I see. The current source will then be only limited by the max output current. 20W seems very optimistic but maybe worth a try. If noone shoots the idea for some reason i'd be willing to give it a try. None of my GCs are heatsinked properly, not to mention the extra heat from the CCS but that should not be such a problem.
Thanks for the idea.
Thanks for the idea.
Steve Eddy said:
Last year, there was a lot of noise on Chip Amp forum about using resistor between negative rail and output. I tried 1k5 resistor but it degraded the sound, so I never bothered with using active CCS.
analog_sa said:
Right. Far as I'm aware, the output stage should be have like a single-ended emitter follower with an active current source. At least within the limit set by the amount of current you're shoving through it.
Well, if you're using a +/- 20 volt supply and if at worst case you swing just 15 volts, that's 14 watts, which is just a mere 1.5dB less than 20 watts.
Yeah, I might give it a go myself. Though I'll have to grab some 3875s. All I've here are some Apex hybrids and I ain't gonna do that kind of experimenting on a $60 device. 🙂
Sure. Though don't give too much credit here. The idea already existed. It's just a matter of scaling it up a few notches.
However if it does pan out, I claim all rights to the name ChipSETTM. 😀
se
Steve Eddy said:
Ah'll be dam! A race to name the next fashion trend! Holy magic smoke! May the current soorce be with you!
MBK said:
I think that was what I was trying to get at. Even 0.01% THD is pretty damn low, certainly low enough that simply dismissing it as a "toy" is unjustified.
In fact, the arguments here are running along precisely the same lines as with any solid-state amp, which (to me) shows that the LM3875 and its friends really have reached the big time.
Cheers
IH
Steve Eddy said:
Some head amp designs use this method like the META 42 but on low power op-amps.
for once I agree with Peter...
I don't think the LM-whatever chip would take kindly to the sustained levels of (asymmetric) Iq required for any substantial class A performance. Which means the 'lower' half of the OPS is gonna come switching on in perfect class B at some point... which might explain the sound degradation Peter speaks of.
If the insistence on chips is to remain, why not rather whack the output into a high input-impedance class A follower ie complementary MOSFET source followers, and then include the latter in the overall feedback loop. The LM o/p stage should then remain in class A, while the discrete devices can truly be biased in full PP class A...?
But that's just one heretic's opinion... anybody tried this?
I don't think the LM-whatever chip would take kindly to the sustained levels of (asymmetric) Iq required for any substantial class A performance. Which means the 'lower' half of the OPS is gonna come switching on in perfect class B at some point... which might explain the sound degradation Peter speaks of.
If the insistence on chips is to remain, why not rather whack the output into a high input-impedance class A follower ie complementary MOSFET source followers, and then include the latter in the overall feedback loop. The LM o/p stage should then remain in class A, while the discrete devices can truly be biased in full PP class A...?
But that's just one heretic's opinion... anybody tried this?
Peter tried a very low current source - about 15mA which some seem to like in this application, but which does absolutely nothing to get the output in class A. The issue at hand is to have about 3A running through one of the outputs and i really see no reasons for unkind attitude. Provided the output stays within the 3 amps there will be no transition to class B.
Of course it's easy to boost the output of a chip amp with an external transistor pair but the advantage of a short signal path will be lost.
Re: for once I agree with Peter...
you could also use a classical (low power) opamp with a mosfet follower. No need for a high power chip
DrG said:
you could also use a classical (low power) opamp with a mosfet follower. No need for a high power chip
Let's forget GCs for a while.
Let's talk generically about chip amps.
There are many commercial amps that have power op-amps inside, some of them high-end products.
Jeff Rowland uses them, Linn uses them too...
DrG, have you ever heard a Linn LK140?
Maby not.
Even then, you may not know, but I tell you.
It has two TDA7293 power op-amps per channel, in master-slave configuration, basically the same as you can find on the datasheet.
And it's a very good amp.
And it drives almost any speaker.
So, there are plenty of ways to make a chip amp.
But please, don't generalize.
And you never heard a GC, I bet.
Let's talk generically about chip amps.
There are many commercial amps that have power op-amps inside, some of them high-end products.
Jeff Rowland uses them, Linn uses them too...
DrG, have you ever heard a Linn LK140?
Maby not.
Even then, you may not know, but I tell you.
It has two TDA7293 power op-amps per channel, in master-slave configuration, basically the same as you can find on the datasheet.
And it's a very good amp.
And it drives almost any speaker.
So, there are plenty of ways to make a chip amp.
But please, don't generalize.
And you never heard a GC, I bet.
Not a very good example Carlos. Many, like myself, are of the opinion that Linn simply can't make a good amp; no matter what technology they use. The Jeff Rowland with chips also does not compare well to the JR without chips.
Linn and Jeff...
Dear Carlos FM...
Linn and Jeff are old friends of mine... and I agree with analog_SA, the models in question don't sound great (my opinion, no reflection on yours). Which means that whatever they contain or are based on is irrelevant (to me)...
And don't bet... I've listened to chip amps, including the popular favorites mentioned here. And my impression is not as favorable as the rest of this community's seems to be. Neither do our opinions iro high-end design concur. But feel free to do whatever makes you smile...
Dear Carlos FM...
Linn and Jeff are old friends of mine... and I agree with analog_SA, the models in question don't sound great (my opinion, no reflection on yours). Which means that whatever they contain or are based on is irrelevant (to me)...
And don't bet... I've listened to chip amps, including the popular favorites mentioned here. And my impression is not as favorable as the rest of this community's seems to be. Neither do our opinions iro high-end design concur. But feel free to do whatever makes you smile...
Of course a good GC sounds even better.
Believe me.
What bet is that you never heard a GC (with LM3875).
THIS is high end.
I've just pick other examples not to talk only of GCs, as there are other good sounding chip amps out there.
My oppinion, of course.
Believe me.
What bet is that you never heard a GC (with LM3875).
THIS is high end.
I've just pick other examples not to talk only of GCs, as there are other good sounding chip amps out there.
My oppinion, of course.
protos said:
Yes. It's a technique that's been used for quite a few years. I first came across it back in the 80s if memory serves. Just that I don't recall ever seeing it used with power opamps.
se
Plenty of this op amp current biasing on headwize.com.
The power follower of a signal op amp sounds like a more elegant approach to me, the best of both worlds so to speak. PMA has some designs on his personal pages, some on Rod Elliot's pages as well...
The power follower of a signal op amp sounds like a more elegant approach to me, the best of both worlds so to speak. PMA has some designs on his personal pages, some on Rod Elliot's pages as well...
short circuit, anyone...?
If you say so, Carlos... but then something sounding "better" than an amp that doesn't sound "good" (assumed concensus ex you) isn't saying very much at all...
This is true. But my idea was to keep the specific GC chips in the picture, overkill ratio notwithstanding. So how about compromising and "using" the GC's output by means of an inverting tranny buffer? Just a thought...
I am unsure how much of an advantage that is, considering the collective metres of interlink and speaker cable common to most systems... and compared to the hundreds of metres of copper coil in the power transformer... Not to mention the chaps implementing all forms of JFET/tube input buffering, thereby lengthening considerably the signal path on the front-end. Surely a class A buffer after the LM would lengthen the signal path no more...
If you say so, Carlos... but then something sounding "better" than an amp that doesn't sound "good" (assumed concensus ex you) isn't saying very much at all...
This is true. But my idea was to keep the specific GC chips in the picture, overkill ratio notwithstanding. So how about compromising and "using" the GC's output by means of an inverting tranny buffer? Just a thought...
I am unsure how much of an advantage that is, considering the collective metres of interlink and speaker cable common to most systems... and compared to the hundreds of metres of copper coil in the power transformer... Not to mention the chaps implementing all forms of JFET/tube input buffering, thereby lengthening considerably the signal path on the front-end. Surely a class A buffer after the LM would lengthen the signal path no more...
Signal path is probably a wrong term. The important issue is the short feedback path. A rare opportunity for NFB to work as intended. And it shows - the microdynamics are stunning for a SS feedback type amp. With additional power devices within the loop there will unavidably be new poles as well. Buffering does not change this.
point (to point?) conceded...
Okay, then the NFB path is shorter... that's fine. Two observations:
1. If the LM is mounted next to the hypothetical buffer trannies on the heatsink, then the FB loop may be... say... 5cm longer, if that much. Does this matter, do you think? I would venture without calculation that the reactive properties of 5cm of PCB track or wire approaches the immeasurable in audio terms. It's not as if you're building an LNB for your satellite dish...
2. If you disagree with 1. then how about no NFB entirely? Simple class A push-pull buffer without NFB, a la Nelson Pass for example.
Without wishing to sound pedantic, exactly what are the "microdynamics" you refer to?
Okay, then the NFB path is shorter... that's fine. Two observations:
1. If the LM is mounted next to the hypothetical buffer trannies on the heatsink, then the FB loop may be... say... 5cm longer, if that much. Does this matter, do you think? I would venture without calculation that the reactive properties of 5cm of PCB track or wire approaches the immeasurable in audio terms. It's not as if you're building an LNB for your satellite dish...
2. If you disagree with 1. then how about no NFB entirely? Simple class A push-pull buffer without NFB, a la Nelson Pass for example.
Without wishing to sound pedantic, exactly what are the "microdynamics" you refer to?
Buffering an LM3875 to make it a single ended class-A sounds interesting. But you need a power mosfet for it as a source follower. A BJT will draw too much base current to keep the LM itself in class-A. The phase margin of the LM is not that big, it is guaranteed stable with gains higher than 10. So stability is tricky with an extra buffer fet, but possibly it can be done IMHO.
Biasing the LM by drawing a large offset current from it is impossible for reasonable output. For sure you get problems with the protection circuit.
Cheers
BTW If it works I will opt to name it “SEC”: Single Ended Clone (or Clown
)
Biasing the LM by drawing a large offset current from it is impossible for reasonable output. For sure you get problems with the protection circuit.
Cheers
BTW If it works I will opt to name it “SEC”: Single Ended Clone (or Clown
)- Status
- Not open for further replies.