Variable operating biass output ?

[AKSA]

Thank you Edmond, seminal information indeed.....

Bloody fascinating, actually.

Hugh

 

[Edmond Stuart]

Thank you Hugh.

Cheer,
E.

 

[Edmond Stuart]

Spectra

Hi Bigun,
.............
As for a 'Hiraga like' spectrum, indeed it looks nice. But remember, this is the spectrum of the OPS in isolation. If implemented in a complete amp with global NFB and Miller compensation, the global loop gain will show a roll-off by 6dB/octave. As a result, you will get a different picture: Although much lower, the spectrum will be CCW 'tilted' by 6dB/octave.
.....................

As shown here: AB2-EC-Harm . As you see the high frequency harmonics rose considerably, that is, relatively of course. Nevertheless, all components are well below 10ppm. So only JC can hear them.

And here again the harmonics of the OPS in isolation: Harmonic products

Cheers,
E.

 

[bobodioulasso]

This Nelson Pass patent is not "variable bias" but allows high bias and low dissipation.

Efficient high fidelity audio power amplifier

 

[Bigun]

As shown here: AB2-EC-Harm . As you see the high frequency harmonics rose considerably, that is, relatively of course. Nevertheless, all components are well below 10ppm. So only JC can hear them.

And here again the harmonics of the OPS in isolation: Harmonic products

Cheers,
E.

Thanks for posting this ! and this does look nice, and milli % too

I have taken to looking at O/P stages in isolation because I find it helpful to understand how things behave before applying gnf and also because I like the idea of having a 'catalogue' of O/P topologies to select from. It's the 'lego' approach !

 

[Edmond Stuart]

Thanks for posting this ! and this does look nice, and milli % too

You're welcome.

I have taken to looking at O/P stages in isolation because I find it helpful to understand how things behave before applying gnf and also because I like the idea of having a 'catalogue' of O/P topologies to select from. It's the 'lego' approach !

BTW, I've also simmed the LT1166 with error feedback enable, i.e. feeding the input signal to pin #2 (see fig. 8 in the data sheet).
THD figures were almost as good as with my circuit, but the step response was horrible. Also the local FB loop was hardly stable:
Unity loop gain frequency: 900kHz (rather low)
Phase margin: 50 degrees (too low)
Gain margin: 6dB @ 3MHz (also too low)
I have tried to improve it, though without success.
So I wouldn't recommend to use the features behind pin #2 (not compatible with 'Lego' )

Cheers,
E.

 

[gaetan8888]

Here it is: Auto Bias part II

Cheers,
E.

Hello Edmond

A bit of topic question.

Any ideas wen your DiAna PC based distortion analyzer web page will be ready ?

Thank

Bye

Gaetan

 

[Edmond Stuart]

DiAna

Hi Gaetan,

Hard to say. The trouble is that I can't find a better sound card. At the moment I've a few 'Waveterminal 192X', one of them modified. But the performance at higher frequencies is still below my requirements: a measurement floor of -140dB at 20kHz.
so I shelved the project temporarily until I find something better.

Cheers,
E.

PS: This is what I'm looking for: 32bits, 192kHz SR, strict separation of the channels and the best op-amps you can get (or at least easy to upgrade). This means no stereo ADCs of DACs or dual/quad op-amps used for more than on channel.

 

[Edmond Stuart]

erratum

[snip]
BTW, compared to the LT1166 (which is supposed to do the same job), the distortion at 1kHz is about 25 times lower.
.......

I have to retract that claim. Although the distortion of my circuit was indeed 25 times lower, it was an inappropriate comparison, just not fair.
The VAS as used in my simulations was okay for my circuit, but totally incompatible with a LT1166. Therefore I got such a gross discrepancy. As a matter of fact, the output impedance of the VAS was way too high for that chip. Decreasing the impedance cured the problem. Now I get comparable results (see my website).
My apologies for the misleading info.

Cheers,
E.

 

[Edmond Stuart]

Hi all,

Please see an improved version of auto bias + EC: AB2-EC-V2

THD-20k is 10dB lower.

Cheers,
E.

 

[tiefbassuebertr]

Hi all,

Please see an improved version of auto bias + EC: AB2-EC-V2

THD-20k is 10dB lower.

Cheers,
E.

is it possible to implement your auto bias circuit in an amplifier topology like that one from schematic of the follow link ?
http://alteclansingunofficial.nlenet.net/proelectronics/proamplifiers/9446A_Service.pdf

 

[Edmond Stuart]

Hmm... Without a considerable redesign, I don't think so.
BTW, why? If works don't fix it.

 

[tiefbassuebertr]

Hmm... Without a considerable redesign, I don't think so.
BTW, why? If works don't fix it.

Not for the Altec amp device. But perhaps for a new project with grounded collectors.
BTW (off topic) - this weekend is a great jazz event in your aera - in Rotterdam
http://www.northseajazz.com/schedule.aspx?eventid=154&date=20100710

 

[Edmond Stuart]

Not for the Altec amp device. But perhaps for a new project with grounded collectors.
BTW (off topic) - this weekend is a great jazz event in your aera - in Rotterdam
Programma 2010 - North Sea Jazz

After having a closer look, I see too much troubles.

Cheers,
E.

PS: Sure, great happening in Rotterdam.

 

[Edmond Stuart]

yet another version; the last one?

Hi all,

Please see an improved version of auto bias + EC: AB2-EC-V2

THD-20k is 10dB lower.

Cheers,
E.

Above version is obsolete and has been superseded by: AB2ECV4

Cheers,
E.

 

[tiefbassuebertr]

Which commercial brands uses such topology like that?

 

[Edmond Stuart]

Pleeeaaassseee tiefbassuebertr, no-one has made use of this topology (and the previous ones) yet. This stuff is NEW.

 

[PHEONIX]

Above version is obsolete and has been superseded by: AB2ECV4

Cheers,
E.

Hello Edmond

I like your work. Just a question on the That 300 series parts needed for auto bias circuit to work, and my question is what are the critical parameters that need matching and how well matched do they need to be.

It sounds like a stupid question but I was thinking about alternative matched parts that might be usable , like HN3A51F (NPN) and HN3C51F(NPN) both are Toshiba parts and have higher voltage operation also.

Regards
Arthur

 

[Edmond Stuart]

Hello Edmond
I like your work.

Thank you!

Just a question on the That 300 series parts needed for auto bias circuit to work, and my question is what are the critical parameters that need matching and how well matched do they need to be.

It sounds like a stupid question but I was thinking about alternative matched parts that might be usable , like HN3A51F (NPN) and HN3C51F(NPN) both are Toshiba parts and have higher voltage operation also.

Regards
Arthur

Hi Arthur,

The maximum allowable offset voltage (delta Vbe) is 1mV. In practice, this means matched trannies, which are dielectrically isolated on a common monolithic substrate.
Also, you can't use duals. Topology 'AB2' needs three matched NPNs and three matched PNPs, while 'AB2ECV4' needs four of each. So the THAT chips are about the only viable option. Matching between NPN and PNP is not critical.

One more thing: regarding 'AB2ECV4' for example, perhaps it looks tempting to use two THAT340s, one for the odd numbered trannies and one for the even numbered trannies. Surely, this will result in less crossings on the PCB, but it's not a good idea. All PNPs respectively all NPNs must reside on the same substrate.

As for a higher voltage rating, you don't need it. The max. Vce is about equal to the max. Vgs, usually below 10V. In case of BJTs in the output, even lower, say 3V or so.

Cheers,
E.

 

[vladn]

Congratulations Edmond ! Your latest contraption looks like a genuine advance in class AB design. Definitely worth a publication and perhaps a patent application (if you want to commercialize the idea).

Works nice in simulation even with switching mosfets (although due to high threshold these would benefit from bootstrap powered CCSs/drivers to get full rail output).