Original Question: AD633 noise SImetrix model doesn't match datasheet by dhhsieh
Hi everyone, I am trying to do a noise simulation of a circuit using the AD633 low-cost analog multiplier. As a first check on my results, I have wired the multiplier in ADIsimPE (Simetrix) as in the attached file. When I run a noise simulation I get a flat noise spectrum of about 200nV/rtHz. However, the datasheet (also attached) shows a flat spectrum of 800nV/rtHz. I am wondering why this is so, since I have been able to get SImetrix models of op amps such as the AD797 and AD822 to approximately match their datasheet noise spectrums. Either 1) the SPICE model is out of date, 2) I didn't do the simulation right, or 3) the difference in noise spectrums isn't a big deal. Please let me know as soon as possible, thank you in advance for your support.
Verified Answer: RE: AD633 noise SImetrix model doesn't match datasheet by jstaley
Hi Daniel,
You are correct, the noise wasn't included in the AD835 SPICE model. I haven't delved into the subject but multiplier noise is unique in character, and functionally dissimilar to noise in a typical op-amp. I once spoke to Barry Gilbert (designer of our multipliers and many VGA products and inventor of the Gilbert Cell) about the topic, enough to know this is an unstudied opportunity but haven't personally followed up on it.
I don't recall anyone asking about it, and we do provide whatever measurements you see in the data sheet. You noticed that the published noise is higher, and it actually is. If you would like to know why you could try to Google the topic. After all this time someone might have dug into the subject, or you might contact Barry himself.
Multipliers are a precursor to mixers and VGAs and people still use them in many applications, we sell a lot of them. However we also offer VGAs and mixers with technologies not available back when the '600 and '800 series of products were developed so this is where the markets have led.
If it were me I'd probably look for a way to add a noise generator to a SPICE model and just use the data provided in the ds as a source for the model. It's usually pretty accurate.
Sorry I couldn't be of more assistance.
Jim
Suggested Answer: RE: AD633 noise SImetrix model doesn't match datasheet by jstaley
Hello dhhsieh,
Thanks for the EZ question.
Years ago the SPICE modelers did not include noise in their models, the AD633 would have been of that vintage.
Best regards,
JS
Question: RE: AD633 noise SImetrix model doesn't match datasheet by dhhsieh
Hi Jstaley,
That helps a lot. I have two follow up questions.
1) Is the AD835 also an model without spice noise? I took the model from the ADI website at this link.
2) If the AD835 is also too old to be modelled, could you recommend a low-noise analog multiplier which does have spice noise modelled?
Sincerely,
Daniel
Question: RE: AD633 noise SImetrix model doesn't match datasheet by dhhsieh
Hi Jstaley,
Thanks.
I am trying to square an analog signal, and I think an analog multiplier is the right tool for the job, I don't know what a mixer or VGA (variable gain amplifier) do very well.
Are the AD534 or AD734 better for the job, versus the 600s or 800s series as you said in your previous reply?
Is there a SPICE model for the AD534? I only see one for the 734.
Daniel
Question: RE: AD633 noise SImetrix model doesn't match datasheet by jstaley
Hi Daniel,
Yes the analog multipliers really do multiply. Once I had an assignment with the same requirement. I was quite pleased and surprised to learn about these types of devices that provide a result with no latency, as one would experience with a digital solution. Of course there is room in the world for both types of technology.
What is the nature of the signals you need to multiply?
Jim
Question: RE: AD633 noise SImetrix model doesn't match datasheet by dhhsieh
Dear Jim, sorry for the late reply.
I need to multiply noise with itself, which has amplitude in the microvolt range, I do not know precisely. Thermal noise itself is my signal of interest.
Perhaps an analog multiplier like AD734 is not the right tool for the job, because of its inherent noise levels which are likely the same magnitude or larger than the thermal noise, and because the noise would couple to the output. Also, it might not work because multiplying a really small signal by itself would just produce an even smaller signal which might have trouble being measured.
Please let me know if you have any comments or questions.
Daniel
Question: RE: AD633 noise SImetrix model doesn't match datasheet by jstaley
Hi Daniel,
That's a really interesting challenge. Yes a small number squared is an even smaller number but what happens if the number is made larger, then scaled? I don't have the answer but you might want to consider taking an indirect path to arrive at a reasonable solution.
Best regards,
Jim
Question: RE: AD633 noise SImetrix model doesn't match datasheet by dhhsieh
Dear Harry,
Good point. Autocorrelation (a noise or signal squared) is not exactly what I want, but it will do. In the current circuit design I am actually trying to multiply together the noise output from two separate but identical lines of amplification, so that the multiplier output will eliminate the noise of the amplifier chains themselves. This is called cross-correlation and it is done in the literature and I am going to try it, though I don't know how hard it will be.
To clarify, I think two noises or signals are 'independent' or not correlated if the average voltage of their instantaneous product in the time domain is zero.
Here is a paper I have read which talks about cross-correlation and using a multiplier.
White, D. R., Galleano, R., Actis, a, Brixy, H., Groot, M. De, Dubbeldam, J., … Gallop, J. C. (2003). The status of Johnson noise thermometry. Metrologia, 33(4), 325–335. http://doi.org/10.1088/0026-1394/33/4/6
Sincerely,
Daniel