Just a last minute post before going on holidays (camping in the Dordogne in France)
An example of the circuit Janneman proposes is here:
http://www.wasedaaudio.jp/English/Documents/AudioLab/2002SES-CC/1.html
The circuit uses a current conveyor for the input amplification and one for the error correction.
Steven
An example of the circuit Janneman proposes is here:
http://www.wasedaaudio.jp/English/Documents/AudioLab/2002SES-CC/1.html
The circuit uses a current conveyor for the input amplification and one for the error correction.
Steven
.
UE,
Are you referring to my diagram or the one Steven proposed? Please note that there is very little gain in this loop. The output stage is less than one, and the correction circuit has a loop gain of one (I think). But, as a kind of bootstrap it has some positive feedback properties. There was no sign of instability in my sims, but I accept that these were not representative for real circuits.
Jan Didden
Are you referring to my diagram or the one Steven proposed? Please note that there is very little gain in this loop. The output stage is less than one, and the correction circuit has a loop gain of one (I think). But, as a kind of bootstrap it has some positive feedback properties. There was no sign of instability in my sims, but I accept that these were not representative for real circuits.
Jan Didden
One interesting but somewhat complex approach I played with when experimenting with error correction was to use complementary diff amps for the V-I converter. This eliminated the funky cross-coupling of buffer input and output signals in Hawksford's resistor network design. It also keeps the output current of the stage that's driving it constant as the error correction current varies. The Halcro also eliminates this cross-coupling with a zener level-shifting network. I don't think the dual diff amp would work with a bipolar output stage though, as there won't be enough voltage drop from rail to rail of the diff amps.
.
Jan,
I am going to start a thread about my finished sample of the new error correction amp soon. I hope you will join with your results.
Cheers, Pavel
several photos:
http://web.telecom.cz/macura/pm_ab1.html
I am going to start a thread about my finished sample of the new error correction amp soon. I hope you will join with your results.
Cheers, Pavel
several photos:
http://web.telecom.cz/macura/pm_ab1.html
PMA said:
Cressy? 1346? One Cressy I know of is a RN warship sunk by a GE U-boat in the 1st WW ( http://www.firstworldwar.com/source/u9attacks.htm )
Or, are you referring to the battle at 'Merelveld' between the Turks and the Slavs in 1346 or thereabouts?
Or maybe this one; http://www.bartleby.com/35/1/108.html , but there was no Jan involved AFAIK. There was a Charles of Luxembourg, but by that time Luxembourg was not part of the Dutch kingdom (yet).
There was a John (Jan) of Bohemia involved, but I don't think he was 'my' Jan:
"It was Saturday, the 26th of August, 1346. On that memorable day, that picturesque old figure of the battle of Cressey, the blind and gray-headed King John of Bohemia, son of one monarch and father of another, was slain linked to two of his knights with their horses fastened to each other by the bridles and fighting most valiantly with his sword to his death. "
Jan Didden
Ouroboros,
Did you already have Spice results?
I like to comment on this:
- The error correction loop is a positive feedback loop (as it is the way Hawksford described it). Therefore, the stability depends on the neg feedback gain staying higher than the pos feedback gain. But that should be clear in the sim. If the output sticks to one of the rails, there trouble in paradise.
- The original Hawksford error correction avoided some stability problems by keeping the pos feedback loop gain exactly one. Your IOP1 has a gain of one, but the IOP2 doesn't. At least, that is what I assume, I don't know what that device is of course. So if you want to use the original Hawksford error correction you should bring the output of the IOP1 directly to the IOP2 input without resistors, and specify IOP2 as having a gain of one. (I can do that in the Proteus sim, don't know if you can).
Then you will see indeed a perfect gain of one amplification, independent of the load, independent of the active devices. (Assuming ideal error-less IOP's). Of course, changing to real-world opamps etc will cause it's own errors.
Jan Didden
Did you already have Spice results?
I like to comment on this:
- The error correction loop is a positive feedback loop (as it is the way Hawksford described it). Therefore, the stability depends on the neg feedback gain staying higher than the pos feedback gain. But that should be clear in the sim. If the output sticks to one of the rails, there trouble in paradise.
- The original Hawksford error correction avoided some stability problems by keeping the pos feedback loop gain exactly one. Your IOP1 has a gain of one, but the IOP2 doesn't. At least, that is what I assume, I don't know what that device is of course. So if you want to use the original Hawksford error correction you should bring the output of the IOP1 directly to the IOP2 input without resistors, and specify IOP2 as having a gain of one. (I can do that in the Proteus sim, don't know if you can).
Then you will see indeed a perfect gain of one amplification, independent of the load, independent of the active devices. (Assuming ideal error-less IOP's). Of course, changing to real-world opamps etc will cause it's own errors.
Jan Didden
Both op-amps are 'perfect' parts with infinite gain. I have results from 'TINA-pro' and I'll try it also in B2SPICE.
As it's so simple it's easy to put it into your version of SPICE to try it out.
I don't see any stability problems, but using real op-amps (OPA445 for the driving high-voltage part, and OP27 for the error amp) then the error-correction falls off quickly as the frequency increases.
By the way, the value of the 1k resistor (R8) feeding the output of the error amp into the summing junction is critical. If you do a distortion analysis with the correct value, and then vary the value slightly each side the distortion rises quite quickly. This is to be expected with error-correction I guess.
As it's so simple it's easy to put it into your version of SPICE to try it out.
I don't see any stability problems, but using real op-amps (OPA445 for the driving high-voltage part, and OP27 for the error amp) then the error-correction falls off quickly as the frequency increases.
By the way, the value of the 1k resistor (R8) feeding the output of the error amp into the summing junction is critical. If you do a distortion analysis with the correct value, and then vary the value slightly each side the distortion rises quite quickly. This is to be expected with error-correction I guess.
IOP1 (the diff amp) has a gain of 0.1 (10k and 100k resistors).
This matches the gain around IOP2 with the values of R2 and R8 as shown.
(A gain of 0.1 in the diff amp allows a standard +/-15V rail to power it) IOP2 has to be a high-voltage part of course.
I'm looking at this architecture to see if I can devise a power amplifier circuit to replace the amplifier boards in a couple of Quad 405 amps I own. I would like to keep the zero-bias output stage but replace the Quad 'current dumping' with an error correction version.
This matches the gain around IOP2 with the values of R2 and R8 as shown.
(A gain of 0.1 in the diff amp allows a standard +/-15V rail to power it) IOP2 has to be a high-voltage part of course.
I'm looking at this architecture to see if I can devise a power amplifier circuit to replace the amplifier boards in a couple of Quad 405 amps I own. I would like to keep the zero-bias output stage but replace the Quad 'current dumping' with an error correction version.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.