• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Difference bewteen ultralinear & triode mode?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I see, negative feedback does seem to add odd order harmonics, though it cancels out even more of it. It is still beneficial from all the evidence. Lower output impedance being the greatest benefit.

There is not enough gain in most tube amps to apply lots of negative feedback. Having installed a NFB switch on most of my amps, I still don't like negative feedback. Having some harmonic distortion in the amp is rather pleasing to listen to. I would only add negative feedback in a SET amp to keep the damping factor greater than 4.

One of my SS amps have less negative feedback than the vast majority of SS amps, and no global feedback, and it sounds great! Low and moderate amounts of negative feedback has worked out great for me so far.
 
Last edited:
cotdt said:
I see, negative feedback does seem to add odd order harmonics, though it cancels out even more of it.
It can add even orders too, although not if the basic amp has no even order distortion at all.

Having installed a NFB switch on most of my amps, I still don't like negative feedback.
Generally not a fair test. When you add negative feedback you not only reduce distortion, you also reduce gain and widen frequency response. You may also cause peaks to occur at LF and HF extremes. NFB must be designed in, not slapped on as an afterthought or simply removed. I suspect that many people who don't like NFB have done the same as you.

It's like painting a wall, then complaining that the paint peels so deciding that you prefer unpainted walls - many paints require some surface preparation and an undercoat. You either need bare walls, or prepared and painted walls: unprepared painted or prepared unpainted are both inferior.

Having some harmonic distortion in the amp is rather pleasing to listen to.
Yes, some seem to like it. Few admit it.
 
artosalo said:
Can you give an practical example with vacuum tubes where these higher-order terms can be observed and measured ?
A variable-mu valve (remote cutoff) with a small cathode degeneration resistor. The natural 3rd order from the valve, and the re-entrant 3rd order from the feedback happen to have opposite phase so can cancel. The resistor should be about 1/3gm (or 1/2gm - I can't remember the factor).

You can see a similar effect, but perhaps for a different reason, in the intermodulation curves for a variable-mu valve (see Philips datasheets e.g. EF85?) without the resistor. There is a sharp peak in the allowable signal voltage for 1% 3rd order distortion, at about the usual bias point - which is why this is the usual bias point. I suspect that cathode coating resistance plays the role of the degeneration resistor.
 
Whether these extra terms are noticeable or not is a separate issue, but they are definitely there - unless somehow tubes manage to avoid using conventional mathematics.

I wanted this time to experiment with mathematics i.e. with LT Spice if I can find recognizable higher order harmonics to rise when small amount of NFB was used.

I used a very linear CCS loaded triode stage with no NFB and 6 dB voltage NFB. In both cases the output level was equal as well as the total load impedance.

Without NFB the harmonics from 7th to 9th were more than 100 dB below reference signal (1 kHz). This level represents THD below 0.001 %.

The results I got were surprising.
6 dB NFB reduced harmonics from 2nd to 6th between 5.4 dB to 7.4, which is expectable, but the reduction of 7th to 9th was much higher ( 12 db to 32 dB ), which was quite unexpected and fully against that has been stated here.

An externally hosted image should be here but it was not working when we last tested it.



An externally hosted image should be here but it was not working when we last tested it.


My conclusion now after two analysis, one with actual amplifier and one with "mathematics", is that it is incorrect to say that low NFB should be avoided in tube amplifiers because this creates high order harmonics.
If it does this, the level of such harmonics are less than negligible.
 
artosalo said:
My conclusion now after two analysis, one with actual amplifier and one with "mathematics", is that it is incorrect to say that low NFB should be avoided in tube amplifiers because this creates high order harmonics.
In your simulation you may have been seeing cancellation. Unexpected with that circuit, but possible.

Whatever your measurements, simulations or whatever say, it remains a fact that re-entrant distortion exists. The maths is clear. The fact that high orders changed by any amount not equal to the feedback ratio confirms this. Whether re-entrant distortion is significant is a separate issue.

I'm not sure whether you disbelieve the maths of feedback, or just misunderstand what I am saying. Maybe I have not been expressing myself clearly enough. I said people who worry about re-entrant distortion should avoid low amounts of NFB. I personally don't worry about it (although I admit I used to). Provided the forward path is sufficiently linear then re-entrant distortion, although always present with NFB, will do no harm.
 
Whatever your measurements, simulations or whatever say, it remains a fact that re-entrant distortion exists. The maths is clear.

I have not seen this clear "math" you refer and I have not seen any test results done with tube circuits that would prove the theory is correct.
Can you provide something more that supports your opinion ?

Whether re-entrant distortion is significant is a separate issue.

To me this is the main issue. I am considering this subject at the practical situation only.

I said people who worry about re-entrant distortion should avoid low amounts of NFB.

Why should they avoid using low NBF ? To just feel comfortable ?
Even low NFB seems reduce the level of higher order harmonics, not to increase. If there were new harmonics created the level of these would be at most negligible level, assuming 120...140 dB below fundamental signal.

