PI noise

[green]

Wanting to lower noise. Integrator input connected to common(low noise). Bypass integrator(low noise). What is causing the noise with circuit connected normal and what could I try to lower the noise?

[ODM]

I'd imagine you are not seeing the very low noise level in the input before your chained gain stages make it stand out. This is the reason for the "appearing" noise when all gain stages are chained. Your oscilloscope has a noise floor that will hide the "quiet" signal. Use the shortest possible ground wire for your scope probe in measurements if you want the lowest introduced ground sling noise. Linear's application note AN 47 is very good for tips on building and benchtesting analog measurement circuits.

Also, all amplifier stages have noise in themselves, the amplifiers' internal noise and feedback resistors' noise. The "common" approach is to lower the preamplifier stage noise, as all following gain stages amplify the noise of previous stages.

[green]

I'd imagine you are not seeing the very low noise level in the input before your chained gain stages make it stand out. This is the reason for the "appearing" noise when all gain stages are chained. Your oscilloscope has a noise floor that will hide the "quiet" signal. Use the shortest possible ground wire for your scope probe in measurements if you want the lowest introduced ground sling noise. Linear's application note AN 47 is very good for tips on building and benchtesting analog measurement circuits.

Also, all amplifier stages have noise in themselves, the amplifiers' internal noise and feedback resistors' noise. The "common" approach is to lower the preamplifier stage noise, as all following gain stages amplify the noise of previous stages.

I'm trying to understand why the low frequency noise(less than 15 Hz) is 4 to 5 times higher with the integrator(gain = 1) in the chain. I'm guessing it's aliasing, generating false low frequency signals. I'm wondering if sample rate and sample time effect the false signal amplitude. Are there formulas to calculate the false signals? Maybe I'm looking at it wrong and someone could correct me.

[Mechanic]

Hi Green,

Ground the output from your preamp and see if the noise has a significant change...

Also try reducing the input resistance to the preamp down to 10 or 20 ohms and keep the gain the same. You might want to try an ad797 for the preamp once you have reduced the input resistance.

Cheers Mick

[green]

Hi Green,

Ground the output from your preamp and see if the noise has a significant change...

Also try reducing the input resistance to the preamp down to 10 or 20 ohms and keep the gain the same. You might want to try an ad797 for the preamp once you have reduced the input resistance.

Cheers Mick

Hi Mick
Thanks for the reply. I've been playing with different integrators and other circuits lately. The results are similar to the attachment in post #1. With the integrator in the chain and integrator input grounded the signal at post amplifier out is low. With the differential amplifier input shorted and the integrator bypassed the signal at post amplifier out is higher but still low. The last try I did (sample rate=1kHz, only one sample time(no EF sample) 10usec, Rfdbk/Rin=100. integrator gain=1) with the integrator in the chain the noise signal at post amplifier out was 4 to 5 times higher than it was with the integrator bypassed. I had thought the low frequency noise might be 1/f noise but sampling the high frequency noise seems to cause more low frequency noise. The difference amplifier has some noise but so far it's the best I have been able to do.

[green]

Hi Mick
Thanks for the reply. I've been playing with different integrators and other circuits lately. The results are similar to the attachment in post #1. With the integrator in the chain and integrator input grounded the signal at post amplifier out is low. With the differential amplifier input shorted and the integrator bypassed the signal at post amplifier out is higher but still low. The last try I did (sample rate=1kHz, only one sample time(no EF sample) 10usec, Rfdbk/Rin=100. integrator gain=1) with the integrator in the chain the noise signal at post amplifier out was 4 to 5 times higher than it was with the integrator bypassed. I had thought the low frequency noise might be 1/f noise but sampling the high frequency noise seems to cause more low frequency noise. The difference amplifier has some noise but so far it's the best I have been able to do.

Some scope traces. Square wave, TLC555 output, 463 kHz(10 ohm resistor in series with a 50k resistor, 10 ohm resistor across preamp input, 9 volt battery reads 8 volts). Post amp gain 100 between 2 and 10 Hz includes integrator roll off (LT spice).

