Is it possible? difficult?
Any existing designs like this?
Nothing obvious has turned up in searches.
Not much power is needed (e.g 1V into 300ohm), neither is high PSRR (but obviously would be welcomed), no gain is needed/wanted, so not a lot of requirements aside from good performance.
I presume it's possible in certain topologies with good matching.
At these low power levels factory matched pair might be an option, assuming their matching tolerance is good enough
BUF634 or LME49600 could have been a simple solution if the offset spec was only a bit better.
edit1:I'm considering just using LME49600 and hoping for a bit of luck that they meet at least typical spec,
it depends on how bad the open-loop performance is, but for the mentioned power levels shouldn't be too bad...
edit2:LMH6321 was mentioned as being very similar to a LME49600 in low bandwidth mode, it's datasheet is much more informative since it is being marketed more as a general use part than an audio part, should give some rough idea of LME49600's performance.
The offset of LME49600 should be better in low bandwidth mode, the typical DC offset is given as 4mV for LMH6321 which is a big improvement.
The internal schematics of LMH6321 and LME49600 are clearly not identical though, I'm inclined to choose the LMH instead for guaranteed lowest offset.
Any existing designs like this?
Nothing obvious has turned up in searches.
Not much power is needed (e.g 1V into 300ohm), neither is high PSRR (but obviously would be welcomed), no gain is needed/wanted, so not a lot of requirements aside from good performance.
I presume it's possible in certain topologies with good matching.
At these low power levels factory matched pair might be an option, assuming their matching tolerance is good enough
BUF634 or LME49600 could have been a simple solution if the offset spec was only a bit better.
edit1:I'm considering just using LME49600 and hoping for a bit of luck that they meet at least typical spec,
it depends on how bad the open-loop performance is, but for the mentioned power levels shouldn't be too bad...
edit2:LMH6321 was mentioned as being very similar to a LME49600 in low bandwidth mode, it's datasheet is much more informative since it is being marketed more as a general use part than an audio part, should give some rough idea of LME49600's performance.
The offset of LME49600 should be better in low bandwidth mode, the typical DC offset is given as 4mV for LMH6321 which is a big improvement.
The internal schematics of LMH6321 and LME49600 are clearly not identical though, I'm inclined to choose the LMH instead for guaranteed lowest offset.
Last edited:
> Is it possible?
Yes.
> difficult?
Not if you know what you are doing.
> Any existing designs like this?
Not yet ready to publish.
But an early design also meets those criteria :
https://www.diyaudio.com/forums/hea...portable-headphone-amplifier.html#post4621143
The 2021 version is all discrete.
Cheers,
Patrick
Yes.
> difficult?
Not if you know what you are doing.
> Any existing designs like this?
Not yet ready to publish.
But an early design also meets those criteria :
https://www.diyaudio.com/forums/hea...portable-headphone-amplifier.html#post4621143
The 2021 version is all discrete.
Cheers,
Patrick
Non-global feedback is called"nested feedback " plenty of articles on it in EW and yes some design engineers use it.
Easy Chip Amplifier: Audio Amplifier Feedback - Nested Feedback Loops
Easy Chip Amplifier: Audio Amplifier Feedback - Nested Feedback Loops
Last edited:
Thanks, the article has a lot of valuable info in it, funny we ended up at similar conclusions with the LMH's low offset.
It's shame PDFs dont show up more readily on google, probably a lot more valuable stuff hidden away.
Paralleling is a good idea, is there a reason you dont recommend LMH or LME for paralleling aswell in the article, since it is cheaper... or maybe it wasn't back in 2016? Ah, I see you have prices and it isn't.
Hmm, then would it be to keep the input impedance higher for the AD844?
Maybe I missed it but couldn't see any THD measurements, I'm most interested in what the high order harmonics are like with these buffers.
Last edited:
Not necessarily. You can also use simple local feedback instead of nested. Nested is sort of between global and local as it can span single or multiple stages.
Nested is not for the faint of heart, you have to juggle multiple loops for stability, that all influence each other.
