I understand what you are arguing but you are missing a vital point. Your new op-amp circuit is now the front-end. You have the original front end too. The op-amp front end is doing the distortion reduction using global NFB. Therefore this circuit relies on global NFB. It doesn't matter that the original front end is not inside the loop, as such.
The original front-end is still contributing to the signal at the OPS input. And it will be directly affecting the GNFB excess loop gain. In this sense it is also part of the loop.
That doesn't mean it is a bad idea, though.
I'm right behind Hugh's encouraging sentiment that you should push forward with this interesting configuration and see whether you can make a better amp. I think it is a a good idea in theory if you can get better IPS performance by using op-amps with a piggy-backed conventional input stage to deal with the voltage extremes. I've not come across this before.
I was inspired by your thread title because I think trying to make a high performance amp without NFB is a great goal and will probably reveal what a truly musical amp sounds like, even if it has modest THD. If you go straight in to a relatively complicated GNFB design with op-amps you may miss this revelation.
The original front-end is still contributing to the signal at the OPS input. And it will be directly affecting the GNFB excess loop gain. In this sense it is also part of the loop.
Unusual to have two, parallel input stages.
Ok, so you are using the original IPS as a sort of voltage shunt, albeit with high impedance so the opamp can add its signal.
I get that you really want to be able to boast this but it is false. Sorry.
That doesn't mean it is a bad idea, though.
I'm right behind Hugh's encouraging sentiment that you should push forward with this interesting configuration and see whether you can make a better amp. I think it is a a good idea in theory if you can get better IPS performance by using op-amps with a piggy-backed conventional input stage to deal with the voltage extremes. I've not come across this before.
I was inspired by your thread title because I think trying to make a high performance amp without NFB is a great goal and will probably reveal what a truly musical amp sounds like, even if it has modest THD. If you go straight in to a relatively complicated GNFB design with op-amps you may miss this revelation.
Last edited:
Member
Joined 2009
Paid Member
I'm right behind Hugh's encouraging sentiment that you should push forward with this interesting configuration and see whether you can make a better amp. I think it is a a good idea in theory if you can get better IPS performance by using op-amps with a piggy-backed conventional input stage to deal with the voltage extremes. I've not come across this before.
I was inspired by your thread title because I think trying to make a high performance amp without NFB is a great goal and will probably reveal what a truly musical amp sounds like, even if it has modest THD. If you go straight in to a relatively complicated GNFB design with op-amps you may miss this revelation.
I can only strongly support this. Years ago when I developed my paX amp with roughly similar ideas, it took a real effort to let go of the 'must have super low THD' adage.
The result was an amp that was widely commented to sound very natural and enjoyable. I must have had the right harmonics in the right ratio;-)
Jan
This is important and this is where I disagree
Opamps configuration does not replace the "original" front-end.
See the what happens, for example, if we drive 1KHz @ 100W into 8 ohm.
Voltage swing at the output of original front-end is 28.4V RMS, driving 5.1K load (resistor).
Voltage swing at the 2-nd opamp's output is 360mV RMS (error signal) in this case.
At 20KHz error signal swing will be higher, but it will not exceed 1.1V RMS.
Does the error correction circuit replace the original front-end? No. It handles error signal only. Original front-end provides full swing signal for driving the OPS. Opamps are not capable of doing it.
Now - if I introduce any simple changes to the original front-end - limit the bandwidth, adjust the voltage gain - those changes will be immediately reflected in the amplifier's characteristics (bandwidth, gain - accordingly).
If I introduce this kind of changes in the error correction circuit - nothing noticeable will happen except the amount of error correction. If I shunt the 2-nd opamp's output to ground - the amplifier will still work, producing higher distortion, at the level of "no error correction".
One more consideration. "Distorted" signal is taken from the output, ODNF correction signal is injected at the OPS input. That's the loop. The fact that I take the reference signal from the input (as the point with the cleanest signal) does not change anything.
