.4nv p-p*1000(amplifier gain)=38.4uv p-p. 38.4uv p-p*(10E6/33)^.5[amplifier BW^.5]=21mvp-p. What did I do wrong?

[Skippy]

Green: Your maths is awful to read, I've got wrinkles in my forehead that will take hours to disappear.
Please: Break it up into seperate lines.
Put your comments in square brackets, not a random mix of square and curved.
Use a capital V for Volts
Use sup and sub to get the powers to appear properly:
Example: 10[ sup ]-6[ /sup ] without the spaces will result in 10^-6
Likewise V[ sub ]A[ /sub ] becomes V_A

[green]

I'm thinking input resistor(4nV/rt-Hz) and NE5532(5nV/rt-Hz)noise cause most of the noise.
Input noise=((4nV/rt-Hz)^2+(5nV/rt-Hz)^2))^.5=6.4nV/rt-Hz
Output noise=input noise*amplifier gain of 1000=6.4uV/rt-Hz
NE5532 BW/amplifier gain=circuit BW, 10MHz/33=300kHz, Output noise=6.4uV/rt-Hz*300kHz^.5=3.5mv
3.5mV*6=p-p noise=21mVp-p
Probably still not right but hopefully fewer wrinkles. Maybe good enough to see what I'm doing wrong.

Maybe someone with a MPP could measure TP3 with a scope to see how much p-p noise there is. 1k input resistor connected to 0V.

[Carl-NC]

I made a gaffe in my calcs, used the wrong GBW. My revised calc is 22.7mV. I used:

- 4nv/rtHz for the 5532
- a 1.22 mult factor on the noise BW since it's a 2-pole roll-off
- a p-p factor of 6.6

So your number is correct.
What do you measure?

[KingJL]

I'm thinking input resistor(4nV/rt-Hz) and NE5532(5nV/rt-Hz)noise cause most of the noise.
Input noise=((4nV/rt-Hz)^2+(5nV/rt-Hz)^2))^.5=6.4nV/rt-Hz
Output noise=input noise*amplifier gain of 1000=6.4uV/rt-Hz
NE5532 BW/amplifier gain=circuit BW, 10MHz/33=300kHz, Output noise=6.4uV/rt-Hz*300kHz^.5=3.5mv
3.5mV*6=p-p noise=21mVp-p
Probably still not right but hopefully fewer wrinkles. Maybe good enough to see what I'm doing wrong.

Maybe someone with a MPP could measure TP3 with a scope to see how much p-p noise there is. 1k input resistor connected to 0V.

green,
Rather than to speculate or assume where your solution might be flawed, I will reference two sources from Linear Technologies that have helped me in the past with understanding total noise:
https://www.analog.com/media/en/refe...tes/dn015f.pdf
https://www.analog.com/media/en/tech...ets/1028fd.pdf

The second reference details the calculations on a amplifier circuit with a gain of 1000 (same as your example).
I hope this helps with your quandry!

[Skippy]

Not that it's going to make a huge change to your maths, but the 1st-stage op-amp input noise is multiplied by 34 x 34 = 1156.
I usually estimate pk-pk noise as 7 x RMS , though I've seen some sources say 8 times ... it depends on whether it's true Gaussian noise or not, and it's subjective, how many of the big spikes you ignore.

[green]

Thanks for the replies. Don't have a MPP so I can't measure the noise. Mainly I wanted to see if I was doing it right(close enough?). Used the MPP as an example calculation. Started a thread few years back,(PI noise)looking at integrator out. Got a lot of suggestions on how to lower noise. Don't remember if anyone suggested the noise level is about as low or maybe 10 times higher than it should be. If I had a MMP and saw 25mV p-p at TP3 with input at 0V I would know noise level is as low as it should be. Have wondered if I could calculate preamp input noise starting with p-p output noise. p-p output noise/6/circuit BW^.5/amplifier gain=input noise. An indication of a noise problem? Still don't know how to calculate expected p-p noise at integrator out. Noise is expected, knowing how much is important.

[Ferric Toes]

I have the same preamp circuit in my viscosity meter, even to using a 5532. The only difference is that I have a total gain of x470 and I roll off the response >200kHz with a small capacitor across each feedback resistor. Tomorrow I will short the input to ground and measure the peak to peak output noise. My understanding is that much of the IC noise that affects a PI RX circuit, is that which occurs below the 1kHz which is specified as 5nV rt.Hz. At 10Hz it is 16nV rt.Hz.according to the Texas data sheet. There are wide band IC's that are much lower e.g the LT1028 with 1nV rt.Hz at 10Hz, as in the data sheet in KingJL's post.

Eric.

[KingJL]

Still don't know how to calculate expected p-p noise at integrator out. Noise is expected, knowing how much is important.

That my friend is much more complicated than calculating the preamp noise out. Many factors are involved: the integrator architecture involved, the charge path cut off frequency (which determines stop band), the output cutoff frequency (which determines pass band), the sample rate and sample width (which will determine integrator noise attenuation/SN ratio/noise floor). Example, having built a couple of HH's, I always felt that the integrator was on "the noisy side" and could be improved. But the HH was designed as a learning platform and is extremely good for that purpose. I attribute the bulk of what I have learned and now understand about PI detectors to the HH platform. The MPP integrator is somewhat more quiet (1000 fold fold improvement in stop band). The GS IV is better yet with over a 3 fold improvement in stop band over the MPP (3000 fold improvement over the HH). Prior to my experience with the VMH3CS, I had always considered the GS IV integrator as close to ideal for a PI as was achievable in the real world with all the trade offs involved.

[green]

That my friend is much more complicated than calculating the preamp noise out. Many factors are involved: the integrator architecture involved, the charge path cut off frequency (which determines stop band), the output cutoff frequency (which determines pass band), the sample rate and sample width (which will determine integrator noise attenuation/SN ratio/noise floor). Example, having built a couple of HH's, I always felt that the integrator was on "the noisy side" and could be improved. But the HH was designed as a learning platform and is extremely good for that purpose. I attribute the bulk of what I have learned and now understand about PI detectors to the HH platform. The MPP integrator is somewhat more quiet (1000 fold fold improvement in stop band). The GS IV is better yet with over a 3 fold improvement in stop band over the MPP (3000 fold improvement over the HH). Prior to my experience with the VMH3CS, I had always considered the GS IV integrator as close to ideal for a PI as was achievable in the real world with all the trade offs involved.

Not familiar with GS IV integrator. I've been using a 1C integrator with normal and invert amplifier outputs. Need to zero both normal and invert outputs to eliminate high frequency. I have thought noise above 1kHz is causing noise in pass band, aliasing on multiples of Tx frequency. Maybe not? http://www.geotech1.com/forums/showt...070#post219070 an example of integrator out I did awhile back. Got it a little better but zapped the circuit.

That my friend is much more complicated than calculating the preamp noise out. Agree, did you ever get where you thought you figured it out?

[KingJL]

Not familiar with GS IV integrator.

Essentially the same as the MPP... just some different values in the charge path and output circuits (that change the stop band and pass band). And the GS IV uses much shorter sample periods than the MPP (which lowers the integrator noise floor)

That my friend is much more complicated than calculating the preamp noise out. Agree, did you ever get where you thought you figured it out?

To calculate it or pre-determine the value.... NO and I am really not trying to. But I have gotten to where I understand the attributes/parameters and how they all interact. That allows me to analyze the relative performance of different designs as they apply to my particular requirements.