UGS Power : it's U.P.

Hi, Cheff!

Amazing work and good design (best I can think of - at least for DIYers).

Originally posted by CheffDeGaar
I've added an on board DC protection, but I must confess I'm amazed by the offset performances : under the mV on diff offset, and under 5mV on absolute offset... :cool:

About the DC protection... is calculating diff really necessary (also it is nothing wrong with it)? I mean if you check them separately and they have no DC offset can their difference have an offset. I didn't yet found a mathematical proof to support my assumption.

What relay do you use for it?

Thanks,
Matej
 
Re: Re: UGS Power : it's U.P.

Tiens, les français se sont réveillés ;)

matejS said:

Amazing work and good design (best I can think of - at least for DIYers).

About the DC protection... is calculating diff really necessary (also it is nothing wrong with it)? I mean if you check them separately and they have no DC offset can their difference have an offset. I didn't yet found a mathematical proof to support my assumption.

What relay do you use for it?

Thanks,
Matej

Thanks Matej :),

Well, what I want is no offset across the speaker. I could have go the way you mention, but it would have make things more complicated.
I assumed it is very unlikely that both sides of bridge go wrong at the same time, so a differential offset makes it perfectly.
And if you own a Son of Zen, you can use this DC protection, since both output voltages are non zero (Vsupply/2), but their difference is zero ;)
The relay is a 24V DCPO (RTE24024 from Tyco/Schrack)

And to reply your email, no I do not have gerber files. I'll post a pcb later (I'm still prototyping), but only the way I use it, i.e. pdf or ps format.

Regards,
 
Hi all,

I spent some time listening and comparing the global/local feedback configuration, and guess what... ;)

How to put words on that... :xeye: I must mention here that all the following is purely subjective, and that it involves pretty tiny and diffuse impressions :)

Compared to the original global feedback scheme, I feel the local feedback comes closer to the AX. I know it's not the same topology, but the the sound becomes a bit more lusher and warmer in the bass/mid part of the spectrum, and that the highs are rolled off a bit. What I call the "AX fuzzyness" comes back again, but with much less noticeable than in the AX's case, and it narrows the scene a bit (it's a very very thin difference). The sound seemed to have lost some air.
Since the output mosfets are no longer in the FB loop, I think they bring their own character to the sound, but may be I'm wrong...
But don't get me wrong, the sound remains quite superb, and some of you may prefer the local FB configuration for its warmer aspect. It all depends on the rest of your system, so feel free to compare and choose ;)

These tests were made with an open loop gain of about 60dB (sorry, I haven't had the time to measure it).

BTW, thanks to Mr Pass and Ian MacMillan - and their breadcrumbs in the AX builders thread, this aspect of the OL gain made me think a lot ;)
So I decided to test and listen what was happening with a lower OL gain. I didn't want to decrease the current mirrors gain, since first it required a heavy manipulation (removing the pcb to access the solder joints), and second since I feel that the current capabilities for driving the output mosfets should be left as is, or quite close.
So I just changed the value of the "ugs" output resistor (10K at the start) by paralleling socketed lower values resistors.
A 5K equivalent resistor didn't changed the things, at least ihmo.
So I tried a 1K equivalent resistor. It lowered the OL gain by 20dB (theroretically), i.e. an OL gain of a rough 40dB.

How to say... with this 1K resistor (in global FB scheme), I felt I was half way the previous local/global FB test : The sound got lusher and warmer, but without getting fuzzy and without loosing clarity, air and accuracy in the highs.
Always with the 1K eq. resistor, I tested the local FB, but didn't like the result this time.... I lost the "aerial" impression, got a darker sound with a feeling of slow basses. Of course these are tiny impressions, but they're mine :D

I also tested a lower resistor value (700R), but I didn't feel any differences.

To sum up, I think I'll stick to the 1K output resistors with global feedback. But your mileage may vary, so feel free to make your own tests, that's pretty easy, and that's the fun with DIY :)

Anyway thanks to NP and Ian for pointing this out to me ;)

Cheers,
 
Hi Cheff,

Sounds like you are making good progress and the results of your investigations are certainly interesting. The differences between local and global feedback are pretty much as expected based on what others have said (I’m sorry to say I haven’t yet tried it myself). Note quite sure why local feedback should reduce the air but the rest is most likely down to the reduced damping factor with local feedback and as you say, no feedback correction of output device non-linearity.

