DIYHiFi.org

For the sake of audio
It is currently Tue Nov 21, 2017 1:46 pm

All times are UTC




Post new topic Reply to topic  [ 21 posts ]  Go to page 1, 2  Next
Author Message
PostPosted: Mon Nov 13, 2017 2:16 pm 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1163
Guys,

So over at the other place they are tinkering with a single rail supply, high (ish) voltage Preamp and if I may say so, while I like the concept of getting high performance the KISS Way, I can see some things I'd like to do different. The original is listed as "Aksa-Lender" and I have no intent in criticising it. Instead here is what I'd do instead and why:

Attachment:
Modest Proposal.png


First, there used to be good sense in using P-Channel/PNP input devices and NPN VAS, but this is not as valid now as it was decades ago. We get pretty decent P-Channel/PNP devices these days.

Flipping the whole circuit around allows a few cute things.

So, we can use J-Fets as Input devices. I like it better that way. For one, the input impedance gets really high and up to sensible frequencies remains signal independent, you don't get that with BJT's, so either you keep source impedance low (Volume control!?) or you add a J-Fet buffer. Why not J-Fet right away?

We have pretty decent P-Channel FET's these days, so why not use one as VAS? The one chose, as long as we have a few volt across it the reverse transfer capacitance is only a few pf...

Output stage - Mosfet, naturally. Resistive tail with a voltage divider bootstraps the tail of the input pair. That significantly improves overall linearity. A lot more (well, not really) than the original, part of that is that resistive tail absorbs extra signal current swing. The FET will dissipate 1.2 Watt, a fairly small heat-sink will do.

Open Loop gain is 56dB with a >100kHz -3dB point. So global loop NFB is a substantial 37dB.

At 4V PP into 3.4kOhm (see original thread why this) the sim suggests 0.0003% THD (pretty much pure H2) which is maintained at 10kHz and rises only 8dB at 100kHz to 0.00075% (still mainly H2).

At 36V PP into 3.4kOhm it looks like 0.005% THD (h2 dominant but higher order stuff creeps in) with 0.01% at 100kHz.

Closed loop FR at 3.4kOhm load is < 2Hz - > 3MHz @ -3dB, quite sufficient, I'd probably place a 100kHz lowpass before the volume control.

SNR with reference to 36V PP should be 127dB unweighted (if avoiding the NXP made input devices), which with 19dB gain is not too shabby. A-Weighted might get as much as an extra 10dB (due the J-Fet LF noise)

Anyway, just thought I show this here.

Ciao T


You do not have the required permissions to view the files attached to this post.

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Mon Nov 13, 2017 3:30 pm 
Offline
God

Joined: Thu Dec 30, 2004 10:50 pm
Posts: 993
Location: Minnesota
I like it


Top
 Profile  
Reply with quote  
PostPosted: Tue Nov 14, 2017 12:52 am 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1163
Hi,

Kuei Yang Wang wrote:
A lot more (well, not really) than the original, part of that is that resistive tail absorbs extra signal current swing. The FET will dissipate 1.2 Watt, a fairly small heat-sink will do.


Lysdexia strikes again, I meant "A lot more standing current in the output stage than the original" (around 4 times as much)...

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Tue Nov 14, 2017 7:49 am 
Online
Cow

Joined: Fri May 30, 2014 6:24 am
Posts: 147
Hi!
I don't know the background of Aksa/Leander Pre.
Just by looking T's schematic, I can see that it's meant to be a simple circuit with good performance numbers!
No split supplies, no servos, no CCS, no current mirrors, no cascoding, no CFP, etc.

Slightly surprising that there's no degeneration in the input diff stage.
Here you only trust to the power of feedback from output?


Top
 Profile  
Reply with quote  
PostPosted: Wed Nov 15, 2017 2:44 am 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1163
Hi,

The Aksa-Lender can be googled.

hanski wrote:
Slightly surprising that there's no degeneration in the input diff stage.