Jan
Non global is "non global " its not an overall feedback control of the whole circuit in meaning and that's what the OP posted .
So encompassing more than one audio gain stage in a comprehensive power amp circuit ( non simple ) is still non global.
But I do agree --certainly not simple to employ due to various parameters having to be taken into consideration .
So encompassing more than one audio gain stage in a comprehensive power amp circuit ( non simple ) is still non global.
But I do agree --certainly not simple to employ due to various parameters having to be taken into consideration .
The circuits I have mentioned are all "local" feedback.
No nested feedback loop.
Just to clarify.
Patrick
No nested feedback loop.
Just to clarify.
Patrick
I have some LMH6321 on the way and will try taking some measurements of their open-loop performance.
Based on DSs there was a couple other appealing things about this part other than the low offset when compared to the BUF634A (and maybe the LME49600 to some degree).
When roughly compared the HD at high power levels is similar in amount but 2nd harmonic dominant vs 3rd on BUF634A.
Output impedance is slightly lower also.
Both these buffers HD performance seemed underwhelming at first glance, but they are always operating well out of class A in the given specs.
The BUF634A has one graph for 10V into 1K load (still not class A) and there is already an enormous difference.
Excited to see how well the LMH will hold up with a real headphone load still operating deep in class A.
There was something unusual though, the graphs for LMH begin to show rising 3rd HD with decreasing amplitude into a 50R load (at 100kHz and 1MHz), the 100R graph HD falls continuously like you'd expect ... how could that be?
It is purely HD, not HD+N.
Based on DSs there was a couple other appealing things about this part other than the low offset when compared to the BUF634A (and maybe the LME49600 to some degree).
When roughly compared the HD at high power levels is similar in amount but 2nd harmonic dominant vs 3rd on BUF634A.
Output impedance is slightly lower also.
Both these buffers HD performance seemed underwhelming at first glance, but they are always operating well out of class A in the given specs.
The BUF634A has one graph for 10V into 1K load (still not class A) and there is already an enormous difference.
Excited to see how well the LMH will hold up with a real headphone load still operating deep in class A.
There was something unusual though, the graphs for LMH begin to show rising 3rd HD with decreasing amplitude into a 50R load (at 100kHz and 1MHz), the 100R graph HD falls continuously like you'd expect ... how could that be?
It is purely HD, not HD+N.
The LMH6321 is up and running as a standalone buffer.
Only 1mV offset and 2mV offset in the channels.
After a few hours of listening I am pleased to be able to say it does sound very good, but more importantly it exhibits some of the same desirable sound characteristics of other zero feedback HP amps (single supply, AC coupled) I've been using, I would say the LMH is an improvement in fidelity over these other amps though.
This is to try and answer the question if the 'desirable sound characteristics' of ZFB amps is due euphonic distortion or some real advantage of ZFB.
I was unsure before, now I am doubtful it is just distortion.
That is not to say GFB is bad, only that GFB is not infallible. I feel GFB amps had certain characteristic strengths compared to these ZFB ones aswell.
Only 1mV offset and 2mV offset in the channels.
After a few hours of listening I am pleased to be able to say it does sound very good, but more importantly it exhibits some of the same desirable sound characteristics of other zero feedback HP amps (single supply, AC coupled) I've been using, I would say the LMH is an improvement in fidelity over these other amps though.
This is to try and answer the question if the 'desirable sound characteristics' of ZFB amps is due euphonic distortion or some real advantage of ZFB.
I was unsure before, now I am doubtful it is just distortion.
That is not to say GFB is bad, only that GFB is not infallible. I feel GFB amps had certain characteristic strengths compared to these ZFB ones aswell.
LMH6321 is OK-ish if you have sensitive phones and requires low current.
15mA bias means maximum 30mA Class A.
In distortion terms it cannot be compared to discrete output stages with 150mA bias.
You can always use 10x in parallel.
Just need to potent source to drive the 25k 35pF Zin.