I can easily arrange the same ODNF configuration, taking the reference from the original front-end's output (still very low distortion) and injecting it inside the OPS - I have already mentioned this kind of approach. It does not change anything. Error correction works the same way, being completely local to the OPS in this case.
Hugh, thank you for the warm feedback
The idea about the tube error processing circuit is brilliant - I will think about it "up"
Brian - sorry, mistyped your name in a few previous posts - I see what you mean.
Having input as a reference, ODNF corrects all the error, including the one, generated by the front-end. It corrects the error of the whole amplifier.
Having the front end output as a reference, we have the front-end out of the game completely, so only the OPS error is corrected.
So my statement "It does not change anything. Error correction works the same way, being completely local to the OPS in this case." is not exactly correct. The result with reference at the front-end's output will be close in case the front-end is rather low-distortion - like Eenbahnstark in post #1 or Diamondstark in post #20.
Otherwise, we will see the front-end's distortion "adding" to the overall picture - not really seeing it in case of the reference at the input.
Anyway, I like the approach with a separate "error" channel - it simplifies compensation approaches and lacks many traditional feedback issues, at least from what I can see now.
OK - good Clear
I still like the idea of involving the tubes - probably both in the front-end and the error channel - correcting mostly (or only) the OPS and letting the front-end's "character" dominating the overall amplifier's behavior.
Having input as a reference, ODNF corrects all the error, including the one, generated by the front-end. It corrects the error of the whole amplifier.
Having the front end output as a reference, we have the front-end out of the game completely, so only the OPS error is corrected.
So my statement "It does not change anything. Error correction works the same way, being completely local to the OPS in this case." is not exactly correct. The result with reference at the front-end's output will be close in case the front-end is rather low-distortion - like Eenbahnstark in post #1 or Diamondstark in post #20.
Otherwise, we will see the front-end's distortion "adding" to the overall picture - not really seeing it in case of the reference at the input.
Anyway, I like the approach with a separate "error" channel - it simplifies compensation approaches and lacks many traditional feedback issues, at least from what I can see now.
OK - good
Clear I still like the idea of involving the tubes - probably both in the front-end and the error channel - correcting mostly (or only) the OPS and letting the front-end's "character" dominating the overall amplifier's behavior.
Valery i have listening expiriences with amps without GNFB.That are Gryphon audio amps and this are best sounding amps i ewer heard at home. I think D'Agostino and perhaps Ayre (bouth without GNFB)are on the saim level-but newer heard at home.
I have expirience with Tabu Century(=NAD S300-built by Gryphon for NAD),Callisto 2100 and Antileon+Pandora wich plays in an other league and is my benchmark for amps(owner a frend).
I hawe made some mods on an NAD S300(audio section complite saim as Tabu Century) two years ago with good sonic results(i aded a small amont of local feedback from pre-drivers and a larger idlling current<first watt>)-schematics on Hifi Engine.
Cheers,Bozo
I have expirience with Tabu Century(=NAD S300-built by Gryphon for NAD),Callisto 2100 and Antileon+Pandora wich plays in an other league and is my benchmark for amps(owner a frend).
I hawe made some mods on an NAD S300(audio section complite saim as Tabu Century) two years ago with good sonic results(i aded a small amont of local feedback from pre-drivers and a larger idlling current<first watt>)-schematics on Hifi Engine.
Cheers,Bozo
Valery, this might be of interest. It combines your desire for no GNFB with Hugh's like of tubes:
Lamm M1.2
Also, designed by a Russian.
I have deduced the following about this monoblock: it uses a tube 2nd stage (voltage amplifier stage). 12 MOSFETs running at a little over 0.4A bias each (8-ohm mode), with 1 ohm source resistors.
The Lamm has to use 12 MOSFETs, heavily biased, just to compensate for their poor transconductance. A lot of power and aluminum! Using BJTs you could, in theory, achieve the same with just 2 devices biased at about 150mA.
