Aerospace runs on requirements. A new airframe takes a few hundred man years effort before it even flies for a big commercial jet. Requirements analysis and systems engineering are fundamental up front. There is no room for 'run it hard for 10 minutes then it will fly straight'. Everything has to be captured as a testable requirement. Testable is the key here. Nothing you talk about is testable. And let's not get into repeatable.
More personal attacks. And lies, too. You can't keep it professional, can you?
I will report this post to the moderators.
Jan
Why things are not easily testable in the audio game is because no-one has come up with a reliable, all-encompassing measuring regime that captures all the performance parameters that count. The plane industry has been going for a long time, in the early days they "measured" planes by how times they crashed, and so over many decades they steadily worked out, evolved a vast array of methods of determining how well the plane would fly, before it ever left the ground ... guess at what point the audio industry is, where does its stage of evolution sit, comparatively speaking, in the timeline of that other engineering endeavour ... ?
If it was paper phenolic, not sure why you'd call it squirrelly. It's dielectric properties are perfectly fine for audio work.
se
The most powerful tool I've found is 'difficult' recordings - the stress test technique. Using "pure" signals doesn't unearth the really important distortion mechanisms, the ones that make the listener lose interest, get bored or irritated with the sound, the very things that are most important to long term pleasure with listening ...
It was something I'd never heard of, I think cardboard was mentioned in part. However the implication was it was somehow crucial to the supposedly great performance, and by inference, that everyday FR-4 would be unacceptable. I remained skeptical, however...
There was a problem with early glass-epoxy laminates discovered by Tektronix, with some very strange dielectric behavior. They had needed some small capacitances in the vicinity of certain high-frequency ICs that were part of a vertical amplifier, and they made them with just copper pads over the ground plane underneath. They traced it to the board material, and dubbed the effect "hook" because of the particular waveform distortions (at relatively low frequencies, not many MHz). They traced the material that elicited this to the vendor's particular process, and apparently it is never a problem today---although one can only imagine what comes out of some factories now.
The article about this was in one of the magazines, possibly the now-defunct Electronics, probably circa 1975.
main effect of higher Z FB network
It appears the effect is mostly just a too-low open-loop inverting input impedance, mostly the resistive component, loading the higher-Z feedback networks and slightly reducing loop gain. I see in sim a net input R at 20kHz of 17.7k, which is plausible based on devices with betas of 600 or so. So this loads the higher Z feedback divider somewhat, and the lower Z one hardly at all.
Cascoding doesn't help this.
It appears the effect is mostly just a too-low open-loop inverting input impedance, mostly the resistive component, loading the higher-Z feedback networks and slightly reducing loop gain. I see in sim a net input R at 20kHz of 17.7k, which is plausible based on devices with betas of 600 or so. So this loads the higher Z feedback divider somewhat, and the lower Z one hardly at all.
Cascoding doesn't help this.
yes, a change of, let say 10% of the overall distortion at 20KHz ? Can even go the other sens at low frequencies ;-)
Well, i took the worse example for me.
In a CFA i believe the effect would be more important, but keeping everything equal around was more complicated and i wanted the thing to be clear.
Anyway, this has no major importance, just something to keep in mind, that i tried to enlighten before all this *strange* controversy started.
I believe low impedance of feedback is a plus, anyway, when we can (rfi, emi immunity etc.). It is a very sensitive path in an amp and every evil that can be induced here will be in the output. Even the point where you take the source of the feedback can change the distortion figure. I use to take-it at the exact point where the protective coil is connected (if any), and try to set the paths of the two sides of the power devices perfectly symmetrical relative thereto.
Not sure to understand your point (my poor English)
In the example i showed, with 25K and 1175, the feedback impedance seen by the input transistors is 1122K. A big change in the level of the feedback if the impedance input R is 17.7k ? Enough to justify such changes in the response curve at HF ?
I really believe it is the result of phase rotation of the feedback signal around 1MHz due to the bases parasitic caps (or emitter in case of a CFA).
Last edited:
Yes , just overbuilt 120db PSSR analog supplies , low noise separate regulators
for your digital ... plus generous RF line conditioning will make for the best SQ.
Without magical shampoo's and conditioners !
OS
I know the phenolic board by the name SRBP = speciall resin bonded paper - the preserve of high volume TH consumer electronics. I do see a move to FR4 type material though in that industry and it may be related to the use of very small SMD and double sided in some cases to meet space constraints.
Last edited:
I guess you mean 1.22k.
I backed out the effective input resistance of the amplifier and determined that it alone is 17.7k for my particular model. This is NOT of course the feedback divider impedance, but a good deal higher. And yes with voltage swing, that 17.7k is also not constant, and yes there is a capacitive component. But I suspect most of the distortion is from the subsequent stages, and more loop gain will help. One could show this by conjuring a "perfect" input stage with ideal transconductances, and use this to drive the subsequent stages.
My point is that in terms of distortion, even a resistive input impedance will still represent a lower overall amplifier loop gain when the divider impedance is high. The change noted for the large difference in FB Z is still a small one at 20kHz.
(Did you change the feedback capacitor to make the closed-loop rolloff for the higher impedance FB the same as the low? I guess you would have had to, or the 20kHz closed-loop response would have been way down).
I know. That was not the AN material. I only mention the hook stuff because it was a surprising degradation mechanism, for a material that was unsuspected to be the culprit.
Last edited:
- Status
- Not open for further replies.