Your comments on reducing the amount of feedback are particularly interesting as I am not aware of anyone who has performed this experiment and reported on it in the forum. Of course at present it is speculation on whether the reduced feedback is responsible for some of the effects of whether the additional loading on the “ugs” output may also be in some way responsible. I’d be interested in a similar experiment that varies the diff pair source resistors instead (to reduce the amount of degeneration). Discovering whether the gain in practice is the same as predicted will also be illuminating but I take your point about it requiring time.

BTW, I agree with you about it not being a good idea to reduce the current capabilities of the mirror although I don’t know what value you have set it to. I calculate you have a current gain of about 6 so I guess the “ugs” output current will be in the range of 25 to 30mA depending on the bias of the diff pair. Most people seem to use around 5mA per FET for this with the 2SK389.

Ian.
 
Ian Macmillan said:
Hi Cheff,
BTW, I agree with you about it not being a good idea to reduce the current capabilities of the mirror although I don’t know what value you have set it to. I calculate you have a current gain of about 6 so I guess the “ugs” output current will be in the range of 25 to 30mA depending on the bias of the diff pair. Most people seem to use around 5mA per FET for this with the 2SK389.
Ian.

Hi Ian,

I used the same bias as in the linestage ugs for the Jfets, i.e. around 3.5 mA for the BLs I had. I had to make compromises between bias and current mirror gain, and as this value worked perfectly (for me) in the linestage, I just applied it to the amp. Maybe it's an error, and I should have increased the bias...

The current mirrors output 19/20mA last time I measured. I just made a rough balance between input capacitance of the mosfets, the linearity and gain of the mirror BJTs, the dV/dt I could afford, the input stage, bias, etc... Perhaps it is not optimum, but it seems to perform the way I planned, as I do not have the feeling of a lack in speed. BTW, reducing the current mirror gain will not lead to a drastic reduction of the OL gain, as I don't feel comfortable reducing it by 2 (6dB) for current driving reasons...

For the source degeneration testing, I can't promise anything. I have to make some preliminary calculations to see how it impacts the whole front end, since the bias point will change.

I first want to test the "class A opamp biasing" ;) adding some SE operation with output resistors to supply. That's pretty easy, so that's the next test (along with the OL gain). Just gimme some time, I'm a bit busy for now :)

Cheers
 
CheffDeGaar said:


I first want to test the "class A opamp biasing" ;) adding some SE operation with output resistors to supply. That's pretty easy, so that's the next test (along with the OL gain). Just gimme some time, I'm a bit busy for now :)


Agreed, this will be very interesting. Take your time, there is no hurry.

I don't think there is anything wrong with your choice of current to drive the number of output MOSFETs you are using. I need a little more for my application as I am aiming for 100W in class A and will need more output devices.

I understand that the source degeneration is hard to change once built but I feel it may be a better way to loose OL gain than loading the output of the "ugs". Just a feeling...

Ian.
 
Cheff,

thanks for sharing your listening impressions between local and global feedback (and thus answering my question)!

I had an Ayre C5 here for a few weeks and compared this no global feedback CD-player against my OPamp based (and thus with very high global feedback) CD-transport+external DAC. Both have the same DAC chip, the PCM1792, but the Ayre has a discete JFET/BJT I/V converter with a very large power supply.

The sound difference between these two systems playing the exact same CD was very interesting. My CD-system had more slam (better transient handling), a more firm bass, more powerful treble, and a little wider and deeper soundstage, more details,
but still the Ayre was the more pleasant one to listen to. Especially the mid and treble was more pleasant. Maybe the Ayre did not present every detail that is on the CD, but maybe that is not was it is all about.



