Bonsai,
Sorry, got my wires crossed on the amps part of my last message. I have just remembered it was another regular forum member's design which was a CFB. My apologies.
I wish I could train my mind into a more analytical way of thinking. After reading your post it has just occurred to me that the servo is just an "active" pot doing the same job, injecting current.
Thank you for your reply.
Paul
JCX this is not about comparing low and high feedback designs, CFB as we know cannot match VFB in loop gain as such. CFB designs can be be high feedback too which is the route I take. VFB cant however easily match CFB when it comes to ULGF. This allows CFB designs to match the loop gain at frequencies above 1Khz. Couple this to the fact that a diamond buffer is many times more linear than a LTP and one can deduce that CFB design will be more linear when it comes to high frequencies. This is in typical designs, one can design a very complicated LTP based amp to match this CFB performance but then again one can also design complicated CFB. We talking here of real world power amp designs, not what can be achieved with opamps with GHZ transistors. My CFB design cannot match a blameless design using TMC in THD at frequencies below 1Khz but at higher frequencies the blameless is left in the dust and this with less active devices.
Lately VFB types using CFB principles have been appearing, this allows the best of both worlds. The penalty to pay is noise but this is not such a big compromise as some make it out to be.
Theoretically very high currents can be used but due to stability requirements there is a limit as you could end up with ULGF of 40 Mhz which would need heavy compensation. With this requirement one can then degenerate the diamond buffer to lower the ULGF which would in turn lessen the compensation required. CFB topology is very similar to the LTP when it comes to these details. Not much is written about CFB power amp design because audio amp authors havent studied all the details as of yet. Remember that in CFB design the diamond buffer operates in class AB if pushed. In terms of noise the diamond input operates in he same way as does LTP designs and compromises need to made.
I operate my diamond at around 2 mA and use low rbb transistors. In terms of numbers Id say 6nV/rt Hz. Unless one has ears like a bat one could actually hear this. As for THD its pretty easy to achieve 10 ppm 20 Khz at full power. A LT simulation of my amp gives 3.6 ppm THD20, leaving 99 % of real world LTP designs in the dust and with less complication. I can use modern CFB techniques such as feedforward to get even better numbers at the cost of a couple of caps but at these levels its not neccesary. Carefull attention needs to be made with regard to the power supply though.
Another detail one should look at is stability, CFB is easily compensated and high phase margins obtained. If I lowered margins to the levels I see here stated for LTP based designs using TMC, TPC or Cherry I would get even better THD performance. I will not compromise stability for outright THD performance though.
I think it will be Cordell, he seems more flexible and open to new concepts although CFB is not new. Change from conventual thinking is difficult for us humans as we already know.
Last edited:
There is no technical or psycoacoustics support for rising THD with frequency either but ideally its not wanted from a engineering point of view. Why is it strived for as can be seen by opamp manufacturers ?? Your correct in your assesment though, no disagreement there. I reject all objective audiophile mambo jumbo on this one.
Manso, I simulated my VFB amps 1-2 ppm at 20kHz but never succeeded to came close to it with CFA simulation. I know it is not so important to get so low distortion but still it's techical chellenge. Could you present your CFA power amp here?
BR Damir
Is there a "forum accepted" way of quoting (measuring / simulating) specs like THD? Eg. at specific power levels or percentage of max power / clipping voltage? Then we could compare apples with apples.
If you cap couple the speaker to the amp... you can put a DC supply to the speaker and adjust the speakers vC centering for lowest speaker distortion...... the variables in spkr mfr'ing are such that this can make quit large reductions.
Just an experimental approach to see how much it matters and can be affected... and then listen/measure.
Thx-RNMarsh
I really don't believe these claims of vastly improved CFA high frequency stability, higher available audio loop gain - we use the same output devices with the same ft, packaging parasitics, operating at the same bias points
the high V, power, output Q are the high frequency feedback limiting devices in audio power amps - not the "speed", type of the error amp/loop gain circuitry
the high V, power, output Q are the high frequency feedback limiting devices in audio power amps - not the "speed", type of the error amp/loop gain circuitry
Last edited:
It is surprising, as CFB should have more ramining feedback ratio at 20KHz than VFB, for an equal ratio at 500 hz. (Less phase rotation for CFB).
I remember when 30 years ago my friend bought budget priced system NAD3020/Tannoy Mercury M20. I was shocked because it sounded better than my own expensive system (Quad 34/405 + 15" Tannoy Berkeley MK2 factory made, not kit). It took me years to understand that it sounded better because of the current feedback power amp circuit, very similar to the circuit of old Revox amp that forum member Banat has. It remains mystery for me if the NAD circuit was chosen because of superior sound quality or was it purely economical decision.
Exactly . I soon plan to use this with an OB speaker to serve in place of a brick wall filter . I think it can be made to work . The main amp will be capacitor couple which helps define a lower frequency limit and is excellent protection . The damping factor will be low and Qts high . The DC will be selectively created . It might not work . A bit of fun and so much nicer than a brick wall filter . I can design them so it doesn't bother me . It just isn't the nicest idea .
BTW . I suspect nobody reading this can easily determine if a coupling capacitor is being used to speakers as long as large enough . Also , the worst quality capacitors probably out perform the best transformers . Measurements confirm this . > - 100 dB .
With VFB I can get easily 60-70dB of the 20kHz loop gain but with CFB hardly 40dB.
Here I simulated my GaiWire HEC CFB power amp at 50W/8R and THD20k is 0.002225% with 35dB at 20kHz LG.
This CFB is probably to complicated but at the moment I did not succeed with simpler shematic to come close to it.
BR Damir
Attachments
Last edited:
I yet tried to explain why. But it seems words, even consistent from several people, are not enough to convince. So, don't believe: try.
yes I see "words" - not technical arguments, just bald assertions - not even working sim of CFA to compare (with all part models included - unlike your earlier post - which I didn't dismiss - they just don't run in LTspice from your posted files)
I am arguing from Bode Phase Integral "limits of feedback" perspective - we are limited in audio power amp design by the power output devices - practically the drivers too, and triple follower output is still needed if you really want to match VFA high loop gain
bootstrap cascode a diff pair and you can use just a pair 300 MHz ft Q added to the "signal path" count vs CFA front end - which adds negligible speed/phase shift cost to a loop closed around 30 MHz ft ring emitter BJT output Q
CFA still have to provide V gain - using the same class of Q as VFA VAS - which are still faster than output and driver Q in typical audio power amps
I'm not saying CFA is worthless - just that it doesn't have a big inherent advantage over VFA -CFA may be able to improve UGLF ~50% over typical VFA practice - less if VFA used Wurcer's AD797 folded cascode/bootstrapped "VAS"
and if CFA is single pole compensated while VFA practice today does understand 2-pole schemes fairly well then 2a 2-pole compensated VFA can easily beat CFA in loop gain to beyond 20 kHz
do you even know how how to add 2-pole compensation to CFA?
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.