 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
24.27.41.140
Ralph, just wondering if you ever happened to locate the Crowhurst article about this and might have bookmarked it? I've searched at length, but no luck. I have become more interested in this issue as a result of implementing feedback to the grid in my latest project. I'm still tweaking the design, but preliminary results indicate distortion is exceedingly low and sound quality very high. :) I realize Crowhurst might not have proposed a direct solution for cathode-node distortion, but I would like to read whatever he wrote as to how this type of distortion behaves.
Follow Ups:
https://www.worldradiohistory.com/BOOKSHELF-ARH/Technology/Technology-Radio/Audio-Handbook-No-2-Feedback-Crowhurst-1952.pdf
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
In recent times I am more interested in SS.
This is the best design handbook for SS (mainly Lin topology) that I have ever read.
It is freely available.
Plenty of information on all topics that are usually ill treated in forums...
Omnes feriunt, ultima necat.
Self wrote an excellent tome!
He does not mention the issues with the input differential pair other than in passing in that chapter (chapter 3).
The input pair isn't linear so you have the same problem Crowhurst identified 40 years earlier- that of the the circuit distorting the feedback signal.
This is supposed to be linearized by the coupling of the differential amplifier. However in practice the coupling is usually degenerated slightly (resistors in the emitter circuit; not shown in the schematic) to reduce 'VAS' distortion; we're talking about that capacitor between the collector and base of the VAS stage. Bruno Putzeys outlines this issue in his paper 'The F Word'.
Self mentions the use of a CCS circuit to help with linearity. To do this properly the CCS must use at least two devices else the CCS circuit leaves a fair bit of performance on the table (the latter as seen in the schematic). However good CCS performance is, the degeneration in the differential stage (not shown in the schematic) reduces the efficacy of the CCS so the result is the feedback signal will intermodulate more easily due to the non-linearity present, which will result in bifurcation of the harmonic (IOW generate higher orders).
My proposal is to treat the amplifier circuit as an opamp and so instead of feedback to the opposite side of the differential input pair it should be sent to the same side as the input signal and mixed via a divider network. This means the phase of the feedback signal must be inverted with respect to the feedback in the form as seen in the schematic. So that probably cannot be done in that circuit without some phase inversion means.
...
Omnes feriunt, ultima necat.
The great advantage of the input pair is that their transconductance does not depend on transistor gain but it depends only on temperature.
The classic book equation of input pair differential transconductance is in fact an exponential curve and whatever tract of an exponential curve you consider it is never linear.
By the way the book guides the expert reader to design 'blameless amplifiers', that is, amplifiers that reduce to the a practical minimum the eight sources of distortion of the Lin topology (the eighth is capacitance distortion).
I have replicated in sim world all stages and the sample complete amplifier and in doing so I learned something new that could possibly be useful for tube DIY.
A similar book by experts of tube design would be even more appreciated...
Omnes feriunt, ultima necat.
-
Glass Audio," Vol 7-6, pp. 20 through 30
is here by M G Scroggie (aka 'Cathode Ray')
on the subject here.
I think this article can be summarized as follows:
1. Tubes are inherently more linear than transistors.
2. Triodes without feedback yield the most natural sounding reproduction.
3. The 45 is an excellent tube.
:)
Only a bit missing.
Distortion of SS starts decreasing from idle to some watts and then starts increasing mildly from a minimum until a sudden increase appears at maximum power.
Distortion of tubes on the contrary is monotonic, it vanishes at low level and the goes always up until maximum power is reached.
This is true no matter how much negative feddback is applied.
This is why golden ears appreciate so much tubes.
This is best appreciated with a professional headphone.
Omnes feriunt, ultima necat.
called "the amplifier distortion story"starts on page 24
continues on page 70
appendix has the calcs for all the ordered harmonics.
Edits: 12/26/24
it was a thing in the 1970s. What's old is new again.
the amplifier distortion story—part1 page 35
Thanx Ralph!
this was my "white whale" for months.
the calcs are in the appendix in part 2 after it continues on page 70
I love that Crowhurst documented all this stuff so long ago. Eliminating TIM was a big deal in the late 70's but Crowhurst was pointing to the exact cause 20 years earlier.
Plus that added issue of why amps sound different while the THD and IMD figures were not showing why.
he always brings the math.
if you know anyone who has a child/teenager struggling with math crowhurst wrote the best math instruction books ever.
no BS.
!
The Mind has No Firewall~ U.S. Army War College.
"crowhurst wrote the best math instruction books ever."
Thanks for mentioning this, I had no idea! A quick Google search came up with quite a few hits for his publications.
dreaded 81st harmonic, in yellow no less,
IIRC that article had something in common with the one reprinted in Glass Audio that included the 81st harmonic.
It also mentioned intermodulations at the feedback node.
I've been repeating this stuff for years now so my Google searches are mostly dragging up quotes from me; not helping. I really should have bookmarked it when I saw it.
BTW if you don't know of Pete Millett's site its a good resource.
"my Google searches are mostly dragging up quotes from me"That's been happening frequently to me too on other topics. I must be an expert now! LOL.
