Augmented Feedback Error Correction

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
in a quick skim it looks like "active feedback"

so it is equivalent to adding loop gain - which is sharply limited by the main amp stability limits

if the main amp already uses "maximun feedback" - you really can't improve it this way without degrading stability
 
www.hifisonix.com
Joined 2003
Paid Member
in a quick skim it looks like "active feedback"

so it is equivalent to adding loop gain - which is sharply limited by the main amp stability limits

if the main amp already uses "maximun feedback" - you really can't improve it this way without degrading stability

Yes, the main amp has to be stable for it to work. I would not apply it as a cure all.
 
www.hifisonix.com
Joined 2003
Paid Member
in a quick skim it looks like "active feedback"

so it is equivalent to adding loop gain - which is sharply limited by the main amp stability limits

if the main amp already uses "maximun feedback" - you really can't improve it this way without degrading stability

The loop gain of the design shown in the link is 40 dB at LF and about 25 dB at 20 kHz. The phase margin is c. 60 degrees at th ULG of 1 MHz. AFEC looks like a useful technique for CFA power amp topologies where the loop gain lends to be lower than VFA.

Let me also stress that this is no designed to fix a crap, marginally stable amp It will take a good solid design to low double digit ppm distortion levels at HF at power levels up to the onset of clipping (8 Ohms).
 
Andrew,

This approach by a US designer from an Electronics magazine I see in the eighties. An additional IC to apply gnfb to the main aim, reported to have very low distortion, around 20dB than the conventional design.

I did something with a tube in the nfb loop twenty years. It had extremely organic sound, very like a tube amp, and wonderful imaging. But there were issues with tube turn on, only solved with a relay which activated when the filament pulled in. Distortion was not that low, but the profile was dominated by the tube. I used a SRPP 12AU7 with ground grid.

Cheers,

Hugh
 
www.hifisonix.com
Joined 2003
Paid Member
Andrew,

This approach by a US designer from an Electronics magazine I see in the eighties. An additional IC to apply gnfb to the main aim, reported to have very low distortion, around 20dB than the conventional design.

I did something with a tube in the nfb loop twenty years. It had extremely organic sound, very like a tube amp, and wonderful imaging. But there were issues with tube turn on, only solved with a relay which activated when the filament pulled in. Distortion was not that low, but the profile was dominated by the tube. I used a SRPP 12AU7 with ground grid.

Cheers,

Hugh

Nice! Thanks for te feedback Hugh. Any chance you will try it again? Nothing new under the sun as they say!

I am going to implement this practically on a class A amp. If you take a look at my sx-Amp, it uses a straight EF follower output and this technique will be great to improve the distortion performance on a future iteration using the sx-Amp topology.
 
www.hifisonix.com
Joined 2003
Paid Member
Any reason why this setup would be or is superior to just simply increasing vas gain by using darlington, cfp, super pair in vas. Its simpler.

No, i do not claim it is superior to other methods. Keep in mind it is not much more complex than your average DC servo circuit.

I think the biggest issue this approach deals with is the output stage non linearity when driving heavy loads. You will need a lot of additional feedback (loop gain) to get to c. 5 ppm at mid power with just a simple EF which is what this setup promises. The loop gain on th demo amp is only 25 dB at 20 kHz. You also get a big improvement in PSRR and DC offset nulling as by products.

:)
 
Disabled Account
Joined 2006
Getting additional loop gain is not difficult to obtain but 5ppm with just a EF is. With feedforward techniques like the national CFB opamps the performance could be surpassed but then again it does become more complicated than just a opamp which in my experience should always be employed in this type of design to deal with DC offsets. Interesting concept.
 
So as to not pollute another thread and out of respect I will post my findings of tuning the AFEC circuit using the group delay here.

To start with, I took an AC analysis of the main amp. Then looked at group delay. Then tuned the AFEC gain resistor and compensation cap to create a group delay profile that closely matched the original amplifier profile (wihout AFEC). This new profile was close and did not have any major peaks just some ripple. End result is pretty good square waves as expected as the main amp had pretty clean square waves. The gain resistor is 150R and compensation cap = 50pF. Gain margin and phase margin unchanged.
 
Last edited:
Am I right by saying that it's basically an error voltage booster? If I'm getting it right it is - the result is that the amp acts faster on the error signal, provided it has the bandwidth left to do so which is presumably the case on lower volumes.

I'll add it to my VAS in the sim and see if I can discern an even further improvement in performance :)
 
Last edited:
That's the basic understanding I had. It seems to be a loop gain booster which is what you are describing.

The results can be impressive. It can be developed further. Eg. My favourite tweak (discovered by playing and not based on any scientific reasoning,btw) is a cap+resistor across its feedback resistor to deal with square wave peaking.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.