[green]

Some scope traces. Square wave, TLC555 output, 463 kHz(10 ohm resistor in series with a 50k resistor, 10 ohm resistor across preamp input, 9 volt battery reads 8 volts). Post amp gain 100 between 2 and 10 Hz includes integrator roll off (LT spice).

Wanted to see if lowering amplifier cut off frequency changed the noise. The diff amp is potted so I made a simple two stage amp with a LM4562. Recorded with the input shorted, 1k across the input and the TLC555 output across the input with two different cut off frequency's. Thought I would see a noise change but don't see much if any. Should I see a difference? PI_ noise_3(LM4562), PI_ noise_2(diff amp)reply#6.____ Sample rate 1kHz, target sample time 10usec, no EF sample, Coil command off.

[Mechanic]

Hi Green,

Ok, try this, reduce your preamp gain from 300 down to say 30 and see if the circuit output noise also decreases, this will prove/disprove that the noise is coming from your preamp stage, which I am sure that it is

Cheers Mick

[green]

Hi Green,

Ok, try this, reduce your preamp gain from 300 down to say 30 and see if the circuit output noise also decreases, this will prove/disprove that the noise is coming from your preamp stage, which I am sure that it is

Cheers Mick

Hi Mick, Thanks for the reply. Some more scope pictures. Sampling seems to be the problem. I'm thinking sampling with an A to D would be the same. Maybe I'm missing something or doing something wrong. The AD797 specs look real good. Need to try a SFT1440 isolation switch suggested by mickstv to get the input resistance low. Every thing I've tried hasn't worked. The reason I'm using a diff amp with an IB coil. I still think aliasing is causing the problem but hope it's not and there is a way to lower the noise I get from sampling.

[green]

Hi Mick, Thanks for the reply. Some more scope pictures. Sampling seems to be the problem. I'm thinking sampling with an A to D would be the same. Maybe I'm missing something or doing something wrong. The AD797 specs look real good. Need to try a SFT1440 isolation switch suggested by mickstv to get the input resistance low. Every thing I've tried hasn't worked. The reason I'm using a diff amp with an IB coil. I still think aliasing is causing the problem but hope it's not and there is a way to lower the noise I get from sampling.

The second scope picture PI_noise_4, integrator bypassed doesn't look right. I connected amplifier out to the integrator 3k input resistor causing it to saturate the direct coupled portion, no 1/f signal. Including a trace connected right.

[moodz]

..would you be able to post a picture of the physical circuit ( as built ) ? Thanks.

[green]

..would you be able to post a picture of the physical circuit ( as built ) ? Thanks.

The circuit was put together as a learning platform. Didn't have SMD adapters for the control circuit and coil driver. They have made it a lot easier to wire the circuit.

[Ferric Toes]

The circuit was put together as a learning platform. Didn't have SMD adapters for the control circuit and coil driver. They have made it a lot easier to wire the circuit.

From looking at the picture, your setup is much too open and you are bound to pick up noise from external sources. Here in UK, it is surprising how much radio noise there is a low frequencies, even a few broadcast stations still exist between 200kHz and 170kHz. And then there is the radio clock control signals at 100kHz. All of this can create havoc with noise measurements, plus there is a huge amount of impulse noise.

At the very least put the preamp in a screened box with coax for the input from coil driver and/or coil. Also use a small figure 8 coil, again shielded if possible in a plastic project box that has been sprayed with an emc graphite coating. Any long leads from power supplies should have rf filters where they enter screened boxes housing electronics. All long leads act as antennas.

I have found that computers need to be off and also the power supply for my temperature controlled soldering iron. Wall plug switch mode psu's also often radiate noise.

Without taking these precautions you will be forever chasing your tail.

Eric.

[moodz]

...hmm yes ... I would have to agree with Eric. I would recommend a good ground plane in addition to any shielding .. there practically no other option for low noise amplifier work.

[green]

Thanks for the replies. I have a magnifying lamp with a cfl bulb and a cheap lamp dimmer to control power for the soldering iron. Both need to be turned off when a coil is connected to the preamp. See no difference in the signal at post amplifier out (E) if they are on or off if the preamp input is shorted. I was assuming I was looking at amplifier noise with the input shorted. Maybe not.