Regards,
Sigurd

CheffDeGaar said:
Hi all,

I spent some time listening and comparing the global/local feedback configuration, and guess what... ;)


Cheers,
 
Hi Ian,
Half way done, but busy and lazy... Enjoying music by the fireplace while winter settles in :)
I tested 100R resistors (230mA SE bias) and 50R ones (460mA bias).
The sound changed a bit, mostly with the 0A46 bias, gained some lush, but I felt some kind of haze in the sound.
It comes closer to my AX sound.
I didn't report yet, since I recently grabbed 22R/50W thick film resistors to test a 1A SE bias.
I still have to bolt them to a heatsink, and I think I'll be done at the end of the week.
I'll be able to compare directly to the 50R resistors, to get some fresh memories ;)

Cheers,
 
I am personally familiar with both the busy and lazy bit :) And winter is only barely settling in here so far.

Sounds like you have made some progress anyway and thanks for the 'interim' report. I must say that the results are not exactly what I was expecting as I can't see why adding some passive SE bais should have any kind of deliterious effect. Are you sure thay didn't upset something else such as the bias conditions? More lush I can understand.

Look forward to hearing about the 1A SE bias results as I would have thought this was more than enough. Will be interesting to see whether it adds any more negative contribution.

BTW, did you perform these tests with the original OL gain or with the new 1K loading?

Ian.
 
Ian Macmillan said:
I am personally familiar with both the busy and lazy bit :) And winter is only barely settling in here so far.
Well, first morning freezes, and I begin to spread fresh straw everywhere in the house, eat more, and fall asleep every now and then.
And my fur's never been so abundant. Tell my hormones it's too early ;)


Sounds like you have made some progress anyway and thanks for the 'interim' report. I must say that the results are not exactly what I was expecting as I can't see why adding some passive SE bais should have any kind of deliterious effect. Are you sure thay didn't upset something else such as the bias conditions? More lush I can understand.
Checked the mosfet bias, the output offset (needed a small correction for absolute, relative remained OK), front end conditions, et caetera... All was OK.
With 100R resistors, I couldn't tell a difference. Maybe my system is a limitation here.
As I said, 50R resistors changed the sound, and I felt it came closer to my AX's.
But the lush was "as usual" accompanied by some loss of details in the highs, some air was missing.
The reverberation was damped a bit, the space was tighter.
The warmer midrange had a richer and flattering aspect, but seems to be a bit "more in front" (sorry, hard to find the right english words), thus unbalancing whole scene.
Hard to describe...:xeye:


BTW, did you perform these tests with the original OL gain or with the new 1K loading?
I kept the new 1K loading I was happy with.

But once more, these are only my tiny impressions, and the overall sound remains superb.
Add to that a subconscious reluctance to build another full Class A heater or a new AX...
Welcome to the world of my subjectivity :)


Cheers,
 
Sounds like you already covered most of the bases.

I appreciate that is very difficult to describe how something sounds and even when things are heard first hand, the subjective element often confuses things. Thank you for trying though - I found it helpful.

I will probably be going down a similar route soon in my journey towards an Aleph-X v2 (basically similar to PassLabs AX100.5) so it will be interesting to see how my impressions compare. The general topology is very similar except that the current mirrors are more basic and the output stage will be biased more towards class A.

I also intend to play with front end degeneration - as long as I can get my hands on some V grade devices (I'm hopeful here).

Ian.
 
Very nice indeed :)

Pity I don't read French though so I can only look at the pictures. I'm also more than a little curious about where the casework was done - looks very professional.

My own amplifier is progressing, albeit slowly, but I doubt that it will look this good even when finished.

Ian.
 
Member
Joined 2006
Paid Member
The front "handles" are cnc i would guess, but i have a feeling we will see a lot of discussion of this wonderfull piece of engineering here in the following months.
Casework can always be, hifi2000, not as good looking but prety pro too.

right now i'm concentrating on finishing the preamp.

but i got 2SK and 2SJ in stock.....
paired mosfets arriving soon.....hehe.

exemplary Cheff, Flat, Aldo, Maousse, Angelilo, Idefixes, Gloubou, Psykok .....et al, redifines diy projects.



:cool:
 
Casework can always be, hifi2000, not as good looking but prety pro too.

Hello,

The Hifi2000's 5U 19 inch is the best for easy work.

An externally hosted image should be here but it was not working when we last tested it.


An externally hosted image should be here but it was not working when we last tested it.


An externally hosted image should be here but it was not working when we last tested it.


An externally hosted image should be here but it was not working when we last tested it.


An externally hosted image should be here but it was not working when we last tested it.


Regards,