It sounds very good with classical music. I've not heard it attempt rock so I can't comment. It's curious to me why monster, multi-transistor amps like this invariably demo using classical or simple acoustic music...or maybe it's not so curious.
Lamm M1.2
Also, designed by a Russian.
I have deduced the following about this monoblock: it uses a tube 2nd stage (voltage amplifier stage). 12 MOSFETs running at a little over 0.4A bias each (8-ohm mode), with 1 ohm source resistors.
The Lamm has to use 12 MOSFETs, heavily biased, just to compensate for their poor transconductance. A lot of power and aluminum! Using BJTs you could, in theory, achieve the same with just 2 devices biased at about 150mA.
It sounds very good with classical music. I've not heard it attempt rock so I can't comment. It's curious to me why monster, multi-transistor amps like this invariably demo using classical or simple acoustic music...or maybe it's not so curious.
Hi Bozo, thank you for sharing your listening experience - I have also auditioned Gryphon Mephisto and liked it a lot.
I also enjoyed studying the topology of S300 - I have it in my manuals collection for for a long time, but now I looked at it under a different angle.
I have also seen one schematic from Ayre - I mean, all of them confirm the overall concept
Designs are relatively simple, short signal paths, good bulky power supplies, solid wiring and they don't care about distortion numbers, that are relatively high - those devices sound great anyway
Couple of years ago, I have tested the following configuration - 2 x SRPP stages, based on 12AU7, overall gain 29db, plus Slewmaster OPS, modified for IRFP290/9290 HexFETs (5 pairs @ 80mA each - class AB).
Front-end produced up to 0.5% THD with mostly 2-d harmonic , less 3-rd, all the rest - insignificant. OPS showed around 0.08% with definitely wider spectrum, but still good profile.
It sounded very nice with jazz trios
I started to believe that good sound from the amp comes from good power supply, good grounding layout in the enclosure, and low dist circuitry is the latest less important factor (including crossover dist)
I started to believe that good sound from the amp comes from good power supply, good grounding layout in the enclosure, and low dist circuitry is the latest less important factor (including crossover dist)
Well, distortion nature and profile are also important, plus - I like the tallest distortion components (preferably 2, 3) being at least below -80db.
However - I agree - without those qualities in the first place (good power supply, good grounding layout in the enclosure, proper wiring with "beefy" high-current connections) all the circuit advantages may be spoiled very easily
Hi Valery,
I ran across this rather clever implementation of a headphone amp with a THAT1646 driver. It appears to be zero global feedback.
The usual "inputs" are used to set the bias current. Audio comes in through the common mode inputs.
Hi X,
THAT1646 has got internal common mode feedback, but anyway - a rather clever circuit
Hi Brian,
Sure, I know DarTZeel. Interesting OPS with 2 diodes at the bases of output devices. There is a number of successful designs with feedback loop around the front-end and open loop OPS. Well, the loop around the voltage gain stages is not that critical, although I prefer current drive and local loops there as well.
Interesting good examples - Nakamichi PA5/PA7. Those ones have got CFP arrangement in their OPS, providing internal feedback, however it's local to OPS. So, there are 2 loops - one around the voltage gain stage, the other one in CFP OPS. I have simulated this kind of OPS with BJT "drivers" and HexFET output devices - looks very promising, even better than BJT + BJT pair.
Hi X,
THAT1646 has got internal common mode feedback, but anyway - a rather clever circuit
I wonder why Wayne Kirkwood called it "open loop Class A"? Reading the data sheet, it indeed shows the differential input has built in feedback. However, that's being used to purely set the bias current on the outputs. The common mode input, I believe does not have internal feedback network - and would normally be connected to the outputs to provide global feedback - but in this case, is used to effect the audio signal drive. The output then drives the output stage and this still does not have feedback into the main input block. Could you take one more look and note how pins are connected relative to equivalent circuit in Fig. 2?
Attachments
Last edited: