OK I understand the perils of NFB, because the correction must always be delayed and therefore incorrect, but it seems to be really useful in controlling a speaker - particularly in the bass department.
So looking at the RIAA circuits and the frequency dependent feedback it strikes me that the same idea could be applied to power amps - like for instance a push-pull tube amp. So if you add a bass boost in a driver stage circuit and a NFB filter that adds more NFB with more bass, you'd have an amp that had hardly any NFB at midrange and none at treble, but lots at the bass.
You'd obviously still design for low noise and distortion like in a no FB amplifier, and only use the NFB around the last output stage, but you'd get the proper bass control I think.
Has anyone tried this?
What do you think?
Has it been done before?
If not - no one can patent it now either 😉
I guess one would call it FDNFB Frequency Dependent Negative Feedback, or FNB for short!
So looking at the RIAA circuits and the frequency dependent feedback it strikes me that the same idea could be applied to power amps - like for instance a push-pull tube amp. So if you add a bass boost in a driver stage circuit and a NFB filter that adds more NFB with more bass, you'd have an amp that had hardly any NFB at midrange and none at treble, but lots at the bass.
You'd obviously still design for low noise and distortion like in a no FB amplifier, and only use the NFB around the last output stage, but you'd get the proper bass control I think.
Has anyone tried this?
What do you think?
Has it been done before?
If not - no one can patent it now either 😉
I guess one would call it FDNFB Frequency Dependent Negative Feedback, or FNB for short!
Have a look at RDH4, chapter 7. I believe I referred somebody to this excellent resource only few days ago but I suppose it wasn't you.
Take a guess 😀 (tip: the correct answer starts with hell yea and ends with an h)
In terms of control system theory, you're basically splitting the gain-plus-feedback system into a system where the feedback H has some frequency response, but the result of that is an inverse overall response 1/H (if gain is high, which is usually fairly true, otherwise it's G/(1+GH) to be exact), so you have to compensate it by pre-filtering with H again to get a flat response. In and of itself, that would work fine.
Thing is, you can't have big NFB at LF anyway, because if nothing else, sheer physics of the transformer foils your plans (and such plans are usually foiled with the juicy farting sound of transformer saturation). So your gain G already has this high pass response, meaning your overall response simply won't be the inverse that you wanted, as above.
Due to the OPT, tube amplifiers cannot have large NFB at LF. You could come arbitrarily close, e.g. by tweaking your filters to pinch ever closer to the LF limit (while somehow trying to avoid phase shift..), with a sharper drop at cutoff, but never exact. And trying to push this limit will inevitably expose the inherent nonlinearities of your amplifier, making things just that much harder to analyze.
Tim
Thing is, you can't have big NFB at LF anyway, because if nothing else, sheer physics of the transformer foils your plans (and such plans are usually foiled with the juicy farting sound of transformer saturation). So your gain G already has this high pass response, meaning your overall response simply won't be the inverse that you wanted, as above.
Due to the OPT, tube amplifiers cannot have large NFB at LF. You could come arbitrarily close, e.g. by tweaking your filters to pinch ever closer to the LF limit (while somehow trying to avoid phase shift..), with a sharper drop at cutoff, but never exact. And trying to push this limit will inevitably expose the inherent nonlinearities of your amplifier, making things just that much harder to analyze.
Tim
There is another way to look at this, and that is to design loudspeakers that are compatible with the source impedance of a typical tube amplifier, or at the extreme to design a speaker for a specific amplifier or family of amplifiers as I have done.. 😀
This presupposes the fact that you are into designing and building loudspeaker systems - something I thought I would never do, but now rather enjoy. All the usual caveats apply, designing a good sounding (and measuring) speaker systems requires the proper design and measurement tools, money, patience, and understanding. (Or very good guidance from more experienced individuals willing to advise.) Good results aren't guaranteed, but are attainable with care. I'm a very happy camper freed from the shackles of what someone else thinks I should be listening to.
This presupposes the fact that you are into designing and building loudspeaker systems - something I thought I would never do, but now rather enjoy. All the usual caveats apply, designing a good sounding (and measuring) speaker systems requires the proper design and measurement tools, money, patience, and understanding. (Or very good guidance from more experienced individuals willing to advise.) Good results aren't guaranteed, but are attainable with care. I'm a very happy camper freed from the shackles of what someone else thinks I should be listening to.
Last edited:
Interesting article, he puts in a lowpass filter and then flattens it with the NFB, is that really the same as changing the amount fed back?
I almost think it is but I'm not 100% sure!
While I think this is probably true, the average punchy push-pull seems to manage enough NFB to drive almost anything, whereas the most delicate SE amp has almost or exactly no NFB at all, so I would expect you could vary between these extremes without too much issue - no?
Most people who do this however don't bother with bass and just add a sub, which is to my mind completely wrong and can never sound as good as a well balanced full range speaker that at least goes to 40Hz. At that level you will have to have a decent sized cone (8" and up) to play with and below 100Hz NFB is pretty happy anyway.
I guess it's like building a crossover into the amp itself so that you get a servo action (or at least at much NFB as a typical Williamson design) for the bass driver and clean NFB free mids and highs.
Obviously active crossovers and multiple amps (SE for tweeter & mid, solid state for bass) would also solve the problem - but you end up with loads of boxes and cables then, so it seems like a nice engineering compromise to improve the sound of an amp, SE or pushpull.
I thought about active X-O and can still do it at any time, but that said I am quite happy with the passive X-O (second pass I'll admit) and eschew all those extra amps and electronics. My system is -3dB at about 32Hz and is quite convincing with just 8Wrms per channel in my modest sized room. The speakers aren't modest sized though.. See the old Onken threads for an idea of what I built.. 😀
- Status
- Not open for further replies.