Mr. Krueger possesses a large added entertainment value.
[If you know neither yourself nor your enemy, you will always endanger yourself. If you only know yourself but not your adversary, you may win or lose. If you know your enemies and know yourself, you can win numerous battles without a single loss]
[If you know neither yourself nor your enemy, you will always endanger yourself. If you only know yourself but not your adversary, you may win or lose. If you know your enemies and know yourself, you can win numerous battles without a single loss]
Thank you.
I suspect that the problems are more political than technical.
Notice the liberal doses of gratuitous insults and personal attacks in close sucession.
Could be. However, if you are a physicist, that rather badly damages your uncredentialed status.
Or, not being encumbered with personal entanglements might be the real virtue.
Last edited:
-IN and +IN are tight to ground through 10 ohm resistors. You can see it in the schematics posted. +Vs and -Vs have standard bypass capacitors to ground.
Why there is no problem with AD797 and LT1028? Both are bipolar opamps as well. They were tested in the same circuit with LM49710, which behaved absolutely same as double LM4562. IMO the answer is the chip circuit design.
Why there is no problem with AD797 and LT1028? Both are bipolar opamps as well. They were tested in the same circuit with LM49710, which behaved absolutely same as double LM4562. IMO the answer is the chip circuit design.
It would be most interesting to see the details of the topologies. ADI was quite candid about the 797's departure from the "standard" VFA arrangements. No details from LT about the 1028 and none from National-now-TI about the 4562. Maybe now that the latter is being de-emphasized in favor of some more recent TI parts, they could be persuaded to show a bit more.
Since all three have substantial input stage currents, that in itself doesn't explain the behaviors. Maybe something in the second stage of the 4562?
Likely because the LM parts have an unusual (for a bipolar) very low input bias current (hence much higher input impedance, hence much more sensitivity to magnetic coupled noise, through the test circuit loops, including the 10ohm resistors), coupled with a very high OL gain. Such a test circuit requires a very tight SMD construction with multiple ground planes, triple shielded against electric and magnetic fields. The 79410 and the 4562 have, very likely, the same silicon chip.
LME49710, LM4562 typical Ib=7-10nA, OLG=140dB
AD797, typical Ib=250nA, OLG=140dB (loaded)
LT1028, typical Ib=25nA, OLG=140dB (loaded)
I think you are overreacting, the TI part (the LM4562 is not discontinued) is an amazing opamp for audio use.
Maybe, but second law (yes, it's a bitch) says that a bias current cancellation technique cannot come without a noise penalty. With 1.6pA/rtHz for the TI part, I doubt there is room for that. BTW, the 79410 common mode input impedance is specified at 1GO.
AD797 doesn't have any input current cancellation and has 2pA/rtHz. LT1028 is far behind with 4.7nA/rtHz.
Interesting point.
The AD797 is a folded cascode input stage. I doubt there is any degeneration ( it's got very good noise performsnce). The 4562 internal details are not known AFAIK - also unlikely to have degen as a low noise part as well.
The GM on the JFET input stage is probably much lower than the BIP devices - we know that confers better RF/EMI resistance.
But, it's not clear to me whether it's EMI related - as I noted before, it's easy to test by simply screening the test jig.
The AD797 is a folded cascode input stage. I doubt there is any degeneration ( it's got very good noise performsnce). The 4562 internal details are not known AFAIK - also unlikely to have degen as a low noise part as well.
The GM on the JFET input stage is probably much lower than the BIP devices - we know that confers better RF/EMI resistance.
But, it's not clear to me whether it's EMI related - as I noted before, it's easy to test by simply screening the test jig.
Re the LM4562 and related parts, I think only the DIP version is discontinued. A lot of the THP leaded packages are on their way out.
Better get used to using magnifying glasses and tweezers . . .
I am still struggling to understand how waving your hand over the circuit can cause an issue if the inputs are loaded with 10 Ohms. I agree construction would have to be good to avoid noise pickup.
Better get used to using magnifying glasses and tweezers . . .
I am still struggling to understand how waving your hand over the circuit can cause an issue if the inputs are loaded with 10 Ohms. I agree construction would have to be good to avoid noise pickup.
I think there is plenty of room. My point is, you can't use the spec on bias current to infer a lower input stage current and thus a greater propensity for rectification effects. I don't even know if they mention anywhere that they are doing bias current cancellation, but if they weren't you wouldn't manage 2.7nV/sq rt Hz e sub n with everyday betas, given the (uncancelled) input currents. Note btw that the typical spec on bias current is slightly less that the typical spec on offset current.
We don't know precisely what the collector currents or the betas are, but with some plausible assumptions including the usual early 1/f cornering of base current, the need to have high enough collector current for a low "half-thermal" of r sub e, a traditional differential pair arrangement, and a plausible rbb', it all works out. Note that the part's quiescent current is about 10mA. So there is enough for a few mA of input device collector currents before the cancellation schemes.
I think the problem of higher rectification efficiency may well be in the apparatus of the bias current cancellation details themselves, with my second choice the details of the second stage. Remember that the bandwidth of the AD797 input stage is very high, and the subsequent stages like the initial folded cascode preserve a good deal of it. I don't know what LT does.
But none of this means it's a bad part. I designed the duals (the 49720) into a class D power amp as the front end (not part of the power amp feedback loop) and got exemplary performance, although the contact with the outside world was through a load-compensated JFET follower. I don't recall doing immunity tests, as we didn't get that close to production before four of us departed, but it was hysteretic noise-shaped class D, on a small board with two independent (not slaved-clock) channels, and there were only heterodyne artifacts under highly-contrived conditions, like open inputs.
Sorry Waly that is not true. 2pA/rt-Hz is the noise of 12uA about right for a beta of 150 and 900uA per side bias. Simplified schematics are just that.
No degeneration used, but the later multi-tanh input amps get some real benefit at audio frequencies. No PIM in an absolute sense
the input transfer function over 10's of mV is exactly linear.Sorry Waly that is not true. 2pA/rt-Hz is the noise of 12uA about right for a beta of 150 and 900uA per side bias. Simplified schematics are just that.
No degeneration used, but the later multi-tanh input amps get some real benefit at audio frequencies. No PIM in an absolute sense
Mr. Wurcer, what is not true in what I said? I don't disagree with anything you say, my only point was that I doubt the TI part has input current cancellation. Slightly less input current noise compared to the AD797, but 25 times less input bias current, where would the alleged input current cancellation noise contribution fit in?
I'd rather think the TI part is built in a modern process, perhaps with a much higher beta devices available?
- Status
- Not open for further replies.