Hi Jan,
Yes I understand that the error correction is a voltage sensing current correction circuit, but my point is that it cannot correct for reactive loads when the resultant current might be high at zero volts, or current might be zero at appreciable voltage amplitude.
Whereas Pavel's circuit has the potential to correct transconductance for load current at any voltage amplitude.
Cheers .......... Graham.
Yes I understand that the error correction is a voltage sensing current correction circuit, but my point is that it cannot correct for reactive loads when the resultant current might be high at zero volts, or current might be zero at appreciable voltage amplitude.
Whereas Pavel's circuit has the potential to correct transconductance for load current at any voltage amplitude.
Cheers .......... Graham.
Graham Maynard said:
I believe both circuits respond to output current, not voltage,
so they respond to reactive loads in much the same way as
resistive loads.
Graham Maynard said:
Graham,
I'm not sure about it, I will need to review Pavel's circuit when I am back from vacation (traveling in India). What the Hawksford thing does is to try to ensure that the output stage gain is exactly one. I can imagine that in this case the output load is being relected to the Vas output in some way, and distorts the Vas output signal, if you know what I mean. I have done some sims and indeed for the Hawksford to work as well as possible one should also use a buffer after the Vas to counter this effect, with a defined buffer output impedance to preserve the voltage-to-current mapping I mentioned before.
In my experience, the Hawksford correction conceptually works very well, but there are several secondary effects to take care of in practise. On the other hand, the simple implementation already decreases output stage non-linearity by anything from 20 to 40 dB, so it's worth while for a very moderate cost.
Jan Didden
I'm lost for words here, as you say Jan, does it sound better?
My view is that one of the worst aspects of amplification is capacitance induced propagation delay within the NFB loop. It is great that output stages can be made more linear, but there must be a point where returns diminish for the complexity necessary, and I 'sim' by reverse driving an output stage to see how stably and phase coherently it can respond w.r.t. input (in the presence of dynamic loudspeaker generated back EMFs) which is where global NFB as well as output stage continuity has more influence.
Cheers ....... Graham.
My view is that one of the worst aspects of amplification is capacitance induced propagation delay within the NFB loop. It is great that output stages can be made more linear, but there must be a point where returns diminish for the complexity necessary, and I 'sim' by reverse driving an output stage to see how stably and phase coherently it can respond w.r.t. input (in the presence of dynamic loudspeaker generated back EMFs) which is where global NFB as well as output stage continuity has more influence.
Cheers ....... Graham.
PMA said:
Whether they will admit it or not, most designers trust the
meter more than their ears.
Nelson Pass said:
Nice prejudice, indeed 😉 . I would not have awaited anything else.
I am just listenning to the finished amp. There is only one to compare - Pavel Dudek's DPA380 SSE version.
PMA said:
Pavel,
To fo a fair listening comparison, I suppose that the only difference between the 2 setups would be the "activation/deactivation" of the error correction circuit (I mean with same amp, music piece, speaker and everything the same).
Keep us posted !
Fab
To fab : We don't need to compare the same amp with or without error correction - with it sounds in EVERY time better and if you will listen it, you will not doubt. We are only searching for another amp which sound similar good 😉 , belive or not.
Upupa Epops said:
I agree that you can compare 2 different amps but how can you tell if the improvement is due to the error correction circuit itself and not the rest of the circuit of the new amp that is different than the one in the "reference amp"? I suppose also that the "rest" of the comparison set-up is the same.
BTW, I am still interested on getting the sound appreciation of the new amp!
Fab
Graham Maynard said:
No idea, never heard one in my life. It's one my list though...
Jan Didden
Graham Maynard said:
Graham,
I agree with that, but one of the advantages of the Hawksford correction is that it's not feedback as we usually assume it, i.e. with as much loop gain as we can get away with. That would give you the capacitive/propagation problems you mention. In the H- case, the 'feedback' is only with a loop gain of one (or two, I'm not sure how to compute it), around an output stage with a gain just below one. In a sim, it never showed even a trace of ringing. My educated guess is that the effects you mention, which may well be at the base of 'feedback sound' as you say, could be absent here.
I'm working on a Hawksford implementation taking care also of the secondary effects, so I hope to be able to report some practical results later.
Jan Didden
Hi Jan,
What a change it is not to be told that my thinking is flawed, or that it is me that is wired up, as has happened on some other diyAudio strings before.
