John Curl's Blowtorch preamplifier part II

[john curl]

Returning to a practical 4Q all fet design:
R1-4 are 10 ohms in this design, probably the LOWEST value that I would use. It will change with the Idss, and TYPE of the jfet quad used.
The 1K load is close to the maximum that I would use, and often I use a smaller value.
Most of the distortion performance will depend on the MOSFETS used. Some don't match well, some are just better than others.
Now why do I keep showing this sort of design? Well, I use it a lot. I am a true believer in design elegance and the most linear OPEN LOOP performance, rather than brute force, using lots of feedback. It works for me, measures relatively well, if done carefully, and usually does not even need AC balance (which can be a problem in other ways).

 

[bcarso]

Returning to a practical 4Q all fet design:
R1-4 are 10 ohms in this design, probably the LOWEST value that I would use. It will change with the Idss, and TYPE of the jfet quad used.
The 1K load is close to the maximum that I would use, and often I use a smaller value.
Most of the distortion performance will depend on the MOSFETS used. Some don't match well, some are just better than others.
Now why do I keep showing this sort of design? Well, I use it a lot. I am a true believer in design elegance and the most linear OPEN LOOP performance, rather than brute force, using lots of feedback. It works for me, measures relatively well, if done carefully, and usually does not even need AC balance (which can be a problem in other ways).

What DMOS would you suggest?

 

[bcarso]

At the time we were arguing about the wide band open loop gain nonsense (gee, what else is new). I was showing that trimming DC Aol to essentially infinite (and even a large negative number) made no difference in the AC response at higher frequencies.

Sigh.

 

[john curl]

This is where is 'all hits the fan' bcarso. The BEST parts are virtually impossible to get. Next best, like Fairchild, might be OK, IR can be awful.

 

[dvv]

And if you should find them somewhere, there is no guarantee that they are the originals, chances are they are Chinese copies.

They are selling 2SK170 like its production is still on full force. Yeah, sure.

 

[ticknpop]

John,

how about Hitachi 2K216/2J79 or 2SJ313/2SK2013

 

[john curl]

Normally, Hitachi is not the best, but I don't know the j313, etc.

 

[wayne]

This is where is 'all hits the fan' bcarso. The BEST parts are virtually impossible to get. Next best, like Fairchild, might be OK, IR can be awful.

The best parts are easy only 10 meters away.
I have used IR, Fairchild and Zetex (ZVN3310 smt) they all work but the Toshibas sound the best and are very consistent.

 

[Bob Cordell]

Returning to a practical 4Q all fet design:
R1-4 are 10 ohms in this design, probably the LOWEST value that I would use. It will change with the Idss, and TYPE of the jfet quad used.
The 1K load is close to the maximum that I would use, and often I use a smaller value.
Most of the distortion performance will depend on the MOSFETS used. Some don't match well, some are just better than others.
Now why do I keep showing this sort of design? Well, I use it a lot. I am a true believer in design elegance and the most linear OPEN LOOP performance, rather than brute force, using lots of feedback. It works for me, measures relatively well, if done carefully, and usually does not even need AC balance (which can be a problem in other ways).

Hi John,

Which JFETs are you specifying for this design? 2SK170/2SJ74?

What tail current are you running through the input stage?

Why MOSFETs for the second stage. I'm guessing BJTs would be more quiet and lower distortion.

Cheers,
Bob

 

[john curl]

Fets are better, because they generally sound better.

 

[ticknpop]

John,

The 2sJ313/2SK2013 are a discontinued Toshiba pair that Pass Labs is using in some current preamps and possibly power amps. I believe Nelson has the entire remaining inventory stored in his wine cellar.

 

[Waly]

the most linear OPEN LOOP performance, rather than brute force, using lots of feedback.

Really? It appears to me this design has actually gobbles of feedback. At a quick visual inspection, assuming as mosfets something like the 2SK213/2SJ79 pair (gm=35mA/V @ Id=10mA) the differential open loop gain is about 85-90dB. Subtract the differential closed loop gain of 26dB makes for 64dB of loop gain (at low frequency, of course). That's about as much you would get by using an average opamp.

I would think the open loop bandwidth is about 1KHz, nothing to call home about, even if this would make any difference (it doesn't, actually).

P.S. To make some die hard fans happy, this design may qualify as a "CFA". The impedance at the negative input is low, etc...

 

[Bob Cordell]

Fets are better, because they generally sound better.

Hi John,

Which JFETs are you specifying for this design? 2SK170/2SJ74?

What tail current are you running through the input stage?

What idle current is flowing in the MOSFET stage?

Cheers,
Bob

 

[bcarso]

John,

The 2sJ313/2SK2013 are a discontinued Toshiba pair that Pass Labs is using in some current preamps and possibly power amps. I believe Nelson has the entire remaining inventory stored in his wine cellar.

Now you're talking. I can imagine a whole new set of criteria for semiconductors mapping to wine terminology.

"We can't use those 2SK389V parts yet---they aren't ready."

 

[elektroj]

This is where is 'all hits the fan' bcarso. The BEST parts are virtually impossible to get. Next best, like Fairchild, might be OK, IR can be awful.

John, those were Supertex VN/VP parts, right? What was so special about them?

 

[elektroj]

"We can't use those 2SK389V parts yet---they aren't ready."

"... but lets try BL grade from 2004 batch, they go very good with Fairchilds"

 

[bcarso]

"... but lets try BL grade from 2004 batch, they go very good with Fairchilds"

And those low rbb' bipolars over there by the older Sauternes---I find they impart a sweeter sound in this circuit position, although sometimes stability can get sticky.

 

[SY]

Read the descriptions for NOS tubes. Not that different, really. And just as obsessive as a Burgundy-phile: "Oh, you need the late 50s version with the square holes, black plates, and d-shaped getter from the Glitzenholen factory. The round getter ones from Arboisette and the later Glitzenholens with round holes are horrible."

 

[31697B]

audibility of jitter

The jitter analyzer arrived..... nice with two inputs to do comparisons. Internal or external master clock.

Audibility of jitter? How much is too much? ... and what can be done about it or to improve on it from the user view point;

A few of many re. jitter:

View attachment Jitter ADC 1 USB.pdf
View attachment Specifying_Jitter_Performance.pdf
View attachment AES 2926 jitter.PDF
View attachment AES 6948 jitter.pdf

Too bad this site doesnt allow larger file size... some very good stuff cannot be included here.

Most of the efforts go into the practical design and its limitations and how to improve on them and not about the theory.... which i hear too much about here at DIY-land. IMO.


THx-RNMarsh

 

[elektroj]

.."Oh, you need the late 50s version with the square holes, black plates, and d-shaped getter from the Glitzenholen factory. The round getter ones from Arboisette and the later Glitzenholens with round holes are horrible."

Word!
(I usually stick to Leckmichamarschen production, waaay better than the other brands.)