Bob Cordell Interview: Error Correction

Jump to Latest
#2,061
#2,062
G.Kleinschmidt said:
Now here�s an idea � why not design an amp with EC applied around the entire thing instead of conventional global NFB?
Instead of designing the amp with heaps of low frequency open loop gain and then rolling it off with distortion causing frequency compensation, design the amp as though you were making a zero-GNFB amplifier, with a flat, uncompensated open loop gain � say +26dB.
Then apply frequency compensated EC around the whole thing - just tweak the EC circuit gains to compensate for the forward gain of +26dB.
Click to expand...


All in all you will need frequency compensation, probably more complicated and non-intuitional compared to negative feedback loop.
 
#2,063
darkfenriz said:



All in all you will need frequency compensation, probably more complicated and non-intuitional compared to negative feedback loop.
Click to expand...


I think you'd only need to roll-off the EC loop gain above a certain frequency to ensure stability, just as is required when applying EC to the output stage only.
In fact, you'd want to ensure that the actual amplifier itself has as wide a bandwidth and as little phase shift as possible.

Cheers,
Glen
 
#2,064
#2,066
G.Kleinschmidt said:
Now here’s an idea – why not design an amp with EC applied around the entire thing instead of conventional global NFB?
Instead of designing the amp with heaps of low frequency open loop gain and then rolling it off with distortion causing frequency compensation, design the amp as though you were making a zero-GNFB amplifier, with a flat, uncompensated open loop gain – say +26dB.
Then apply frequency compensated EC around the whole thing - just tweak the EC circuit gains to compensate for the forward gain of +26dB.

Cheers,
Glen
Click to expand...


Don't steal my idea! 😉 . Instead, come to SF in Oct to the Burning Amp festival and you can listen to the thing!

Jan Didden
 
#2,067
EC applied around the entire thing

Originally posted by G.Kleinschmidt
Now here�s an idea � why not design an amp with EC applied around the entire thing instead of conventional global NFB?
Click to expand...

darkfenriz said:
All in all you will need frequency compensation, probably more complicated and non-intuitional compared to negative feedback loop.
Click to expand...


Hi Adam,

Why don't you tell him a bit more about your "little gem" ?

Cheers, Edmond.
 
#2,069
Re: EC applied around the entire thing

janneman said:



Don't steal my idea! 😉 . Instead, come to SF in Oct to the Burning Amp festival and you can listen to the thing!

Jan Didden
Click to expand...


Sorry! :innocent:
When I first opened your pax pdf, the circuit looked at first glance (before zooming in) like conventional opamp-front-end power amp, with EC applied around the whole thing.


estuart said:

Hi Adam,

Why don't you tell him a bit more about your "little gem" ?

Cheers, Edmond.
Click to expand...


???


forr said:
Originally posted by G.Kleinschmidt
---Now here’s an idea – why not design an amp with EC applied around the entire thing instead of conventional global NFB?---

Is this a low power implementation ?
http://www.diyaudio.com/forums/attachment.php?s=&postid=1183839&stamp=1176634068
Click to expand...


Not exactly the same thing - it looks like an attempt at error correction applied to a frequency compensated amplifier with global NFB.

Cheers,
Glen
 
#2,072
G.Kleinschmidt said:



I think you'd only need to roll-off the EC loop gain above a certain frequency to ensure stability, just as is required when applying EC to the output stage only.
In fact, you'd want to ensure that the actual amplifier itself has as wide a bandwidth and as little phase shift as possible.

Cheers,
Glen
Click to expand...


Hi Glen,

Although EC applied globally around an entire amplifier may indeed work in principle, my belief and experience says that EC works as well as it does because it is around a very fast group of components. Speed is really king in the way EC seems to work, and its ability to correct accurately at high frequencies (where it is needed the most) is very dependent on speed and phase.


Cheers,
Bob
 
#2,073
Bob Cordell said:



Hi Glen,

Although EC applied globally around an entire amplifier may indeed work in principle, my belief and experience says that EC works as well as it does because it is around a very fast group of components. Speed is really king in the way EC seems to work, and its ability to correct accurately at high frequencies (where it is needed the most) is very dependent on speed and phase.