OK, so let's consider whether there might be other means to resolve the issue of cathode-node feedback distortions. You apparently work exclusively with balanced designs, so your techniques have been developed for that mode. I'm all about single ended signal chains, so the ultimate topology for this (if there is such a thing) will probably be different. I'm not dissatisfied with the method used in my latest control amplifier, but it would be nice to have something more universal, a method that doesn't require special consideration for the fact that the source is connected to the same node as the feedback.
Recently I've been looking at pentagrid converter tubes, wondering how linear they might be. These tubes have two separate control grids operating in the same electron stream, and this would seem to present an opportunity. Do you know of any work into using these tubes as audio feedback amplifiers (other than DC-controlled compressors/expanders)? I suspect grid 1 provides a relatively linear response, but the second control grid is a mystery.
Edits: 12/26/24
OK, so let's consider whether there might be other means to resolve the issue of cathode-node feedback distortions. You apparently work exclusively with balanced designs, so your techniques have been developed for that mode. I'm all about single ended signal chains, so the ultimate topology for this (if there is such a thing) will probably be different.
Since the cathode isn't available as a feedback node that has forced us to use the grids. We apply feedback in the balanced domain but the technique for doing it single-ended is identical. The issue is phase and impedance.
Phase obviously so its negative rather than positive. You'll find that feeding the grid requires higher resistance values which should be unsurprising.
We had a customer send us a Leak Point One preamp for which he needed a power supply; it was our job to build that supply since the Leak runs an umbilical which normally plugs into a Leak amplifier for its power. The Point One employs EF86 pentodes with feedback applied to the grids. This allows for a single tube to operate as a phono preamp for MM cartridges and a second EF86 as a line stage with tone controls. It turned out to be a very musical preamp- flying in the face of the traditional 'triodes only' and 'zero feedback' assumptions. It seemed a nice reinforcement of the ideas that Crowhurst wrote about. I really wish I could find that article again!
What Crowhurst mentioned, IIRC, is that the feedback applied to the cathode resulted in the generation of higher ordered harmonics as well as IMD. That correlates nicely with the bad rap feedback has gotten in some circles, such as the SET community. I don't think the SETs guys would like it though simply because as you clean up your amplifier power its very natural to use more of it. So people that say '7 Watts is plenty', if using feedback may find that they need more like 70 Watts (of course it would be playing louder but not sounding louder).
So why not use the two inputs with the same gain?
Cathode has the same gain as the g1. Apply signal to one, and FB to the other. Say, FB to g1 and signal to the cathode?
cheers,
Douglas
Friend, I would not hurt thee for the world...but thou art standing where I am about to shoot.
its more than just gain. The tube isn't linear.
So when the feedback is mixed with the audio in the tube, it gets distorted. When that happens the feedback is generating harmonics of its own. So it would seem that what you want to do is mix the audio and the feedback in a more linear way.
That can be done by using a series resistance with the incoming audio signal and use a simple divider network to mix the feedback with it, before it enters the tube. Now the feedback is able to correct for the non-linearity of the tube.
Let's forget feedback for a moment.
Are you saying that the tube will respond to signal applied to the cathode in a fundamentally different way than signal applied to the control grid?
I have always looked at input signal as something that is simply modifying the bias. Bias being the instantaneous voltage potential between the grid and the cathode.
How does the tube know if that signal is being applied to the cathode vs. the grid?
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
Are you saying that the tube will respond to signal applied to the cathode in a fundamentally different way than signal applied to the control grid?
No.
It will be harder to drive the cathode because the impedance is lower. But the tube responds the same. And that's the problem: its not linear.
Resistors are a lot more linear. The correction Voltage is taken from the output; we're assuming that the circuit isn't linear so that Voltage has distortion which we are tying to reduce. If we mix it with the incoming signal using a divider network rather than the tube the feedback won't be distorted. Instead it will be the undistorted correction signal to compensate the non-linearity of the circuit.
Think of the tube as a house of mirrors. If you try to mix the audio (which is being distorted by the mirrors) in that space with the feedback (which is being distorted by the mirrors) then the result isn't the audio mixed with feedback. Its a distorted version of that.
The article I read by Crowhurst mentioned that in addition to harmonic distortion generation you also get IMD because the feedback node is non-linear. By mixing at the input (in a manner identical to how its done with opamps) the feedback node is far more linear (I'm assuming not perfect since resistors aren't perfectly linear either but they are a lot more linear than any tube).
If feedback is returned to the cathode, and feedback is instead applied to the control grid, is there a problem because the grid is not in phase with the cathode?
In order for this to work the phase has to be opposite. In a PP amp this isn't a problem; you can change it out the power tubes or the drive to the power tubes.
"In order for this to work the phase has to be opposite. In a PP amp this isn't a problem; you can change it out the power tubes or the drive to the power tubes."
This also works out correctly in a line stage with a cathode follower output. It's a mystery why designers haven't exploited this technique to a greater extent.
Reducing one or the other( FB or signal ), is not going to linearize anything. Resistance is by definition linear... :)
Douglas
Friend, I would not hurt thee for the world...but thou art standing where I am about to shoot.
