Scott's looks more like linearizing than canceling. Is there subtle difference?
It interesting that this same non linear functioning is also seen in mechanical valves.
Not this again.
There's cancellation via displacement current in the capacitors. jcx and PMA have shown a simplified sim several times. The voltage error across the output stage is "s" shaped the derivative gives you crossover "spikes" (the derivative because that is the displacement current in Ccomp that has to be supplied by the input stage when you close the loop). Cn sees only that same error voltage so it's displacement current is only the spike which is injected in the opposite phase into the current mirror so the input stage does not have to. At low to moderate frequencies Cn is bootstrapped by the output stage and technically does not appear in the signal path. We debated whether this was neutralization, positive feedback of -1, etc. At this point in time I'll stick with the pictures.If Cn goes to ground the displacement currents totally cancel and there is no compensation at all. A portion of Cn to ground cancels part of Ccomp and there is some effective decompensation, this was less useful than hoped. I'm sure there are some of those transitional compensation tricks that could be used with access to the internals.
Last edited:
Not this again.
If Cn goes to ground the displacement currents totally cancel and there is no compensation at all. A portion of Cn to ground cancels part of Ccomp and there is some effective decompensation, this was less useful than hoped. I'm sure there are some of those transitional compensation tricks that could be used with access to the internals.
Thanks Scott. You saved me from having to dig through thousands of post to find this.
It would be nice if there were place on the forum to put these things to separate them from the noise.
"We debated whether this was neutralization, positive feedback of -1, etc."
A rose is still a rose...
Cheers,
Last edited:
http://www.diyaudio.com/forums/attachments/solid-state/362091-bob-cordells-power-amplifier-book-cn-.jpg
But just to clarifiy my naive thoughts .. Scott, if this was a power amp, would you take Cn to CLEAN earth? ie the same one used by the feedback network.
I think its time AD promoted a DIP9 package .. one big enough for us oldies to solder and with an extra pin for stuff like this.
The extra pin can stick straight up like EF37A
I'll resist claiming this is the dual of my 'pure Cherry' mod
But just to clarifiy my naive thoughts .. Scott, if this was a power amp, would you take Cn to CLEAN earth? ie the same one used by the feedback network.
I think its time AD promoted a DIP9 package .. one big enough for us oldies to solder and with an extra pin for stuff like this.
The extra pin can stick straight up like EF37A
Last edited:
Not this again.
Hi
IIRC you once expressed an interest in a complementary version of the AD797 concept. Did you ever develop this any further?
Edmond's post reminded of this because his complementary cascode Has-anyone-seen-front-end-before seems to be in the same spirit.
Wondered if you overlooked the thread or just felt the similarity was not in the essentials?
Best wishes
David
For interpreting my own comments and use of the words.. linearizing would mean to reduce. Cancel is cancel or to null out. Zero. Just different techniques to get lower/lowest distortion.
Thx-RNMarsh
I guess I just missed that one (sorry Edmond a perfectly nice circuit), in fact the discrete op-amp thread had yet another folded cascode with recapture of the base current. These make a near perfect gm/Rl gain stage with low DC and AC distortions, very versitile. I think Bob Cordell used one in a phono article (without the base current recapture).
What I wanted to do was a complimentary discrete version with the output crossover error cancellation. This would add an output buffer rather than just an rload.
Richard, in normal operation the Cn samples the voltage across the output stage so its (Cn's) current has the correct (derivative of the voltage error) relationship. Remember to the first order an op-amp's input error signal is in quadrature with the output (d(cos) = sin) the current in Cn is also so 90+90 = 180 or in other words subtraction.
I think by now some can see my style is to point and illustrate concepts. I learned a long time ago that it really isnt good people skills to tell them what to do. If I do this well, I can influence others to connect the dots in thier own minds and often come up with something better than I might have thought of. or to get fresh ideas to brainstorm with.
Final point as this isnt my forum.... just an observer... a lot can be gained by using cancellation with linearization (like neg feedback). Especially when it simplifies the circuit and makes it easier to get to the goal.
My question to myself.. thinking out loud ... is what else can be done with FET/MOSFET's in the output stage of a power amp to reduce the distortion from thier highly non-linear C's via cancellation techniques.