J-Fets have degeneration "build in" and generally differentials with J-Fets are better without degeneration.

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Wed Nov 15, 2017 9:36 am 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1163
Kuei Yang Wang wrote:
Instead here is what I'd do instead and why:

Attachment:
Modest Proposal.png


So, there comes the reply:

The other place wrote:
This design is a good (and very clever) attempt to use a similar topology but it has one issue - MOSFETs when used as a common source in the VAS lack depth of imaging, and this is something that is only audible from listening.


So, there you go, any attempt to use any Mosfet, regardless of electrical parameters or application "lack depth of imaging". That's it. Great.

:rotfl: :rotfl: :rotfl: :rotfl: :rotfl: :rotfl: :rotfl: :rotfl: :rotfl: :rotfl:

Naturally, I have used this Mosfet as VAS in a number of applications and it is just dandy on imaging depth and being holographic. On listening.

And yes, it is better than using a modified Baxandall Pair (NPN Emitter Follower before a PNP VAS), which in turn is better than a simple temperature compensated BJT VAS (which is used in the original) again on listening.

Now there are some really good reasons I selected a Mosfet as VAS. If we look at the transconductance (gain) slope it is much different from a BJT, almost a pure straight line that will only produce H2. For more on this it is worth reading Samuel Groner's commentary on D. Selfs "Audio Amplifier Design hand book".

https://pdfs.semanticscholar.org/1375/ec0b3f02e765fd493437363e90abcf2c65bd.pdf

If we combine this with the 2-quadrant J-Fet input, which has been extensively analysed by Pavel Macura, we find that the differential J-Fet input creates ONLY H3 and in very low amounts, UNLESS degeneration is applied (hence there is non).

http://pmacura.cz/diyaudio/jfetdist.htm

The result is that combining the Input stage and the VAS stage produces an open-loop system that creates only H2 & H3, certainly at reasonable signal levels. For 1kHz 4V PP into 3.4kOhm the simmed THD (and I have measured similar circuits on AP2 and know this is not far off) is 0.2%, practically all H2 at -54dB with H3 at -100dB and nothing else of consequence.

As we are going to apply a LOT of negative feedback which invariably leads to a growth in the order of the created harmonics, starting out simple keeps it simple further down the line.

Had we used bipolar transistors instead, ignoring non-linear loading effect distortion on earlier stages, the more curved transconductance curve of the BJT's, even degenerated would create more higher order components. Ok, in the end it may end up below the noise floor, who knows. But I prefer to start from a simple harmonic distortion structure before applying NFB, hence J-Fet & Mosfet.

Now, how about we review some of the cuter tricks in the circuit, that got totally missed over the terrible, unforgivable, terminally stoopeed use of a Mosfet as VAS Stage and its "lack depth of imaging"...

As the original circuit was single rail, it featured the classic input coupling cap, output coupling and DC-Blocking cap for the NFB Loop. And with all these capacitors "looking into" fairly low impedances, we will maximise the audibility of anything they do wrong and we need large value electrolytic cap's to get decent frequency response. To me this kind of a "perfect storm" of inappropriate use of coupling capacitors.

Should I ever desire to make another test circuit to evaluate the "sound of capacitors" (I have both a high voltage Tube based and a low voltage one for that in the lab), that would be a really good way to do it.

Well, at least we can can go full bore into exotic capacitors and declare that the circuit can only sound any good if build using NOS Black Gate NX type capacitors in "Super-E-Cap" configuration. And anyone using anything else cannot hear or is too poor to afford Black Gate at NOS Prices...

I tried to show an alternative approach that attempts to mitigate as much as possible the problems created by these capacitors. Naturally, that went immediately over everyone's head.

First, by having both inputs of the differential amplifier AC coupled using identical RC circuits, these in effect become extensions of the differential amp and most first order non-linearities (including microphonics etc. with careful mechanical arrangement) will cancel out. By AC coupling the inverting input, we can not only do away with the DC Blocker for the feedback loop, but we can take the negative feedback from AFTER the output coupling cap, thus enclosing it in the negative feedback loop and thus reducing it's nonlinear contribution by the feedback factor.

