It was beta, again in a complicated heterojunction experimental device. I didn't ask about breakdown. I did ask to explain how it was achieved, hoping for a Feynman moment, but he said Read Sze. He then lectured me on the failings of op amps... I then recited a few of the things I did in research instrumentation at UCLA, and this ended the exchange.
Well, he's (or he was) a very busy man. This was several years ago.
I hunted down an old-dog programmable op amp, the LM4250, that has externally set input stage and other-stage currents. But it is very limited in that range and has 10k emitter resistors. It looks as if it would be truly horrible for audio use, but might be useful in researching the rectification issue.
I am tending back to my hypothesis that with the shunt network Pavel has, a very large portion of the feedback is shunted to ground to the point that there is not enough signal feedback current, so the device in effect operates OL, or partially OL for a portion of the input signal.
The JFET opamps don't have this problem since the required signal feedback currents are orders of magnitude below the bipolar opamps.
However, this does not explain the AD797 and the fact that it does not appear display this phenomena. It might be that the Ib cancellation scheme may have something to do with it.
The JFET opamps don't have this problem since the required signal feedback currents are orders of magnitude below the bipolar opamps.
However, this does not explain the AD797 and the fact that it does not appear display this phenomena. It might be that the Ib cancellation scheme may have something to do with it.
I place my bets on the test jig implementation. Again, bipolar opamps with very low input bias currents are much prone to collect, perhaps rectify, and certainly amplify mains harmonics through magnetic induction in the test jig circuit loop(s).
Of course, the wealth of mains harmonics in PMAs spectra could be due to rectification in the input circuit (and would explain why jfet opamps don't look so bad) but having the input bias current as a variable, the current test jig implementation precludes comparing bipolar opamps.
I would consider a 4 layer PCB with 2 ground planes, with SMD parts and triple shielded against electric and magnetic fields a minimum, and a test protocol per the TI document, to be able to e.g. compare the LM4562 design to the LT1028 design "EMI sensitivity".
So far, the only conclusions that I see as valid are:
1. Jfet opamps are less sensitive to "EMI rectification" (we already knew that, isn't it?)
2. Bipolar opamps with low Ibias (call them "precision") are more "EMI sensitive" (makes sense, but is this a good enough reason to avoid them?)
I don't see so far any valid proof that (e.g.) the LM4562 is intrinsically (that is, by design) more prone to "EMI rectification" or more "EMI sensitive" than the AD797 or the LT1028.
P.S. To add insult to injury, "EMI sensitivity" and "EMI rectification" above are in fact, so far, ill defined concepts. A precise metric is required at least for the "EMI sensitivity" for an apple to apple comparison. Again, the TI document quoted above addresses this.
Last edited:
It's important I think to look at the overall architecture. These op amps are complex systems.
If one had an op amp with sufficient bandwidth and if the parasitics, especially the inductances, could be small enough, RF would simply be amplified with good linearity. I recall an experimental amp, I think it was done by some people at Hughes, that had a 10GHz crossover frequency and was close to unity-gain stable. The article remarked that they were pleasantly surprised that the modeling worked so well, and although many of the conventional parameters differed a lot from more conventional ones of the day (this quite a few years ago), the prototype agreed well with the model.
The part had only a modest open-loop gain. Scott remarked that some internal transistors today sport ~100GHz ft's, and are necessarily integrated since even the tiniest SM packages would assure excessive parasitics if applied to low-pass broadband-gain applications like discrete op amps.
I don't know what's inside the LT and TI parts exactly, but the 797 departs significantly from the traditional diff pair/current mirror followed by the somewhat-misnamed VAS and output buffer. The folded cascodes hold the input collectors at about a constant voltage, making the common-mode Miller multiplier of Ccb just about unity and simplifying the phase-gain behavior. Thereafter iirc there is no additional common-emitter amplification. So the parasitic pole accumulations common to other topologies are minimized.
Input bias current sees 10 Ohm to ground. Shouldn't be much of a problem
That's on the schematic, not in the test jig implementation, the HF impedance could be significantly larger. Not to mention there's more than one possible loop where the EMI could be randomly be injected.
As a rule, if waving your hand over the circuit (as already mentioned by Mr. PMA) changes the output, then you have a test jig implementation problem.
What I'm still trying (apparently unsuccessful) to convey is that the differences recorded between the LM4562 and the AD757, LT1078, etc... are not necessary because of internal design differences, but because of the lower input bias current of the TI part.
I agree that it does not change the experiment significantly. 80 dB gain, 86 dB gain, whatever.
It was a minor detail, and should be handled as such.
I only mentioned it this time because someone made a post that said that my inability to properly calculate the gain proved that I was technically incompetent.
Fact is, the gain was slightly different than initially represented, and I was basically disrespected and called an idiot for catching the slight detail difference.
I'll live but I will consider that source for what he has proven himself to be. ;-)
This proves that you are unfamiliar with the well known fact that a voltage mode signal source is properly considered a short circuit to ground all frequencies.
Engineering 221. Not exactly a rookie mistake but your inability to recognize and react properly to a clear presentation of correct technology is noted.
This is pretty much possible. On the other hand, it shows that the TI part might be, in the same setup, much more prone to parasitic impedances and to shielding than the competitive parts with BJT input.
LM4562 is also produced in the TO-99 package. Will it behave the same?
> Will it behave the same?
~ 1970 (?) We switched from ua709 (metal can)
to lf356 (dip) based on listening ..... After ~ 6 months
customers complained the 709 units were better.
The 356's (we then found) had degraded. The Signetics
guru said they were finding (with the new plastics)
that some sort of contamination was getting into the chips
............................ FWIW
~ 1970 (?) We switched from ua709 (metal can)
to lf356 (dip) based on listening ..... After ~ 6 months
customers complained the 709 units were better.
The 356's (we then found) had degraded. The Signetics
guru said they were finding (with the new plastics)
that some sort of contamination was getting into the chips
............................ FWIW
The new plastics don't have this problem. There were a lot of issues on power mosfets in the 90's related to plastic contamination (auto applications, so very stressful) - all solved about 15-20 years ago. If its really important, they also put a layer of nitride over the chip - but I don't know if this is done in the amplifier business.
Yes, all old problems long solved we have a 1fA op-amp electrometer in plastic. Except possibly clear plastic packages there are some with serious limitations and precautions.
To me, your first sentence is completely nonsensical, whereas the post this triggered was very much about explaining what we are looking at. If you don't agree with my explanation, why not post technical arguments?
Anyways, Pavel, can you chip in? Or Scott? I just want to know if it is 80 or 86 dB noise gain, because if it is not 80, I have to improve my understanding. That is called learning, and I am not affraid of it. Unfortunately, I have not yet mastered the art of gaining understanding by means of insult.
- Status
- Not open for further replies.