Thx-RNMarsh
That comment was for the other Richard who wanted to know about Cn to ground. I leave the VAS/TIS or current feedback semantics alone.
and my comment was refering to myself and not your comments and explaination. I used your Cn example because it was a cancelling method. Thx-RNMarsh
Hi Richard,
This reminds me of the Hawksford Error Correction (HEC) that I used in my MOSFET power amplifier long ago. It can be looked at in numerous different ways, and that was discussed here at DIYaudio many times. It acts like a cancellation method, although it does so without a feedforward path. It has a place in the circuit where an adjustment pot can be placed to optimize the null, and it is interesting to watch the distortion residual as the pot is adjusted through the null.
Of course, the HEC can also be viewed as a negative feedback system wherein there is enclosed a positive feedback loop with a loop gain of +1, implying near infinite contribution to loop gain. Alas, as a feedback system viewed as such, it must be compensated in some way to maintain stability. The end result is that at very high frequencies the ability of the HEC to "cancel" distortion products becomes limited.
Of course, the bandwidth over which virtually any cancellation technique is fully effective is limited to some extent.
Cheers,
Bob
kgrlee said:
Du.uuh! Scott, you've confused my small brain even further. Are we using the same terms for Cn & Cc? In the above pic, Cn is from the output to the current mirror.
You haven't actually answered my question about whether Cc should be taken to CLEAN earth in a power amplifier .. rather than Vss as in AD797
Is the quadrature nature of the error signal cos most OPAs are sorta integrators?
(PS I've cheated in this post cos I erroneously referred to Cn rather than Cc in my original post
)IMO -
It would be great to see even more cancellation techniques included in your book.
Thx-RNMarsh
Hi Rick,
I believe the old analog scopes are full of this sort of thing. Since they couldn't use FB because of bandwidth limiting. Perhaps we should look there.
RF is another arena for canceling techniques. Lot's of predistortion and feed forward stuff.
Cheers,
Last edited:
That's a good idea for the Second Edition.
BTW, I just thought of another cancellation technique that is in the current book. It is the cascomp input stage, developed at Tek by an old friend of mine, Pat Quinn. Barrie Gilbert was his mentor at Tek.
I also show a JFET cascomp in my book. Even though the JFET is not an exponential device, the cascomp principles still allpy.
Cheers,
Bob
Du.uuh! Scott, you've confused my small brain even further. Are we using the same terms for Cn & Cc? In the above pic, Cn is from the output to the current mirror.
You haven't actually answered my question about whether Cc should be taken to CLEAN earth in a power amplifier .. rather than Vss as in AD797
Is the quadrature nature of the error signal cos most OPAs are sorta integrators?
(PS I've cheated in this post cos I erroneously referred to Cn rather than Cc in my original post
Sorry if I was confused, yes Cc goes to ground in a discrete amp. Funny, we have op-amps with a Kelvin connection at the output but no op-amps with grounds even though there are compensation schemes that would benefit not being referenced to a supply.
Bob,
I'm not enamoured with the feedback system approach in some cases, the vector subtraction approach gives the same degradation at high frequencies and makes more sense to me. The gain in the loop to actually cause instability can come from some parasitic process like the impedance into an emitter looking inductive at high frequencies.
Hi Scott,
Yes, I personally have always preferred the "error correction" view of looking at HEC, since it made some aspects of what was going on easier to see and understand. However, I always felt that the feedback system view of it made it easier to understand that it needed some form of compensation. Depending on what we call parasitics, I think its stability is more than a matter of just parasitics - otherwise I think we would be getting a bit of a free lunch.
Stanley Lipshitz also felt strongly that HEC should be viewed as a feedback system.
I have not tried it, but I expect that if we implement a HEC circuit at the idealized block level, where the various blocks had reasonable, sometimes multi-pole rolloffs, and simulated it, we would see evidence of instability if it was not compensated in some way.
One think that I like in the arrangement that I use is that the main form of compensation actually is somewhat of a shunt that feeds the signal forward in the main signal path, while at the same time reducing the correction loop action at high frequencies.
Cheers,
Bob
Scott, I have often wondered why op-amps with a ground pin are not more common (in fact as far as I know there's only ever been one integrated op-amp that had a ground pin). As an industry insider I would appreciate your opinion on this. Single internally-compensated op-amps with no offset trimming have three "spare" pins in an 8 pin package so I don't really see what the issue is. Having a ground pin means that the relationship between PSRR and CMRR in "traditional" op amps can be broken.