|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
77.43.73.35
In Reply to: RE: Feedback, no feedback and selective feedback posted by Lew on October 10, 2010 at 15:08:35
'Mutatis mutandis' I think this selective feedback network can be applied to the majority of commercial circlotrons.
The selective feedback network is made of R13, R22, R21, R20 and C5.
The feedback signal is picked across the normal speaker output (paying much attention to the polarity to prevent wild oscillation).
In this example there is room for about 11 dB feedback at low/mid frequency, 7.5 dB at 10 KHz, 3 dB at 20 KHz and zero feedback above 30 KHz.
Changing the value of C5 will allow to increase or decrease the 10-20KHz feedback.
Changing the value of R13 and R20 allows to trade in more low frequency feedback (i.e. lower output impedance) in exchange for reduced mid-band gain.
Best Regards
Luca
ecc230
Follow Ups:
By installing a resistance between the cathodes of the input tubes, the coupling between them is reduced. There is a price paid for this: reduced differential effect, which results in lower gain and higher distortion. There is also the need for some means to adjust the resulting plate voltages- a big mismatch in them means distortion.
A variable resistance installed in the cathode circuit, with the wiper tied to the CCS can be used to solve this problem, but it will be immediately seen that other problems are introduced, leaving the distortion issue nearly untouched.
Also, to be truly effective, a 2-stage CCS is highly recommended any time a differential amplifier is executed, along with a B- voltage of sufficient value such that a true high impedance effect can be realized in the CCS.
It is possible to mix the feedback signal with the incoming signal with a resistive divider network, thus eliminating the cathode coupling problem in the first stage. The increased gain can then be used to assist with additional feedback, if desired.
However, it should be noted that while feedback **seems** to decrease output impedance, that if that were really the case, the output power at lower impedances would increase, and it doesn't. What it *does* do is push the amplifier design towards behaving more like a voltage source, but unlike a transistor amp which might double power as load impedance is halved, in this case the total output power into higher impedances will be limited by the power that the amplifier is capable of into lower impedances.
IOW, a set of ZEROs will still be a good idea on lower impedance speakers.
Thought I saw an article once in Glass Audio in which it was shown that a pair of small value resistors between the cathodes, such that the current source feeds in at the junction of the resistors rather than directly to the cathode(s), actually resulted in a reduction of distortion, at the expense of a small amount of gain. Otherwise, I am not qualified to argue. In that article, the dual-differential cascode was referred to as a "Hedge Amplifier".
These are the (simulated) output spectra @ 1.2A peak current (in 8 Ohm speaker) of my dummy schematic with (top) and without (buttom) the 100 Ohm resistors.
As at 300 Hz negative feedback (with the 100 Ohm resistors) is working 'full steam ahead' the overall distortion is MUCH BETTER than without the 100 Ohm resistor.
And this is exactly as it should be according to known electrical theory, because negative feedback not only reduces output impedance but distortion too.
You can also see that in the case at hand odd distortion with zero feedback exceeds 0.1%, the supposed upper limit recommended by the still unknown/unpublished General Electric study of 1960 mentioned so many times by Ralph.
Best Regards
Luca
ecc230
Luca, The 3rd order harmonic distortion is clearly reduced by the presence of the 100-ohm resistors (what I use in my amps), but is it not true also that higher order distortion products are increased, WITH the resistors, based on what I am seeing in your data? Thanks.
That's feedback.
Moderate feedback (like in this case) reduces low order harmonics (about 12 dB), but creates higher order ones.
In real world you can not have, at the same time, a completely filled in cask and a drunk wife (this is my best translation of an Italian proverb).
I suppose this is the reason why excessive feedback has such a bad reputation and why it is generally accepted that a good basic design (with inherent low distortion) is needed before ever thinking of applying NFB.
I realized this fact by extensively playing with simulations and this is the main driver for adopting frequency selective feedback instead.
If feeedback is vanishing away (starting from a few KHz), the creation of potential high order harmonics, that would be otherwise created by flat frequency feedback, is automatically prevented.
In conclusion, I think that, based on a careful case by case assessment, frequency selective feedback can be effective in improving audio performance.
Best Regards
Luca
ecc230
In conclusion, I think that, based on a careful case by case assessment, frequency selective feedback can be effective in improving audio performance.