I described the difference between mixing at a non-linear node as opposed to one that is linear (the resistor divider network).
I think you missed my point or else I misinterpreted your comment.
You said to 'linearize' the FB with a resistive network. Not needing to quote you( I think ). Resistors, being linear are not going to make any non-linear input, linear.
Douglas
Friend, I would not hurt thee for the world...but thou art standing where I am about to shoot.
The context lost was the bit about how the feedback has to be mixed with the incoming audio. To do that it must be inverted with respect to the feedback signal that is used when feedback is applied to the cathode.
In a PP amp this is fairly easy since there are usually coupling caps involved with driving the grids of the output stage. They can be set up to drive the other power tube, thus inverting the output. Or the plate connections of the power transformer can be reversed.
Well...yes. Seemed you were going for something other than basic NFB application when you wrote of mixing NFB in a more linear way:
So it would seem that what you want to do is mix the audio and the feedback in a more linear way.
and then using resistors to accomplish this:
That can be done by using a series resistance with the incoming audio signal and use a simple divider network to mix the feedback with it, before it enters the tube.
Thanks for clarifying anyway... :)
Douglas
Friend, I would not hurt thee for the world...but thou art standing where I am about to shoot.
Well...yes. Seemed you were going for something other than basic NFB application when you wrote of mixing NFB in a more linear way:
I'm looking for the same loop gain and the same damping the feedback loop might have if applied to the cathode. The difference is because the grid has so much higher impedance the feedback values have to change to accommodate that.
Opamps use this exact technique to do their feedback. This does mean you need an inverting input. If the feedback is applied to the cathode the amp circuit is usually non-inverting. So some sort of accommodation needs to be made to allow the input to be inverting.
Yes, Leak used the Baxandall tone circuit in the output stage. The phono EQ likewise uses feedback from plate to grid. The EF86 is one of my favorite small-signal tubes, and I'm not surprised this unit performs well.
I uncovered two online references with schematics detailing the use of the two control grids of the 6BE6 for audio. Both sources are from the early 1950s, and they show the tube being used as a mixer. Perhaps the question now becomes whether feedback applied to a separate grid with different distortion characteristics is a valid means of avoiding cathode node distortion. I might test this with the choke-loaded follower I'm fond of when my current project is complete.
UL for a pentode like the EF86 is done with a divider network for the screen grid between the plate of the tube and ground. So the screen thus is providing feedback.
This technique can be done with a couple of triodes in cascode too. But the more gain the better as far is the feedback is concerned- more linear, lower output impedance. No idea if a 6EB6 or similar would do the job but it might be fin to try it out.
I believe nested feedback that includes both stages is key to overall performance. I also suspect the best approach won't be the use of two grids, although I do intend to try the pentagrids when time permits. It just appears likely to me that in the end, the most effective method will see the two signals applied to the exact same node (grid). This seems to be supported by SPICE, as well as the circuit I'm using now. 1kHz distortion is under .002% at the medium output levels I would use most of the time.
Ralph, I completed the difficult switched input wiring in my preamp project yesterday. Still need to finish a few cosmetic items, but it's in my system today, driving a Citation V. I understand now why you're so adamant about this method. Compared to the SS backup preamp I've been using (which is no slouch), the sound of this unit is remarkable. In fact, it's so good, I believe for the first time ever that my system is limited only by the speakers.
I'm thinking now about building a few separate line stages using this technique. Perhaps grid feedback could be used at the input to the phase inverter in a power amp, in order to avoid issues regarding input isolation.
Many thanks for bringing this matter to the forefront. I can confirm that it has an audibly significant advantage over traditional topology.
"With current feedback in a cathode circuit the feedback is effective right down to DC at the low end. At the high end the only modifying factor is the stray capacitance of the tube and its associated circuit. This eventually deteriorates the tube's gain and hence also the feedback. So current feedback in the cathode does not modify the low-frequency response at all, and the high-frequency response is modified according to the distribution of tube capacitances. "
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
.
Have Fun and Enjoy the Music
"Still Working the Problem"
Excerpt:
"Multiplying this out, the fundamental term is Aeg, as before;
the distortion terms are - AFeg and Deg, combined as (D -AF) eg.;
and the other term, - DFeg, represents a distortion of distortion.
As the original distortion is usually small, and is reduced by feedback to still smaller proportions, any distortion of distortion will
be reduced to infinitesimal proportions, and can be neglected."
I read this the other day and briefly thought it might relate. Later I realized Crowhurst is referring to the effects of feedback generally, not the specifics of whether it's injected at the cathode or grid.
one of the great mysteries.
I searched all over american radio history site through all the magazines and ralph told me "great article, but nope, that's not the one"
so it may be
That's a nice article with some good fundamental concepts explained about feedback. I particularly like the discussion of positive feedback, used by both RCA and Harman Kardon (and probably many others) to good effect. Unfortunately, neither this nor the second article hit the point Ralph is making regarding cathode-node feedback distortion. I'm hoping we find it one of these days!
this
if you scroll down you'll see a blue box with next/back or you can go up the tree using home.
its a whole chain of stuff that's not quite like any other crowhurst stuff.
| FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
||||||||||||||
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
