Noise Analysis

[JC1]

EE

Hi Eric,

Actually the Voltage noise of the op amp is typical, if you are doing a specific application you may be able to substitute opamps from different date codes or brands and maybe find one a bit less noisey.

Make that 10 to 15 mV. Maybe get less than 10mV at a gain of 1000 {with or without the second amp}, but changing resistors around and stuff, don't get much improvement.

Like I said before lower feedback and input and get to 3.9uV, so between there and you should be able to realize 7.1uV for sure. but that's about it.



If you have been doing this on the bench, I feel your pain.

[JC1]

EE

Hi Eric,

If you can drop the input resistor to 500 ohms, and then you can use 5k, 50k, or 500k feedback and gain up to 1000 in each case (ignoring any noise from gain up amp) and you get 5.5 uV/rtHz in each case.

5.5uV * 316 * 6 = about 10 mV p-p.

So if you can double up on the clamp diodes or something and get a 500 ohm or lower in there, you might see some difference.

[Ferric Toes]

So Eric you should see about 15 to 20 mV peak to peak noise on the scope on your circuit. If alot higher then you got some noise that is not part of these parts.

On my dual preamp (5534/8055) with a total gain of 500, and 1k and 10k resistors for the 5534, I get 20mV pk-pk noise at the output of the AD8055. This is with 1k grounded, rather than going to the coil circuit. TX, coil and everything else is connected and running, so this result doesn't look too bad. I have not noticed, so far, any differences in noise level with different 5534's.

Eric.

[JC1]

EE

Hi Eric,

I never tried too many 5534A s so may not make too much difference. of course the A s are lower in noise and what I have been talking too. Grounding the input resistor is right. I guess you meant no noticeable change with the coil etc. hooked up which is probably right.

Measuring the noise with an analog scope can be a function of intensity of the beam there is an article and a good way using two probes on the same output to measure noise.
I will see if I can find again on the net.

Of course your noise bandwidth may not be exactly 100 khz . and if you believe alot of this is 1/f noise low freq noise then a high pass filter somewhere might help or otherwise reduce bandwidth. these simple calculations don't really take into account the spectral components of noise and assume flat across spectrum which isn't true.

Why I said you will probably see more noise.
plus theory is minium.

Well there's my 2 cents for what it is worth.

Industrial application? plugged in wall? maybe thermoelectric cooler to cool the opamp?

Grasping at straws at this point.

[JC1]

EE

Hi Eric,

Maybe I will turn something on around here and see what i get on a 5534. With a gain of 500 instead of 1000 I would still expect to get more like 10 mVp-p but thats from the data and could be worse. Not familiar with ad8055 but with only gain of 10 in first stage its noise is probably in there a bit.

[JC1]

EE

Hi Eric,

Found that pretty article pretty quick. here it is.

http://www.reed-electronics.com/tmwo.../CA187331.html

This technique uses two scope channels to measure the random (Gaussian) noise at any point in a circuit. You can use one channel and roughly measure the peak-to-peak noise in your circuit, but you will get a more accurate measurement with two channels.

[JC1]

EE

Eric,

3. Gruchalla, Michael E., “Measure Wide-Band White Noise Using A Standard Oscilloscope,” EDN, June 5, 1980. p.157.

4. Franklin, Gary, and Troy Hatley, “Don’t Eyeball Noise,” Electronic Design , November 22, 1973. p. 184.


Actually it was one of these articles, I read the one I posted above. If I remember right there is a correction factor that has to be applied to the voltage seperation and gives you RMS. not sure the answer is just the seperation of the traces as stated in this article, but don't remember.

[Ferric Toes]

Measuring the noise with an analog scope can be a function of intensity of the beam there is an article and a good way using two probes on the same output to measure noise.
I will see if I can find again on the net.

I will have a look at the two probe method. As well as my standard scope, I have a Tektronix storage scope, and I set that up for a fairly long persistance, so that the noise would add up over a period of time and give a thick band on the screen. Still about 20mV.

Yes, I always use 5534A's.

The AD8055 is quite noisey at low frequencies, but this does not seem to be noticeable with the x10 gain before it. At one point I substituted the 8055 for another 5534, and got a similar result. Also used a LM6171 low power, high speed opamp, for the x50 stage, and that worked OK too.

Eric.

[JC1]

EE

Hi Eric,

Well using a 1k grounded and a 1 Meg I get about 17 mVpp of noise "eyeball". 5534A in breadboard.

so with gain of 500, noise = 9 mVpp.

You said 20mV is with detector running, but 1k ground, so 20 mVpp may be more real, you may also be able to get rid of 10 mVpp with good ground, or may be there now and scope and ground showing more than circuit is really seeing, with rest of circuit banging the ground so to speak.

[Carl-NC]

Page 4... opamp noise

[bbsailor]

Carl,

While looking at the published specification for an op amp only shows it's typical characteristic, is there much actual noise variation in any one batch of TI NE5534AP op amps?

If I ordered a quantity of 25 TI NE5534A op amps, might there be a few that have lower noise than others?

Is the variation too small to be seen in actual performance?

I am enjoying this noise thread.

bbsailor

[Carl-NC]

The TI datasheet on their version of the NE5534 has vn=3.5nV/rtHz typical, 4.5nV/rtHz max. This is at 1kHz. They also list 30Hz numbers: 5.5nV/rtHz typical, 7nV/rtHz max. This is a pretty normal max for vn (40-50% above typical), because vn depends on internal biasing, as well as transistor resistances, and the gain of the first stage.

Datasheet numbers for "in" are typical only; usually, it is tighter than vn as it depends mostly on the bias current of the input stage, which might vary 10-20%. And, as my example shows, "in" doesn't even matter for a simple gain stage.

- Carl

[Carl-NC]

More noise...