I think people who worry about re-entrant distortion should open their eyes and see the reality, which surely is much better than they have supposed.

Below is a graph that mostly will be referred when this subject arises.
We all can now see that the reality is far from what the graph shows.

An externally hosted image should be here but it was not working when we last tested it.
 
artosalo said:
I have not seen this clear "math" you refer and I have not seen any test results done with tube circuits that would prove the theory is correct.
Can you provide something more that supports your opinion ?
Take a polynomial. Square it. Notice that it now has higher-order terms which were not there in the original polynomial. That is the maths (actually, this is a simplification - reality is even worse). If you understand how feedback works and how distortion is generated then that will convince you. If you don't understand these things then you first need to learn them before you can be convinced. It is not a matter of opinion, just fact.

Many real circuits are likely to have enough intrinsic higher order distortion that the effect will be hidden, as I keep saying. Not absent, just hidden.

I think people who worry about re-entrant distortion should open their eyes and see the reality, which surely is much better than they have supposed.
I agree. I have admitted that at one time I did not agree. The point I was trying to make was that people who worry about re-entrant distortion caused by feedback should not suppose that they can have less of it by having less feedback. To get less of it they either need no feedback at all (to avoid re-entrant distortion), or lots of feedback (to suppress all distortion). People who understand distortion and feedback will not worry, but simply use the appropriate amount of feedback.
 
I may be possible that doing the mathematical operations you mentioned the result will be as you said, but I am looking for the connection between the mathematics and real life.
So I look for any test results or any reliable reference that proves that this is also the case with vacuum tube amplifiers.

I found this:
http://www.its.caltech.edu/~musiclab/feedback-paper-acrobat.pdf

Please study the pages 17...23.
The results achieved here conform with my earlier test results and later simulations.
 
The results in that paper are compared for equal input levels (see p22), not equal output levels. Therefore not comparing like with like.

You still seem to be asking for proof that feedback in valve circuits obeys the normal laws of mathematics. I have no answer which will satisfy someone asking such a question, because by asking the question they are denying the basis of the correct answer.
 
Yes, I am still looking for information that will connect the math with actual tube amplifying circuits. Even one reference would be fine to see.
But I am not denying anything. It may be possible that low NFB can create high order harmonics, but I am wondering why I can not find those and why nobody else has not found them, as it seems to be.
It would be unusual if the behavior you claim to exist could not be found in any research or publication, if it really existed.

Unless you or somebody else have any new information to show, I am ready to finish this subject.
 
No new information, but I will have one last final try at getting you to correctly interpret the information you already have seen.

To keep things simple, assume an amplifier or device which is fairly linear but also has smooth non-linearity. This could mean a triode or BJT or a complete amplifier. All I am excluding is nasty stuff like crossover distortion or peak clipping. Put a signal through the amplifier. Coming out at the other end you will see the original signal plus distortion (plus some noise, but we will also ignore that). The distortion will go up to all orders, but we can typically only measure down to around the noise level. As the non-linearity is smooth the higher order products will be smaller than the lower order products. These all arise from the input signal being multiplied by itself by the terms in the in-out function of the device/amplifier. There is not a direct one-to-one correspondence: for example, a third-order term causes both third-order and first-order outputs, a 5th-order term gives 5th, 3rd and 1st-order outputs.

Now add some feedback. Even with low amounts of feedback the situation gets more complicated. The output now gets not just products from the amplifier distortion, but the amplifier distorting its own distortion. So at every level above 2nd, the output now consists of both intrinsic distortion and re-entrant distortion - but in both cases also reduced by the feedback. The re-entrant distortion can be the same phase as the intrinsic distortion and so add, or the opposite phase and so subtract and partially cancel. Then the result is reduced by feedback. Generally, the worse the amplifier was before feedback the greater will be the contribution from re-entrant distortion. Anyway, the net effect depends the balance between new distortion being created and all distortion then being reduced by feedback. An FFT of the output cannot distinguish between intrinsic distortion and re-entrant distortion; you just see the total at each frequency. That is why you 'cannot see it' - it looks exactly the same as intrinsic distortion. To see it clearly you need to start with a device with large low-order terms and very small high-order terms, so the effect is not hidden by intrinsic distortion.

Note that nothing I have said is device-specific. It applies to all amplifers and all devices. That may be why you have never seen a publication showing the effect for a particular device: there would be little point. The paper you referenced in your post 76 probably would show the effect if they had used equal outputs instead of equal inputs. As a rough estimate, take their results (e.g. on p22) and multiply each distortion product by 2^(N-1), where N is the order (or add N*6dB to each). I think you may then see the effect. They were using 6dB of feedback, so the input voltage seen by the device was halved. Doubling that, to get the same output, means that distortion products are potentially increased by 2^(N-1) (apart from complications created by changes of sign causing partial cancellation). Can't guarantee anything, as the device might have too much intrinsic distortion.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.