Luca, I hope you see that in this quote we are in complete agreement, and that in fact that has been the point of my initial link, as well as the succeeding arguments.
I read about the GE study in a book, but so far I have not seen that anyone has published a report of it online, FWIW.
To be clear though, GE was saying that the 3rd, as well as the 2nd and 4th, are not regarded by the ear as irritating; it is the 5th, 7th and 9th where the problem lies. If you spend any time on the SET Asylum, you will find that that is a commonly-held viewpoint that is not unique to me. As your simulations show, these troublesome distortions seem to arise as result of the use of feedback.
It is for this reason that the cathode resistors as mentioned in the differential amplifier are not desirable- sure, they reduce lower orders (although the even orders will be canceled, so we are really only talking about the 3rd harmonic), but at the price of increasing the higher orders.
Agree with You Ralph about GE harmonic distortion study and their conclusion.Same is for SET example You give,everybody know why in SET Amps GNFB is more Problematic than PP Amps:Basicly beacose there is no automatic distortion canceletation like PP amp doo,SET Amps generate allmost all harmonic distortion products,so GNFB for SET Amps is `must` and value of inserted GNFB is allways compromise beetwen real `Ear` Audio performance and acceptable measurment(Osciloscope) performance,and this is allways valid for SET higher Audio spectrum reproduction.
Any way beside that from above,the Bigger problem for most SET Amps is actually OPT(price/sound limitation/ degradation isue),think the acceptable and real good `Ear` sounding SET Amp is relative low output power(~10W)A/A2 class SE OTL Amp with moderate GNFB.
Best Regards
__
Enlightened Evolution-Astral Projection
Edits: 10/21/10
Best performance corresponds to 6-13 dB feedback.
http://www.audioasylum.com/cgi/vt.mpl?f=tubediy&m=190082
This is my last project and I have recently added a top panel switch to select 9dB or 14 dB NFB.
Even though the amplifier is very natural with zero feedback too when using 14 dB setting the absolute performance is (believe me or not) undistinguishable from my OTL!
In my opinion high plate voltage, almost zero PS impedance and high bias current (i.e. true class A operation) allow tetrode amplifiers to match OTL triode performance.
Best Regards
Luca
ecc230
And all inmates are now in the position to fully developing their own (informed) opinion.
As a final remark, I think that we all also agree that the ear is the final judge.
If somebody prefers zero feedback this is good for him.
If somebody else prefers some moderate feedback this is also good for him.
And those who still like the old Leak (30dB NFB) or McIntosh (about 20 dB NFB) tube amplifiers, again, it is good for them too.
Best Regards
Luca
ecc230
So, should I get those resistors out of there, in my amps? Wish I could find that Glass Audio article which gave me the idea to use them in the first place. Without the FB resistors, my amps would have a very high input sensitivity (meaning a very low signal voltage would drive them to full output). I am using type 6900 triodes in the dual-differential cascode. These are super-5687s with tremendous Gm and low Rp.
Hi LucaHere is some old(1966) Philips AG9007 60W OTL Circlotron HI- FI Tube amplifier design.
Found interesting in this design that Philips designers used 2x2 symetrical Negative Feedback.
First NFB loop is symetrical from output stage to driver stage only,and second GNFB loop is symetrical from output stage to input stage.
Best regards
Check this link:
__
Enlightened Evolution-Astral Projection
Edits: 10/14/10
Hi Lucca
Basicly agree with Your posted schematic,results and conclusion,but I think next:
Is better to have Only for Anode Follower driver stage separate supply direct from oposite output(Circlotron)power stage two B+ supply.
This way of`bootstraping` between driver stage, AC coupled to grids of output power stage, achive some low value of positive feedback for anode of driver stage and results in bigger voltage sfing and overall lower distortion.
Of course ewerything else including GNFB remain the same like You posted in Your schematic
This solution for driver stage is actually esential idea of Mr Tapio Koykka,inventor of Counter Parallel Push Pull(CSPP) Amplifier(Circlotron),and this solution is succesfully used in many Circlotron Amplifier,VOIMA,Philips,Electro Voice...
Best Regard
__
Enlightened Evolution-Astral Projection
Thank you for your suggestion.
Best Regards
Luca
ecc230
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: