Member
Joined 2009
Paid Member
I found this an interesting paper, it compares ideal square law devices with ideal tubes.
http://www.its.caltech.edu/~musiclab/feedback-paper-acrobat.pdf
http://www.its.caltech.edu/~musiclab/feedback-paper-acrobat.pdf
Eugene Karpov showed 7th order and more in a paraphase SE amp. Google Karpov SPECTRA and you will find the article.
Never mind, it is here:
http://www.next-tube.com/articles/spectr/spectrEN.pdf
...and it goes to 10th harmonic.
Never mind, it is here:
http://www.next-tube.com/articles/spectr/spectrEN.pdf
...and it goes to 10th harmonic.
There were higher order harmonics at readable, but low level.
I limited the results to 7th.
Here is the plot of 0 dB:
Kuva 17850 - Kuva.termiitti.com
I limited the results to 7th.
Here is the plot of 0 dB:
An externally hosted image should be here but it was not working when we last tested it.
Kuva 17850 - Kuva.termiitti.com
Good reference. He is showing that most tubes have significant harmonics to at least the 5th if not 7th. Conditions: pentodes connected as a single ended triode putting out 1 watt. Most have 3 to 5% 2nd harmonic and about .04 to .07% 5th harmonic. Maybe .01% or less 7th, just above his noise floor of -85dB or so.
One thing to keep in mind is that a modern spectrum analyzer with a wide dynamic range will show quite a few harmonics before they fall into the noise floor. The presence of .01% 7th harmonic is not inconsistent with a view that a single ended amplifier has "primarily lower order distortion". If you are an early researcher and you can draw out a sine wave with 5% 2nd and a little bit of 3rd and it looks like what you see on a scope, then that is a proper generalization.
What we can't say is that a particular tube has just 2nd or just 3rd.
David S.
Also a very good paper. He shows a vast range of intermod products for all types of amps (FET, bipolar and tube. Single ended and push-pull) but is simulating and plotting on a 200dB window.
He comments on a single ended triode simulation: "the 3/2 power law in equation 39 can be expressed as a power series that has terms of all orders...Thus we can expect the triode results to be somewhat like the BJT results above, in that distortion products of all orders should appear in the spectrum of a simple amplifier. The question is: does the presence of feedback make matters better or worse?"
Answer: better.
David S.
Definitely my design philosophy and experience as well. All too many lousy open loop designs out there where they just poured on the gNFB to sweep their mistakes under the carpet, and hoped no one would notice. I always listen open loop only for at least a week before finalizing a design so far as corrective NFB is concerned. No choice since I "eat my own dog food", and what I listen to is what I designed. It better sound right because I have to live with it.
Add too much gNFB to a hollow state design and what you get is a very "solid statey" sound, and not good SS sound either. (And there's no reason SS has to sound as bad as it does in way too many cases.)
It probably would, but that's also topology dependent. I didn't do such a design, and P. Millet doesn't seem to have one either.
Listen, decide, correct and improve.
That's the point! There is no ingredient - independent recipe in any cuisine! And what a single loop can't correct multiple properly applied loops definitely can. However, I personally prefer to use topologies that don't create errors that other topologies struggle to minimize, but since everything is optimization and search for acceptable compromises, this recipe is not always right.
negative feedback seems like a design compromise.
What are you trying to fix when using NFB?
Noise?
if your answer is noise most commonly its the power supply design. other things can be used.
Gain?
Well, input impedance mismatch is the culprit most of the time.
100k is rarely a valid value, that has popped up in "oh it works with 100k" schemes on those DIY designs. lets use this fancy attenuator.
There is something about the negative feedback that many people either overlook or not really understand-- It modifies the frequency resonance of the whole circuit regardless where the NFB injection point is at.
but there are ways to remove noise without feedback.
actively injecting into a current amp ( Akido ) or using a part not everyone knows about, the common mode capacitor.
its not really talked about. on a push pull amp remove the negative feedback. now its noisy. now take something for example of a 2 output tube 5 k plate to plate push-pull circuit, put a 200-120 pf cap from plate to plate on the output stage.
I like to see the analysis.
What are you trying to fix when using NFB?
Noise?
if your answer is noise most commonly its the power supply design. other things can be used.
Gain?
Well, input impedance mismatch is the culprit most of the time.
100k is rarely a valid value, that has popped up in "oh it works with 100k" schemes on those DIY designs. lets use this fancy attenuator.
There is something about the negative feedback that many people either overlook or not really understand-- It modifies the frequency resonance of the whole circuit regardless where the NFB injection point is at.
but there are ways to remove noise without feedback.
actively injecting into a current amp ( Akido ) or using a part not everyone knows about, the common mode capacitor.
its not really talked about. on a push pull amp remove the negative feedback. now its noisy. now take something for example of a 2 output tube 5 k plate to plate push-pull circuit, put a 200-120 pf cap from plate to plate on the output stage.
I like to see the analysis.
My understanding is that multiple loop feedback is generally wasteful of potential feedback (gain) in that you end up having feedback on circuit sections that are more linear, as well as on the non-linear sectons. I would imagine that most amps have the higest distortion (and highest signal strength) on the output stage. Having minor loops elsewhere will usually reduce the potential distortion reduction of the less linear output stage.
David S.
Hi David; your generalization as well is limited by certain topologies that have "most distortions in output stage". Here is the example: trying to compensate "most distortions" in output stage the driver stage goes out of linear regime, especially on higher frequencies, creating nasty dynamic distortions.
Yes, feedback is waste of amplification factor. No matter, is it global, or local. But strictly speaking, it is not waste: it is trade-off. We loose what we have in excess for something that is harder to achieve directly. And usage of multiple loops is the way to optimize the whole system, carefully and well balanced.
Yes, feedback is waste of amplification factor. No matter, is it global, or local. But strictly speaking, it is not waste: it is trade-off. We loose what we have in excess for something that is harder to achieve directly. And usage of multiple loops is the way to optimize the whole system, carefully and well balanced.
These curves of small amps may shed some light on what can happen with global feedback.
Notice how some stages become distorted before clipping sets in and change the overall curve and others do not. So which is better low distortion over most of the range but creeping up near clipping or slightly higher distortion but clean until clipping?
Trick question, the best amplifiers are low distortion overall until clipped!
Notice how some stages become distorted before clipping sets in and change the overall curve and others do not. So which is better low distortion over most of the range but creeping up near clipping or slightly higher distortion but clean until clipping?
Trick question, the best amplifiers are low distortion overall until clipped!
Attachments
It is not necessary right. Consider crest-factor of real musical signals. If all signals in all channels before the mixing are well preserved you would need an enoemous dynamic range of the whole channel, from microphones to speakers. But when say vocal microphone preamps gradually increase distortions the overall system would sound better.
It is not a waste if it is traded for distortion reduction, but to do that it needs to encompass the most distorted stages.
For example 20 dB of feedback will potentially reduce distortion by 10 times. If it is a global loop it should do so. (The graphs of the OP show this very clearly.) If instead you split it into two 10dB loops around 2 different stages (first half and second half of the amp, lets say) and one section is relativley high distortion while the second is relatively low, then both of those sections will have their distortion reduced by 3.16 (sqrt of 10). If the cleaner of the two sections has very low distortion relative to the dirtier, then in the end you are only getting significant reduction of distortion for the dirtier stage by the lesser amount (of 3.16 rather than 10). You have "wasted" half of your feedback.
I'm not saying that local feedback is a bad thing, just that if you can only afford a certain amount of feedback (in dB of gain loss) then the greatest distortion reduction comes from global feedback rather than multiple local loops.
Read Crowhurst on this. He discusses it in several older magazine articles.
David S.
Thank you.
Unfortunately I have no time to explain in details here what articles in magazines don't.
If you want to talk about compression, that is a different issue. It is unusual to expect a power amp to provide the compression. Of course if your source is a CD the dynamic range of the music can be less than the power amplifier!
But you just wanted to point point out all absolutes have exceptions. I am absolutely sure of it.
If you want to talk about compression, that is a different issue. It is unusual to expect a power amp to provide the compression. Of course if your source is a CD the dynamic range of the music can be less than the power amplifier!
But you just wanted to point point out all absolutes have exceptions. I am absolutely sure of it.
Right. And my power amps actually use compression (well, optical limiters actually). It sounds in power amps less nasty than clipping. It is trade-off exactly for what you advocated: minimum of distortions up to clipping level.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.