FET inputs or bipolar inputs on a power amp?

[john curl]

For audio, it can be significant.

 

[lumanauw]

Mr. Curl, what is actually happening? When I make bipolar differential input, it seems without RE sounds better than using RE

 

[john curl]

I can't tell you what to like. I have not built a bipolar transistor input for many decades. I sometimes use them in my test equipment, when fet inputs (IC based) are too noisy.

 

[traderbam]

FM distortion is generated by the non-linear S curve that exists WHEN you want BOTH low noise and a bipolar transistor input.

I don't understand. This S curve...what are its axes?

Degeneration resistor free bipolar input stages have a significant amount of FM distortion, unless the open loop bandwidth is very, very high, or the gain bandwidth is very, very high.

Could you be more specific? Are you talking about phase modulation due to the non-linear transfer characteristic of the bipolar transistors in conjunction with their junction capacitances, or something else? Why is this different for FETs which also have non-linear gm, albeit a lower order distortion, and junction capacitance?

 

[john curl]

Please contact the Walt Jung website for further information. Specifically, the link to Barrie Gilbert's articles on op amp design.
If an individual does not understand that the distortion characteristics of a differential bipolar transistor pair maps like an S curve, then I recommend some beginner courses on transistor theory, perhaps from an internet course before going any further on this topic.

 

[lt cdr data]

one of the bottom lines is that both types have been used sucessfully in many designs.

Audio subjectivism seems to defy the theory that frequently something is objectively better, yet good things and good sounds can result from something measuring less well.

Another is my pet theory that the more complex an amp. gets, the more 'clinical' the sound.

ie, start mirroring and cascoding and lots of elaborations, and the sound suffers, even tho the measured performance may improve, as elicited by amplifiers in the late 60s/70s, some of these are the finest soudning amps I have ever heard, we have not really moved forward one iota, yet modern ones measure near perfect?

why do they not sound better then?

 

[PMA]

Nuts. Under direct comparison with good up to date amps the oldies from sixties and seventies cannot compete. Experience, no guessing.

 

[john curl]

I agree with you on this, PMA. The only exceptions that I know of personally are quality tube amps and the early Electrocompaniet power amp. The Electro-Research power amp would probably qualify as well. Many early power amps had TIM (SID) limitations DUE to the undegenerated bipolar transistor input stages then used, because WE did not know any better. Today, we usually either resistively degenerate the input transconductance or use a fet input which does essentially the same thing, as fets don't have as high a transconductance as bipolar transistors at normal operating currents.
Remember folks, transconductance (Gm) is the enemy of high slew rate, especially in bandwidth limited F(t) power amps.

 

[forr]

J. CURL
---Electro-Research power amp ---

Its author is Jon Iverson, I think. Has anybody the schematics ? I am looking for it since twenty years.

 

[jcx]

http://www.diyaudio.com/forums/showthread.php?postid=501789#post501789

please point to the spectral components that support the claim that degenerated bjt has higher order distortion than the fet diff input

also note that the the loop gain is integrating for both examples so we expect that (and can see by) the spray of multiple sidebands evidence that the primary distortion is in fact "FM distortion" and is present with both fet and bjt inputs


it sure looks to me that degeneration has reduced "FM distortion" very effectively in the bjt example

 

[john curl]

jcx, you make a good point. I cannot say that a jfet input is really lower distortion than a degenerated bipolar input with the same effective Gm for the diff pair. I haven't added the weighting factor of N(squared)/2 or (N-1)!/4 which are common weighting factors for higher harmonic distortion.
However, jfets have OTHER advantages, such as LOWER NOISE and VERY LOW input current under these conditions. These advantages are most important to me, and make it easier to make direct coupled input stages,

 

[Netlist]

/Mod head on
I cleaned a bit as the core business of DiyAudio is to be a place for everyone and I believe everyone can learn from another.
Let’s not pull our hairs about ranking and lets discuss the second most important core business: Audio.
/Mod head off

/Hugo

 

[darkfenriz]

John, by "S" you mean tanh function of transconductance of long tail bipolar pair?
It is not so easy to follow your imagination of this curve, you could call it 'snake' as well, so try to be a little more understanding. By "S" some may also understand curve in S-domain. You may believe yourself to be genious enough to be a little arogant, but please do not.

regards

 

[john curl]

In the USA, it is called an S curve, because it looks like an S. Your name for it could be different, such as a 'tanh' curve which is perfectly acceptable to me. In future, I will just ignore inputs that I feel should be answered by the questioner by other means than myself.

 

[traderbam]

In future, I will just ignore inputs that I feel should be answered by the questioner by other means than myself.

How to win friends and influence people.

If anyone should have a question of me I shall respect your interest in my opinion and either attempt to answer it as best I can or admit I have no answer or respectfully decline to answer it for IP protection reasons. I encourage other experts to do this in this "diy" forum for the benefit of the less experienced participants.

 

[LineSource]

There are several good threads on cascodes and folded cascodes. The thread below cites several good papers and provides some data comparing Bipolar-FET, FET-FET, FET-Bipolar, and Bipolar-Bipolar cascodes. There are also threads on "modulated cascodes" using FET-FET topologies that readers here might find interesting. The paper data show the lowest distortion for a bipolar input with JFET cascode, but also very good results with a JFET input and bipolar cascode.


I've been using a complementary JFET-Bipolar folded cascode topology with good results in my power amps. I find this gives me the lowest noise plus a very easy load for my on-board active crossover using the PASS diff amp with diff JFET input. I pay extra to purchase gm graded JFETs, but still find it necessary to do hand matching.

http://www.diyaudio.com/forums/show...10&pagenumber=2


http://www.diyaudio.com/forums/showthread.php?s=&threadid=63187&perpage=10&pagenumber=2

 

[PMA]

The "S curves" (Walt Jung):

P.S.: another possible cure is the Current Feedback topology.

 

[courage]

Traderbam: If anyone should have a question of me I shall respect your interest in my opinion and either attempt to answer it as best I can or admit I have no answer or respectfully decline to answer it for IP protection reasons. I encourage other experts to do this in this "diy" forum for the benefit of the less experienced participants.

I second that!

 

[traderbam]

Thanks for answering my question, PMA. These particular 'S' curves are more formally known as the tranconductance curves for the two input types. Nothing new here. But how about this claim of FM? What's the rationale behind this? Assuming this acronym was intended to stand for frequency modulation.

 

[darkfenriz]

http://web.archive.org/web/20010821120135/http://www.edtn.com/analog/barrie4.htm

Hope it does not hurt to link to archieve article. If it does violate anything, then sorry and please dear moderators delete it.

regards