I have not done any work on 'Error correction' so I am 'shooting from the hip' so to speak.
When reactive loudspeaker back EMF impinges a NFB loop controlled amplifier output node wrt input, the output terminal voltage reverse 'commutates' through a portion of class-B output transistor bias before the NFB regains control. Yes - 'feedback sound', and it affects first asymmetrical cycles, not the steady sinewaves that the simulator programmers expect their customers to run.
We need to watch the leading edge reaction of a reverse sine sim to observe this. Amplifiers that use 'Miller' connected VAS C.doms are particularly poor in this regard, thus some class-A designs can be affected too.
When the output stage has its own correction I could not begin to imagine what might happen due to both loops interacting wrt input without simulating this on a specific circuit.
Pavel's arrangement might be better in this regard, especially if used in a symmetrical circuit where there would not be any imbalances due to the leading + or - polarity of any toneburst, either input or output in origin.
Does your Hawksford implementation also use a Miller connected C.dom. Does it use a series output choke which would be even worse ??
Cheers ............. Graham.
What a change it is not to be told that my thinking is flawed, or that it is me that is wired up, as has happened on some other diyAudio strings before.
I have not done any work on 'Error correction' so I am 'shooting from the hip' so to speak.
When reactive loudspeaker back EMF impinges a NFB loop controlled amplifier output node wrt input, the output terminal voltage reverse 'commutates' through a portion of class-B output transistor bias before the NFB regains control. Yes - 'feedback sound', and it affects first asymmetrical cycles, not the steady sinewaves that the simulator programmers expect their customers to run.
We need to watch the leading edge reaction of a reverse sine sim to observe this. Amplifiers that use 'Miller' connected VAS C.doms are particularly poor in this regard, thus some class-A designs can be affected too.
When the output stage has its own correction I could not begin to imagine what might happen due to both loops interacting wrt input without simulating this on a specific circuit.
Pavel's arrangement might be better in this regard, especially if used in a symmetrical circuit where there would not be any imbalances due to the leading + or - polarity of any toneburst, either input or output in origin.
Does your Hawksford implementation also use a Miller connected C.dom. Does it use a series output choke which would be even worse ??
Cheers ............. Graham.
Graham Maynard said:
Nothing intrinsically wrong with an output choke....if implemented properly....
They are pretty usefull at attenuating RF before it gets into your feedback network....and...
....isolating the amp. from adversely capacitive loads.....
Graham Maynard said:
Graham,
I am currently working on the output stage, but have a design in mind for the Vas without Cdom (not a real Vas really), which I like to keep to myself for now.
On the series choke: If we can make the output stage such that it can stably handle anything down to say 2 Ohms, +/- 60 degrees of phase shift, I would hope we can lump the output choke (if needed), cable, xover and speakers in one single box called 'load' and be done with that. We probably wouldn't need the choke in this case anyway.
Jan Didden
janneman said:
Bad idea....the choke has to be connected as close to the feedback pick-off point as possible, if it's to do any good as far as attenuating RF spuriae is concerned......
Yeah sure, the 'box' was meant in a metaphysical way to lump everything after the output pin from a design/amp stability point of view. I thought that was self-evident. Sorry for the confusion Mike.
Jan Didden
Jan Didden
Graham Maynard said:
Yes, these "strings" have gotten much more interesting and informative lately.
Well, as a programmer that works at a company that makes simulators, I can honestly say that the simulator developers have to write their code such that the customer might use any and every combination of supported simulations with as few restrictions as possible. That's how the company they work for makes their money. They don't care about the results of anybody's simulation, except in the case when there might be an error caused by the simulator itself. Then it becomes top priority to fix. Audio mystics are of no interest to any of the simulator developers I know. Also, it's well known that SPICE has supported many other waveforms than sine waves in its transient analysis for over thirty years now.
As to the other items of yours that I've quoted, I'd be interested in seeing some actual technical justification for what you're saying. You complain about people challenging your assertions, yet I've not seen any rigorous technical explanation for a single one of them so far.
Graham Maynard said:
i am not averse to learning new stuff.....so.....can you show how negative feedback is a 'bad thing' with respect to back EMF?
Graham Maynard said:
This will tell you next to nothing......the turn-on gyrations in transient simulation are the result of circuit reactances being charged or discharged....usually invisible in real circuits...
andy_c said:
Yes...
- Status
- Not open for further replies.