That should not be taken as encouragement to use garbage worthy capacitors, but it can go some way to make reasonable industrial quality devices quite usable.

Oh if only I had not used this accursed, terrible, terminal Mosfet VAS, someone might have actually stopped to think and notice these subtle (actually, trust me, NOT SUBTLE AT ALL in listening) twists...

I already mentioned the boot-strapping of the differential pair's tail resistor, by using the resistive tail of the output follower, so no need to elaborate again.

Anyway, the reaction over there is really rather amusing. Reminds me why I'm outa there. Most of the self proclaimed Guru's holding court there have little more than prejudice and limited listening experience and an Ego that needs a 20ft Hanger door as workshop entrance. Good riddance.

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Wed Nov 15, 2017 1:26 pm 
Online
Cow

Joined: Fri May 30, 2014 6:24 am
Posts: 147
Oh yes, I missed the "high" output voltage point here.
Isn't this the point where you usually say: Just use 5687 tube? :grin:

But this circuit follows your told design principles. Start with the most linear circuit first and then apply feedback.
And it's simple!
Thanks for enlightening us again!


Top
 Profile  
Reply with quote  
PostPosted: Wed Nov 15, 2017 2:53 pm 
Offline
Goat

Joined: Mon Nov 28, 2005 1:10 pm
Posts: 535
Kuei Yang Wang wrote:
First, by having both inputs of the differential amplifier AC coupled using identical RC circuits, these in effect become extensions of the differential amp and most first order non-linearities (including microphonics etc. with careful mechanical arrangement) will cancel out.

I always thought that would be the case. Good to know I´m doing that right. Excellent :good:


Top
 Profile  
Reply with quote  
PostPosted: Thu Nov 16, 2017 1:39 am 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1163
Hi,

hanski wrote:
Oh yes, I missed the "high" output voltage point here.
Isn't this the point where you usually say: Just use 5687 tube? :grin:


Absolutely. But it fun to to conceptually play with other stuff. And when I looked at the original I felt that there was something simple that could take a few adjustments to make something simply interesting (pun intended).

So, anyway, have fun, not necessarily with this exact circuit, but with the concepts and ideas embodied therein...

Things like "minimise capacitor sonic impact" and "if you a resistor somewhere where you might use a CCS there is probably a way to re-arrange the circuit so the resistor is bootstrapped, long as you look hard enough".

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Thu Nov 16, 2017 1:08 pm 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1163
Hi,

Kuei Yang Wang wrote:
"minimise capacitor sonic impact"


Incidentally, for anyone asking/moaning "Oh well, how bad could it possible be", I recommend a little light reading:

http://pa2hs.kg-uitdaging.nl/documenten/Capacitor%20Misunderstandings.pdf

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Thu Nov 16, 2017 10:43 pm 
Online
Cow

Joined: Fri May 30, 2014 6:24 am
Posts: 147
Thanks for the link T!

Totally avoiding caps is not easy in amp applications.
The only way that comes in mind is balanced SE (tube) operation with choke loads. (a transformer connected between SE output stages).
Like Vinylsavor has been writing in his blog.

Well, more thinking reminds me WE push-pull amp with a choke in series with B+ before the push-pull transformer. There's no cap in signal current path.
So there must be many other ways also.


Top
 Profile  
Reply with quote  
PostPosted: Fri Nov 17, 2017 4:12 am 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1163
Hi,

hanski wrote:
Totally avoiding caps is not easy in amp applications.
The only way that comes in mind is balanced SE (tube) operation with choke loads. (a transformer connected between SE output stages).


Well, next to ANY magnetic component I have encountered, capacitors tend to be fairly mild. Of course, for magnetic component suitable amounts of money can remove a lot of the problems, but it'll cost you. A lot. Problems caused by magnetic components can also be reduced by clever circuitry.

So, there is no silver bullet, I guess we could do balanced with CCS Load - the bigger original Pass Amplifiers are like that (active current source - it is a parallel take on a white follower, but instead in a CCS loaded common source stage).

For arguments sake, one could re-configure the output stage of the french croatian squid amplifier shown elsewhere here to eschew capacitors, for both the White Follower and the Input, but honestly, I'd think just carefully selecting these capacitors will probably be ok. Make this a bridge and the power-supplies are removed from the equation. Eliminate capacitors elsewhere...

Of course, if you replace (for example) a capacitor in a bootstrap circuit, the required level shifters and active circuitry with the addition of extra high PSU rails etc. add massive complexity.

hanski wrote:
So there must be many other ways also.


There are many just remember:

Microphones have many nonlinearities causing distortion
Speakers have many nonlinearities causing distortion
Bipolar Transistors have many nonlinearities causing distortion
Mosfets have many nonlinearities causing distortion
J-Fets have many nonlinearities causing distortion
Tubes have many nonlinearities causing distortion
Inductors and Transformers have many nonlinearities causing distortion
Capacitors have many nonlinearities causing distortion
Resistors have nonlinearities causing distortion
Wires have nonlinearities causing distortion
Even Air has nonlinearities causing distortion

Clearly the best/perfect audio system removes all sources of distortion.

Thus the best/perfect audio system contains from the actual instruments to the ear no air, microphones, speakers, active or passive components of any kind.

Then we have no distortion. Without air also no music, naturally and we are likely to loose consciousness after around 90 seconds and will die shortly afterwards, but it would be the "PERFECT" system and a small to price to pay, n'est-ce pas?

Me, instead of chasing best/perfection, what I find interesting is to try to find the optimum within the limits of imperfect components, recordings and humans.

I know, it will have never the ring of "I made the best/perfect system" and thus gets no cult following and if commercialised would not sell that well, but it's honest and often a fair bit better what certain people try to pass off as "best/perfect".

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Fri Nov 17, 2017 2:35 pm 
Offline
God

Joined: Thu Dec 30, 2004 10:50 pm
Posts: 993
Location: Minnesota
Best laugh I've had in a while T.

Isn't Atmosphere doing a bridge with out caps like you mention?


Top
 Profile  
Reply with quote  
PostPosted: Mon Nov 20, 2017 4:38 am 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1163
Hi,

Mikeg wrote:
Best laugh I've had in a while T.


Dang. And I was being so serious... :rotfl: :rotfl: :rotfl: :rotfl:

Mikeg wrote:
Isn't Atmosphere doing a bridge with out caps like you mention?


That would be Atma-sphere?

What they do is a circlotron. E.g. this...

Attachment:
circlotron.png


Even presuming the inputs are DC coupled, I still see capacitors in the power supply (the two electrolytic caps) that complete signal current loops.

I guess we could double up the number of tubes and make them into a class A shunt regulator. Then no Capacitors if inputs are DC coupled.

Does not look like this one says "Do me - I'm fun"...

Ciao T


You do not have the required permissions to view the files attached to this post.

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Mon Nov 20, 2017 11:55 am 
Online
Cow

Joined: Fri May 30, 2014 6:24 am
Posts: 147
Hi!
Off topic here, but I was making a 10mA CCS for 5687 gain stage.
I built and measured this. It was slightly better than the sim suggested. It was 2.2MOhm @ 10kHz.
Well, the sim didn't anticipate 50kHz resonance. It was worse there, but not horrible.
Attachment:
CCS_1.PNG


You do not have the required permissions to view the files attached to this post.


Top
 Profile  
Reply with quote  
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 21 posts ]  Go to page 1, 2  Next

All times are UTC


Who is online

Users browsing this forum: Baidu [Spider], hanski and 2 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB® Forum Software © phpBB Group