Some time ago, I prototyped a modified version of Acaudio's UGS module, and showed it here as I was very pleased with the sound :D .
But I encountered some problems when using it as an Unbalanced to Balanced converter :
the two outputs were of opposite phase, but the peak to peak amplitude was slightly different from one output to the other...
I investigated a bit, and it appeared that this phenomenon was due to the common-mode feedback resistors (R5 & R6 on the linked schematic),
whose value was too low.
So I moved from 4K75 to the hundreds of KOhms (namely 180K) for these resistors, and it cured the problem.
But unfortunately, this modification had a negative impact on the absolute output offset stability,
which began to wander too much for my taste.
This stability is mostly related to the thermal stability of the output mosfets, and their dissipation being quite high,
I began decreasing their bias current (increasing their source resistors).
The offset tamed down, but the sound was worse... Some veil on the highs.
So I searched for a solution allowing both to keep the high bias and to lower the mosfet's dissipation.
It ended up by using a second cascode on the output stage :
A bit more complicated than the first version, but it met my needs.
The offset was down to what I was used to (+/-1mV), as for the sound. It was even better,
with a greater extent to th top of the spectrum, some added speed. I was a lucky man :D
Then I stumbled upon one of Nelson's crumbs on the UGS, and I immediately wanted to try it ;)
It's based on the use of current mirrors for the "output stage" (the level shifters), instead of mosfets.
Being born to the world of High End using mostly mosfets (thanks Nelson),
it was somewhat conceptually hard for me to move to BJTs, but I made the step,
and here's what I tested :
To go straight to the conclusion, I must confess that (for me) this version is by (a lot of) far superior to the mosfet version, cascode or not.
Ultimate speed, added life to the music, truth everywhere, instrument location, dynamics, widening...
Wowwwww... What a beast !
And from the technical point of view, it's a pleasure. I love the elegant simplicity of the design.
And no more need of common-mode feedback resistors... Provided that the transitors are thermally coupled,
it sticks to 0V, typically 0.1 to 0.2mV, both relative and absolute, without any parts matching.
I've still got some test to perform, e.g. to see if a lower current gain of the mirrors has effect on the sound.
This would allow to dicrease the dissipation, and may be to avoid a thermal coupling.
Just for the pleasure, the whole family ;)
From back to front : The original version, cascoded mosfets, and current mirrors.
And at last, a very warm thank to Nelson for allowing us to have a taste of heaven. Many many thanks :cool: :cool: :cool:
I followed this adventure on homecinema-fr
great design, wouahhhhh
I am trying to make a pcb of your V3 mirror, a lot of work to do..
Do you have a typon, or anything?
thanks for publishing here
I will replace my 2 CCS-X-Bosoz bords with these !
what to say-you ruined my day-what I can do now , except to try something different to make ;)
good job !
These look great!!!
Nice heatsinks too, Great job!
I am not used to make Pcb's and with Osmond_X on my mac it's hard...
I think at first I'll make 2 bords with predrilled Vector circuitbords
Then continue with the pcb...
thanks so much.
Thanks Alain for kind comments, and sorry Zen-Mod for having ruined your day ;)
As Marc said, I will post soon a pcb, but I'm afraid it will be double-sided. But the circuit is quite simple, and should be easy to route by yourselves. Feel free to draw your version ;)
I just made some measurements last weekend, comparing the three versions, mainly on THD basis. It's only valid as a qualitative reference, and by comparing the curves, since the PC soundcard is not the best device to measure such low levels...
you can find the results here : http://idefixes.phpnet.org/illu/av/p...Comparison.htm
where V1 is the first version (posted some time ago),
V2 is the cascode mosfets one, and V3 is the current mirrors version.
Beautiful work Cheff. Very nice indeed!
Thanks, Graeme :)
How about your version you mentionned here ? Pretty close, as you can see, but I removed the mosfets... But there was some inspiration for me in there, thanks ;) Did you power it up ?
I have built a one channel proto but I have not yet powered it up. I have made corrections regarding the polarity of the SuSy connection based on your suggestion at the time. I really like the fact that you only had minimal offset.
I used mosfets only as followers. That was how I increased the output current. I noticed that you included current gain in your mirrors.
I have a couple of questions:
1) What bias current are you using in the JFETs and how do you set it? There is a lot of Idss variation in these parts and other people may have to tweak some values to make your circuit work with BL parts from different batches. The two pots you show seem to adjust only the relative and absolute offset.
2) Did you Vbe match the current mirror transistors? I did, but I don't know how important it is.
3) Why are the emitter resistors on the mirrors so large? It seems that this would limit the output swing. I have always understood that minimal values are best - with a voltage drop of only a few tenths of a volt. Note that I used 100 ohms here. Did you choose your values based on listening tests?
4) The 450/550 transistors on the output have a Vce max of 45V. It seems risky to run them off 24V rails. This is why I used the 696/796 pair.
Once again, I really like your design. I have recently come up with a requirement for a UGS stage and you have inspired me to get back to work on it.
|All times are GMT. The time now is 08:50 AM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio