John Curl's Blowtorch preamplifier

Status
Not open for further replies.
Jump to Latest
#6,943
Hi Terry,

You are comparing apples with oranges (just as Pavel). To make a fair comparison, you should use two LTPs as show in my previous post and keeping the total amount of drain currents the same. Then you will get a THD difference of a factor 2 instead of 8.

Cheers,
Edmond.

edit: In case that the P and N-channel fets are exactly complementary, you can of course use all fets of the same gender for the LTP test version.
 
#6,945
Edmond Stuart said:
Hmm.... but to what extent and why? The nonlinear behavior of the PNPs may be quite different from the NPNs.
Click to expand...

Hi Edmond,

that is why I said "may be reduced" . In some cases thought this reduction could be substantial (several times) .

Edmond Stuart said:
I'm also a fan of fully complementary designs
Click to expand...

I did use a complementary LTP on the input of one of my commercial designs, namely Creek Audio 5350SE (Stereophile Class A rated). Would prefer to make it FET input but that was considered too expensive at the time.

Cheers

Alex
 
#6,947
Edmond Stuart said:
Hi Terry,

You are comparing apples with oranges (just as Pavel). To make a fair comparison, you should use two LTPs as show in my previous post and keeping the total amount of drain currents the same. Then you will get a THD difference of a factor 2 instead of 8.

Cheers,
Edmond.

edit: In case that the P and N-channel fets are exactly complementary, you can of course use all fets of the same gender for the LTP test version.
Click to expand...

Hi Edmond,

This is arguable of course. Yes we could run the n's at double the
current, but practically there would be 2 x PS draw.

It could also be said that one could run 2 parallel pairs of n
devices running at double current.

Still, the complimentary jfets are a beautiful thing, being self biasing,
near zero tempco, having very low drift etc.

If I have time, I will do a few more sims on weekend, but am very
busy learning a new 3D cad package and my brain is hurting.
This is the 6th CAD system now, where does it end 😱

cheers

T
 
#6,949
Terry Demol said:
Hi Edmond,

This is arguable of course. Yes we could run the n's at double the
current, but practically there would be 2 x PS draw.

It could also be said that one could run 2 parallel pairs of n
devices running at double current.

Still, the complimentary jfets are a beautiful thing, being self biasing,
near zero tempco, having very low drift etc.

If I have time, I will do a few more sims on weekend, but am very
busy learning a new 3D cad package and my brain is hurting.
This is the 6th CAD system now, where does it end 😱

cheers

T
Click to expand...

Hi Terry,

In the meantime, I too realized that it is a bit arguable, but I think, when it comes to a comparison of different topologies, it's fair to start with the same amont of silicon and power consumption, theoretical speaking. In practice, it's a slight different story.

Indeed, it's a beautiful thing, but where to get them nowadays (in quantity and in the future)? That's why I'm forced to use an arrangement with BJTs plus the inevitable and evil current sources.:sad:

As for CAD, software developments will never end and and you will never stop learning. Good luck with mastering your new package.

Cheers,
Edmond.
 
#6,950
forr said:



Hi Bob,

Do you have an example of a practical schematics of such a cascode stage using a biasing signal derived from the output of the amp ? My concern relies in the DC bias of the bases of the devices providing the cascode scheme.

By the way, albeit knowing cascodes for thirty years, I discovered the term "telescopic cascode" only today on a french forum.

Cheers
Click to expand...


Hi forr,

The cascode bases in the non-driven version are biased at +15V. The introduction of the drive to the cascode is DC coupled from the output of the amplifier. The closed-loop gain of the amplifier is 20. The feedback network is 8870 ohms in series with a shunt of 464 ohms. An identical network is used to drive the cascode bases from the output of the amplifier, but in that case the shunt resistance is returned to the +15V supply reference. The actual DC voltage on the cascode bases is then approximately 14.25 volts, as it is pulled down slightly by the resistive network. For those who would worry about the 1.6 mA of imbalance current drawn from the output node of the amplifier, I suppose that an identical dummy resistive network could be attached to the -15V supply to balance it out, but I have not bothered to do that.

Cheers,
Bob
 
#6,951
G.Kleinschmidt said:



Asymmetrical and symmetrical circuits are just different means to the same end. Your need to constantly reference your 50W amp paper is bordering on the pathological.
There have been a few symmetrical designs presented here that do not lack over your design in terms of linearity and speed and some lesser ones that certainly lack nothing in terms of adequate performance for the intended application. Why not offer some new ideas after ~25 years?
Click to expand...


Hi Glen. I have presented a lot of new stuff; witness the driven cascode I just discussed. I cite My MOSFET amplifier paper from 25 years ago because it is readily available on my web site and is very convenient to reference when discussing that unipolar approach to a very high performace input/VAS combination.

You are perfectly right that there have been many improved versions of the full complementary input/VAS presented whose performance is just as good as my unipolar design. I'm just trying to point out that those who say that the full-complementary approach is superior are wrong, or at least they do not present any evidence to back up their claims.

There is nothing WRONG with the full complementary approach if it is executed well, but it does suffer the disadvantage of needing both sex of input JFET pairs, whose non-availability has been bemoaned. This of course is not a problem for those who are happy to use BJT input pairs. For those of us who prefer FET input stages, the high performance unipolar input stage is a good alternative to the full complementary approach

Cheers,
Bob
 
#6,953
Bonsai said:
I disagree with the assertions.

You need to look at the final result.

Bottom line is Cordell's amp produced a superb set of measured results that set the benchmarl for 20 years. I'd venture that PGP sets th e latest breakthrough performance levels.

If we ar e going to be even handed in this discussion, lets also hammer (or is it nail?) JC for continually referencing his 1960's work.
Click to expand...


Thanks, Bonsai. Your kind comments are much appreciated.

Cheers,
Bob
 
#6,954
Bob Cordell said:
[snip]
There is nothing WRONG with the full complementary approach if it is executed well, but it does suffer the disadvantage of needing both sex of input JFET pairs, whose non-availability has been bemoaned. This of course is not a problem for those who are happy to use BJT input pairs. For those of us who prefer FET input stages, the high performance unipolar input stage is a good alternative to the full complementary approach

Cheers,
Bob
Click to expand...

Hi Bob,

I'm perfectly happy with BJTs. I know that certain jfets produce slightly less noise (courtesy of John Curl), but do we really need them in the input stage of a power amp, where noise is less of a concern?
Do jfets produce less distortion? Maybe, but still I don't need them for my power amp. My latest front-end does ~50ppb distortion at 20kHz. Will those who think that's too much, stand up.

Cheers,
Edmond.
 
#6,956
Edmond Stuart said:


Hi Bob,

I'm perfectly happy with BJTs. I know that certain jfets produce slightly less noise (courtesy of John Curl), but do we really need them in the input stage of a power amp, where noise is less of a concern?
Do jfets produce less distortion? Maybe, but still I don't need them for my power amp. My latest front-end does ~50ppb distortion at 20kHz. Will those who think that's too much, stand up.

Cheers,
Edmond.
Click to expand...


Hi Edmond,

That's fine. I just think it is a personal preference, and it may not always be completely logical. BJT front ends can do great. Like many other things audio, people's preferences for BJT vs JFET front ends may not always have to do just with measurable distortion alone. I tend to agree with you that in many cases the BJTs will win in that department. Some feel that JFETs just sound better. Others feel that they are more tolerant of EMI effects.

This is probably not unlike the MOSFET vs BJT issue for output stages.

Cheers,
Bob
 
#6,959
Edmond Stuart said:


Hi Bob,

I'm perfectly happy with BJTs. I know that certain jfets produce slightly less noise (courtesy of John Curl), but do we really need them in the input stage of a power amp, where noise is less of a concern?
Do jfets produce less distortion? Maybe, but still I don't need them for my power amp. My latest front-end does ~50ppb distortion at 20kHz. Will those who think that's too much, stand up.

Cheers,
Edmond.
Click to expand...

Edmond,

Escuse my ignorance, but how do you isolate the linearity of a
power amp front end that works in a GFB amp?

Don't you have to make a lot of assumptions such as the reactive
OP stage load, etc.

Or is this figure based on a purely resistive load for the front end.

WRT 50ppb, I have designed an open loop front end with just some
hawksford correction, and a few other tricks that simulates the
same. No global feedback.

cheers

Terry
 
Status
Not open for further replies.
Top Bottom