| GRollins |
Well, for one thing, a "non-inductive resistor" is hardly a good stand-in for most real world loudspeakers. You could start a prolonged argument as to whether the guy's logic is flawed on this point. I'm just noting this in passing.
You're losing power when you remove a pair of output devices. In the case of one pair, when you add a pair to act as proxies, you're going to load the power supply down; possibly with dire consequences.
I'd question some amplifiers' stability when faced with modifications to the output stage.
Assuming that you can get the thing to work (for this, I'd recommend designing an amp from the ground up, rather than modifying an existing one), I have no doubt that it sounds different...but does it sound better?
Sounds like sophistry to me.
Grey |
|
|
| jam |
A very interesting idea. He is actually compensating for non-linearities in the output stage (with feedback) but the reactive nature of the load will not affect the the feedback loop.
Might be a good, idea the only problem I see is the output stage (not in the feedback loop) has to be large to maintain a low output impedence.
This might make a good comparison to a circuit with nested feedback loops.
Jam |
|
|
| Ren Hoek |
His idea is not new. I bet your friend Nelson Pass knows this one, man.
Ren |
|
|
| Nelson Pass |
Actually I have not encountered this, although it is
an interesting and imaginative idea.
Of course it won't correct for things that a real life
load will do, but then again we aren't always looking
for that either. |
|
|
| Pete Fleming |
| I have also never seen this idea before and think it has great merit |
|
|
| janneman |
I looked at it, I wouldn't say interesting, rather weird. Look at the schematic under 'this is it':
He uses feedback to linearize the output of the block 10. Due to the nature of feedback, this will result in the input to '10' to be heavily distorted. Then he uses this heavily distorted node to feed the stage that drives the speaker. And claims it sounds much better....
As I said, weird. |
|
|
| GRollins |
I've been mulling this over and it might be interesting to pursue for drivers that present a purely resistive load, i.e. ribbons, planars, etc. I still don't like the idea of modifying an existing circuit, but a from-the-ground-up design might be interesting.
Hmmm...wonder if it's too late to stick this into the Aleph-X...
Grey |
|
|
| Nelson Pass |
I guess Altmann is more than a one idea guy. :)
It strikes me that the SPLIF amp will work better if you
can presume matching between the output stages, and
if you adjust the load of the fed-back stage to look like
the speaker load. |
|
|
| janneman |
Nelson,
It surely will work better IF the two output channels on the SPIF are matched, AND if the dummy load on the feedback-output stage completely and at all times tracks the characteristics of the speaker load at the other output stage. Because only then can you guarantee that the signal at both output nodes (the dummy load and the speaker) is identical at all times. But, if that is the case, you can just connect the two output nodes together, right? And if you do THAT, you may as well delete one of the output nodes and the dummy load, right? Guess what, it just looks like a normal feedback amp now.......
Janneman |
|
|
| Pete Fleming |
"Tube-o-lator" 'eh?
Groan! |
|
|
| tiroth |
Janneman,
The idea is to attempt to correct for the nonlinearities in the amplifier, not necessarily linearize the output (into a reactive load). As such, you really don't want a highly reactive load at the feedback node...I would guess you'd want a resistance of similar impedance, or a less reactive simulation of your target load.
Whether or not this is a good idea is another thing entirely. To me it seems as though this really is a no feedback design that happens to be extremely linear (into a resistive load). Most low-distortion, no feedback designs I am familiar with are stages driving high-impedance resistive loads--not the case with loudspeakers. I think you'd have to build it to really be able to say much about it as a concept. |
|
|
| tiroth |
| On a distantly related note, can you imagine paying $6000 for his DAC when the only pictures of it are renders? Scary... |
|
|
| janneman |
| quote: | Originally posted by tiroth
Janneman,
The idea is to attempt to correct for the nonlinearities in the amplifier, not necessarily linearize the output (into a reactive load). As such, you really don't want a highly reactive load at the feedback node...I would guess you'd want a resistance of similar impedance, or a less reactive simulation of your target load.
Whether or not this is a good idea is another thing entirely. To me it seems as though this really is a no feedback design that happens to be extremely linear (into a resistive load). Most low-distortion, no feedback designs I am familiar with are stages driving high-impedance resistive loads--not the case with loudspeakers. I think you'd have to build it to really be able to say much about it as a concept. |
Tiroth, Im not sure I fully understand you're saying. But this idea is wrong in principle. Either your goal is to get a distortion free signal at the load, and then you end up with simulating the load at the output stage where the feedback is attached, and 'two identical output stages as I described above, and that's a no-go.
Or your goal is to get a distortion free signal at the output of the output stage where the feedback is attached, loaded by a linear load or at least a load different from the speaker. In that case, you really have no idea what the signal at the speaker is doing. Again, a no-go. Sorry. |
|
|
| jam |
I think Trioth is tying to say that the designer is trying to get rid of non-linearities in the amplifier without resorting to global feedbak...kinda. He is getting rid of one distortion mechanism (non-linearity of the output stage ) not all of them.
Jam |
|
|
| tiroth |
Yup...as I read it, Altman's philosophy is "deliver (what would be) a linear signal to the load, it knows what to do with it"; the assumption being that whatever nonlinearities result from the difference between the actual and simulated load are less objectionable than the application of global NFB.
