andy_c said:
Hi Andy,
I agree completely. He made us think.
Great job on your link. To me, your math is breath-taking. I don't think I would have the patience for it, much less the skill.
There are two other things about PIM that I think might be true, but which I have not explored theoretically or with measurements.
1. Can PIM exist in the absence of conventional distortions like SMPTE IM and THD?
I suspect not, except in a pathological case such as with a nonlinear all-pass filter somewhere in the signal path. Even then, I think PIM would show up as sidebands on a spectrum analyzer.
2. Does an amplifier with deliberately restricted LF open-loop gain create more PIM?
Using the input-referred feedback analysis, the low-frequency error signal driving the LTP will be larger in an amplifier that has less open-loop gain at low frequencies. This would seem to stress the LTP harder and cause it to compress more, creating more PIM. This observation, if true, may only apply to PIM as measured by Otala's technique, wherein the agressor frequency is a high-amplitude 60 Hz sinewave.
Cheers,
Bob
john curl said:
Thanks John.
john curl said:
The datasheet for the AD825 is here (PDF) and the AD859 datasheet is here (PDF). Looks to be around 10 kHz for each.
Bob Cordell said:
I suspect not as well, but I don't know how to prove it one way or another.
Bob Cordell said:
Hmm, that's an interesting question. I suspect the total IM would be higher, but as to how this breaks down into AIM and PIM components, I don't know. The input-referred distortion thing gets quite confusing in the multi-tone case.
As far as single-tone AM-to-PM, if the open-loop gain has no AM-to-PM and its input-output phase shift is identically zero, there can be no AM-to-PM in the closed-loop amplifier. Trouble is, you can't get identically zero phase shift, and reducing open-loop phase shift requires reducing the open-loop gain quite substantially. So on the one hand, you don't get the PIM hit from having Vdm be in quadrature with Vfb, but on the other, Vdm will be much larger than it would have been in the integrator (quadrature phase) case. So where do you end up? Back where you started (assuming equal GBW)? Dunno.
Btw if you want to compare the Lohstroh Ottala amp with the subsequent Electrocompaniet commercial offering you'll find the latter circuit on a Greek site: try "Sam Electronic Circuits" in Google. That seems to work. It is in the power amp section.
There used to be a good site detailing the evolution of the original (LO) design in the early years but only part of that is now up and the "re-construction" work to be announced has not yet turned up after sevaral years. It would be good if that was finished as it contained a lot of interesting material. The main changes seemed to be around the front end, to do with noise from memory, but the overall design remains the same.
There used to be a good site detailing the evolution of the original (LO) design in the early years but only part of that is now up and the "re-construction" work to be announced has not yet turned up after sevaral years. It would be good if that was finished as it contained a lot of interesting material. The main changes seemed to be around the front end, to do with noise from memory, but the overall design remains the same.
stinius said:
Hello Stinius
You seem to know a little bit about Electrocompaniet, can you tell us if the philisophy of high open loop bandwidth was carried over into their preamplifiers .
Regards
Arthur
andy_c said:
Thanks Andy
But, they are different AD 825 70dB Open Loop (Only SMD
)AD 829 100dB Open Loop.
A designer of my country did AD 825, reported some Offset in output.
Already tried to use Opamps without negative feedback?
andy_c said:
What if the output impedance of the VAS was raised? The pole-splitting effect would be diminished and the ULGF subject to greater modulation due to non-linear capacitive loading at the VAS collector (driver input capacitance)??
Rafael.luc said:
But, they are different AD 825 70dB Open Loop (Only SMD
Scott gave us some DIP samples during Burning Amp festival in San Francisco. This year I'll bring my class A+C+C amp made with this chips. If somebody wants to measure, welcome! 🙂
G.Kleinschmidt said:
You've answered by big eyes' smiles when I wrote about that.
G.Kleinschmidt said:
Hello Glen
Thats what I think is the problem in cascoding the VAS in blameless amp, the non-linear capacitance effects can increase the distortion on the voltage swing at this stage, is this what you are implying. Having lower impedence at this node all things being equal should have lower distortion.
Regards
AR
These chips must be compromised, because they HAVE a high open loop bandwidth. Everybody knows that!😕
Analog Devices, will sell at DIP?Wavebourn said:
There is also the AD817 ..............~70dB/ 10Khz open-loop.
http://www.analog.com/static/imported-files/data_sheets/AD825.pdf
G.Kleinschmidt said:
Well, what you're talking about is an actual real-world type of situation. What I've been talking about is a vastly oversimplified situation which nonetheless has math describing it that's so messy there's no more real insight to be gained from it than one could obtain by just doing a simulation 🙂. That's one reason why I haven't pursued this beyond what I've already done.
...which got me to thinking... I wonder if there's a way to make a "virtual" version of Bob's PIM analyzer in simulation. Hmmm...
I suspect that the AD825 might actually sound better than the AD797, but I use both, and the 797 will measure better. Just my imagination, I guess.
andy_c said:
Well, OK, but with respect to the assertions made about the PIM implications high LF loop gain and NFB, the VAS-output stage interface just came to my mind.
In an ideal amplifier with an integrator VAS you have a 90-degree phase margin. In a real world amp you may have, say, a 70-degree phase margin. This means that the unity loop gain frequency is determined by the dominant pole capacitor and, less significantly, by some other pole(s).
If one eschews NFB and high LF loop gain, I can think of a few reasons why one may have a hard time splitting into oblivion that pole formed by the finite-minimum output impedance of the VAS with the (non-linear, and substantial) input capacitance of the driver stage.
Without running a sim or pressing any buttons on a calculator, that just appears to me as an obvious (and needless) source of PIM (and other distortions as well, of course).
G.Kleinschmidt said:
If you use the usual approach of a resistor to ground from the VAS output to get a wide OL BW while keeping GBW constant*, does this affect the pole-splitting of the VAS? It's been ages since I've looked at the pole-splitting formulas, so I can't remember.
*Not that I recommend this...
Andy, I did try the technique of using a resistor to open the open loop bandwidth to approximately 20KHz back in 1973, with a 1 meg resistor on pin 8 of an HA911 IC op amp. It worked amazingly well, did not effect the stability in any significant way, but LOWERED the SMPTE IM distortion a great deal, AND improved the slew rate somewhat. Figure that one out.