Cheers,
Bob
Click to expand...


Hi Bob.


I agree, hence my comments on the bandwidth and phase shift of the voltage amplification part of the amplifier.
I doubt that such an approach could be demonstrated to be generally more effective than conventional global NFB, but I think a decent performing amplifier could be built this way, on the condition that particular attention is paid to open loop linearity.

Cheers,
Glen
 
#2,074
Speed is really king in the way EC seems to work, and its ability to correct accurately at high frequencies (where it is needed the most) is very dependent on speed and phase.
Click to expand...
Hi, Mr. Cordell,

I'm very interested in this. Actually, I'm thinking alot about an EC based on floating opamp like this one :
http://www.diyaudio.com/forums/showthread.php?postid=1230139#post1230139
or this one :
http://www.diyaudio.com/forums/showthread.php?postid=1271976#post1271976
Because (before I read your post above), the OL gain is the important factor in how well an EC will do its work. And opamp provides very high gain.
Now comes a factor of "speed and phase". Opamp based cannot beat simple discrete transistors in these terms (like Hawksford EC).
What is your opinion on opamp based EC (floating +/-15V). Will it work better or worse than discrete EC?
 
#2,075
lumanauw said:

Hi, Mr. Cordell,

I'm very interested in this. Actually, I'm thinking alot about an EC based on floating opamp like this one :
http://www.diyaudio.com/forums/showthread.php?postid=1230139#post1230139
or this one :
http://www.diyaudio.com/forums/showthread.php?postid=1271976#post1271976
Because (before I read your post above), the OL gain is the important factor in how well an EC will do its work. And opamp provides very high gain.
Now comes a factor of "speed and phase". Opamp based cannot beat simple discrete transistors in these terms (like Hawksford EC).
What is your opinion on opamp based EC (floating +/-15V). Will it work better or worse than discrete EC?
Click to expand...


I've never tried any kind of op-amp-based EC. I don't think it saves much complexity at the end of the day and I believe that in most cases it will compromise speed that the EC loop can work with. Discrete-based EC can work remarkably well, and is really not that complex. It adds only about four or six small-signal transistors to what you need to have there anyway. It is particularly effective with vertical MOSFETs because of their lightning speed, but also appears to work well with RET bipolars.

Cheers,
Bob
 
#2,076
Hi, Glen,

About your schematic in here : http://users.picknowl.com.au/~glenk/CLASSA.HTM
I got some things that I don't understand.
1. Why is the anode of D1 (and katode of D? = 1N4148) are tied together to R1/input? If it is part of current limiter, maybe connecting those 1N4148 to collectors of Q29 and Q30?
2. After input R1 (750ohm), the signals are level shifted via VR (LM336B-5V). Is this makes the base of Q1 sitting at +5V (and base of Q4 sitting at -5V), or these 2 LM336B-5V at input are just voltage limiter?
 
#2,077
lumanauw said:
Hi, Glen,

About your schematic in here : http://users.picknowl.com.au/~glenk/CLASSA.HTM
I got some things that I don't understand.
1. Why is the anode of D1 (and katode of D? = 1N4148) are tied together to R1/input? If it is part of current limiter, maybe connecting those 1N4148 to collectors of Q29 and Q30?
2. After input R1 (750ohm), the signals are level shifted via VR (LM336B-5V). Is this makes the base of Q1 sitting at +5V (and base of Q4 sitting at -5V), or these 2 LM336B-5V at input are just voltage limiter?
Click to expand...


G'day

The two LM336-5's just bias the first emitter follower (Q1-Q3, Q4-Q6). The two 1N4148's from the collectors of the I-limit transistors connect to R1, because it is from here that the "error voltage" is sensed by the EC. Clampling the signal (with respect to the output) here stops the EC transistors from conducting harder when the current limiter operates.

Cheers,
Glen
 
You must log in or register to reply here.
Top Bottom