Patrick
15mA bias means maximum 30mA Class A.
In distortion terms it cannot be compared to discrete output stages with 150mA bias.
You can always use 10x in parallel.
Just need to potent source to drive the 25k 35pF Zin.
Patrick
I'd say most dynamic headphones get extremely loud (to my ears) before hitting 30mA. Does having a large surplus of bias often help reduce distortion considerably or are you just talking about cases where you'd realistically need near 150mA (some planars)?
BTW Are you sure it's 15mA and not 11mA? I left CL pin floating, as it was simplest, which disables the current limiting circuit entirely but noticed the datasheet implies a lower supply current (11mA) when leaving it open compared to the 15mA when limiting to the max of 300mA. I thought it was added power consumption from the current limit circuit.
Would any of those discrete designs have low enough offset to not need a servo/cap?
This is another advantage of LMH (though I have no experience with servo effects on sound).
Oh and still waiting for a DAC to arrive before I can try to take some measurements of the LMH
BTW Are you sure it's 15mA and not 11mA? I left CL pin floating, as it was simplest, which disables the current limiting circuit entirely but noticed the datasheet implies a lower supply current (11mA) when leaving it open compared to the 15mA when limiting to the max of 300mA. I thought it was added power consumption from the current limit circuit.
Would any of those discrete designs have low enough offset to not need a servo/cap?
This is another advantage of LMH (though I have no experience with servo effects on sound).
Oh and still waiting for a DAC to arrive before I can try to take some measurements of the LMH
Last edited:
> Does having a large surplus of bias often help reduce distortion considerably ?
Yes.
> Are you sure it's 15mA and not 11mA?
You can change bias to 15mA.
> Would any of those discrete designs have low enough offset to not need a servo/cap?
Yes. Many examples already exist and built by others.
The DAO SE all-FET Class-A ZGF Headphone Amplifier
DAO - Original-Ton
UTHAiM -- Just for Fun
The Pioneer Super Linear Circuit
Patrick
Yes.
> Are you sure it's 15mA and not 11mA?
You can change bias to 15mA.
> Would any of those discrete designs have low enough offset to not need a servo/cap?
Yes. Many examples already exist and built by others.
The DAO SE all-FET Class-A ZGF Headphone Amplifier
DAO - Original-Ton
UTHAiM -- Just for Fun
The Pioneer Super Linear Circuit
Patrick
This design by Wayne Kirkwood is direct coupled (all DC) and only local feedback within the THAT1646 or DRV134 driver IC. Overall, no global feedback. It sounds and measures excellent. One of the lowest distortion headphone amps I have measured.
Simple Class A Headphone Amp Using THAT1646
It can be taken to an extreme and used with a large output BJT and big heatsinks for a 10W class A speaker amp:
Simple Class A Headphone Amp Using THAT1646
It can be taken to an extreme and used with a large output BJT and big heatsinks for a 10W class A speaker amp:
Last edited:
Ok, the LMH at least was an inexpensive and easy way to test the water before building a discrete one.
Will the 2021 version you are designing also use EOL/rare parts?
The original requirement should have been 'open-loop' instead of 'no global feedback', my mistake.
I bet this does sound great though, considering there are feedback loops within the DACs I've been using with the LMH and it can still 'sound' open-loop that would seem to reinforce the idea that the output stage is where feedback loops are most detrimental.
> Will the 2021 version you are designing also use EOL/rare parts?
The Pioneer SL is already all active parts.
Patrick
The Pioneer SL is already all active parts.
Patrick
Ah, and this whole time I thought the SL and the current drive SLT were the same amp...
Would it be possible to build it without the input/gain stage?
It probably still performs better than the IC buffers with gain, but if the gain stage could be eliminated that would make it a whole lot easier to build (less matching) and probably perform even better... that is if the gain stage isnt integral to the design.
Last edited:
Another relevant design: the https://www.diyaudio.com/forums/solid-state/235695-nfb-line-amp-gainwire-mk2-97.html#post4629715