Please don't think I'm gung-ho on this idea, I just think that it is possible that it would produce acceptable quality, especially with a load with a large resistive component. |
|
|
| janneman |
Hi Jam & Tiroth,
Well, it's all very nice, but this guys makes money from it, that gives me the right to voice my criticism & opinion, isn't it?
You guys talk about 'he is trying to get rid of, sorta...', well, he may be trying, but isn't. The point is that having a linear or distortion free signal at the resistive load doesn't do diddly for the signal driving the speaker, UNLESS both output stages are identical (may be done for double the cost) and the speaker is resistive. Yes, probably there are some speakers with a large resistive component, in part of the freq range. But, be honest, knowing what you know, would you gamble on it with your present speakers? I wouldn't. But, of course, if I missed something, I'm sure Mr Altman will explain what it is.
Cheers,
Janneman |
|
|
| djk |
Consider the ESP 03,
http://sound.westhost.com/project03.htm
If we move the feedback connection to the center of the two 1N4000 diodes we can run the output stage open loop. Moving the 0R22 resistors up to the emitters takes care of the errors here. Funny, this now looks just like the amplifiers in the Nakamichi Stasis receivers. Those sounded very good too. Hmmm, didn't Nelson Pass have something to do with those? |
|
|
| mrfeedback |
Anybody have a pair of ESP 03 amps lying around, and willing to try djk's modification suggestion ?.
It looks as though the mod should not take too long (switchable) and I am sure we are all keen to hear the results.
Eric. |
|
|
| Nelson Pass |
Another thought strikes me: the output stage
which is used as the "dummy" stage could be a single
device while the speaker powering device could be
a number of parallel devices. This way, the "dummy"
output stage could replicate the real amplifier if it's
load is a multiple of the impedance of the speaker load.
In that case the "dummy" stage would not have to be as
large, saving energy and expense. |
|
|
| AKSA |
As usual, and with minimal words, Nelson goes straight to the nub of the matter.
Take it further; if the purpose is to buffer the feedback from a real world load, why not just take the feedback from the drivers? If we use a simple double emitter follower output stage, and we join the two driver emitters with a simple resistor (common practice, typically 220R), why not split the resistor and take the feedback for the amp from the midpoint of this resistor?
Then the output devices see the load, the feedback sees the crossover disjunction and at least partly corrects for it, and the phase anomalies of the output devices as they work into an extremely reactive speaker crossover are not seen by the global feedback network.
Of course, the load now suffers the intrinsic source impedance of the output devices, and this could be highish, largely depending on the size of emitter resistors, but if it is true that the imaging and fine sonic detail is lost in the phase distortions of crossover interaction with a conventional feedback network, then this little number should do the trick.
In truth, I have tried all this. The result: the amp sounds very twee, delicate and with a ravishing sound stage. It presents very like a 2A3 SET amp on steroids. But it lacks slam and impact, with a slightly wooly but musical bass, and really is only suitable for small jazz combos and chamber music. It was completely and utterly unsuited to rock 'n roll, and eventually I came to the conclusion that a combination of voltage amp and output stage feedback, in the ratio 1:2, was the optimum mix, but even then, the sound was still a little precious, at least to my battle-hardened ears.
Those interested in an even more detailed analysis than this, go to www.harmonicdiscord.com, and look at the thread entitled 'Tweak that works for every SS PP amp'.
Cheers,
Hugh R. Dean
www.printedelectronics.com |
|
|
| HarryHaller |
This idea of spliting the feedback has been used by several commercial designs. I have built circuits with class A mosfets outputs with feedback taken at the gates of the output stage mosfets and noticed some of the sonic attributes Mr. Dean has outlined. The first use that I am aware of of the dual feedback loop would be John Curl's JC3. Definitely an idea worthy of some experiments.
H.H.
http://marklev.com/marklev/JC3/jc3schematics.jpg |
|
|
| jam |
Ah Ha! Nested feedback loops an area I want to experiment more with.
Harry and Dean would you care to elarobate more on the designs you are familiar with and if you played with changing the raitos of those loops, or is it better not to nest the loops and have the voltage amplilier have its own loop and a seperate output stage with gain having its own loop.
Jam |
|
|
| Tube_Dude |
yes there are some comercial amplifiers that use the spliting feedback system. I remenber the Densen MA10 that is publicited
of a nom negativ feedback amplifier and i have found that it uses a normal voltage feadback satge and only the outputs followers
are out of the feedback loop...(it woud be advertised as a overal
feedback free amplifier..a la Mister Pass...)
I have also try that topology and my findings are the same...speakers d'ont like be left floating in the breeze...the sound
is very sweet but d'ont have guts...the bass became slow and
d'ont stop in time...specialy the timing of music became dirty... everything nice but not exciting...
Just my opinion
Regards
Jorge |
|
|
| Nelson Pass |
I have often taken feedback from the Gates of the
output stage, and it works well, but this would be
a whole other technique.
As to nested feedback, it works particularly well if
you have an IC op amp for a front end, as you can
set the open loop gain of that stage, and thus define
just how many decibels of feedback you want around
the output stage. |
|
|
|
