Phasure NOS2

[acg]

I was going to pester you again for more information Peter but you beat me to it!

[dcgold]

Hello Peter,

Long time, no talk. Say hi to the family, always welcome for a visit to So CA. Sounds like all is well.

Have you considered a silicone oil for cooling your amplifier? Along with heat-sinks around the edge, heat would flow through convection and even heat / cooloing to all parts.
Great to hear you are onto another great piece of gear - cannot wait to hear. Laurie & I loved the NOS1 at the show.

David

[PeterSt]

Hi there David ! Hope all is well with you all. And with the cooking of course.

Thanks for the tip; most probably this will not be needed, but if it is ...

First let's get this thing to work as I want. And, closer already ...

Regards !
Peter

[Scroobius]

Hi Peter - just out of interest will NOS2 DSD playback have the (obligatory?) low pass filter to remove all that out of band ultrasonic noise that is a feature of DSD files. Or have you done something clever to avoid it?

[PeterSt]

Hi Paul,

The latter. There won't be any ultrasonic noise to begin with ...

Peter

[acg]

Have you anything more to show us Peter?  Are the amps operational? 

[Scroobius]

There won't be any ultrasonic noise to begin with

Nice, very nice!!!!

[PeterSt]

Have you anything more to show us Peter?  Are the amps operational? 

Anthony,

It goes in bits and pieces with not too much time for it. I must give priority to XXHighEnd and if one issue in that is solved a next pops up. But at least a beta went out last weekend so that must be getting close. But the more of that, the less time for other stuff.

Still every day a few things are attempted on the power amp. Point is : something is wrong and for the life of me, so far, I can't find what it is. So it's a process of flying in different chips, rewire them to the existing PCB, try again, see what might be wrong with *that*, sit back and think and ... all sorts of iterations. I *still* think it is mainly a measurement issue but hard to prove.

...

The above is what I typed 2 hours ago. Meanwhile I again was ordering some other parts (or more of them for new trials) and then I saw a mistake in the initial order for these parts ...
yep
A 6.3V blocking capacitor in the supply which should be 16V.
And since I always see problems which lead to something wrong with the gnd this has to be it.
The correct (smd) part should be here Monday (now it is Friday). So a bit of waiting again ...

Peter

[PeterSt]

It is still premature but at least I now have something I aimed for.

What you see below is without any power supply regulation (that is the part which momentary fails actually); additionally I use the worst power supply imaginable (high ripple) and still this (THD+N of 0.00079%) is close to the analyser limit.

Input signal is 3Vp-p which will be the ladder output and the output signal is 8.4VRMS (24Vp-p).

This is with 37.5 Ohm load and after finishing the board this afternoon and sticking on some heatsink it should do 4 Ohm (loudspeaker) load. If the figures are still the same this little part is finished ...

not.

Not really because next up is something which allows a deeper interpretation of the figures because of hitting the analyser limit for the THD+N number. Maybe it can work visually (like see dropping the noise line and harmonics) or maybe I can do something to the analyser squeezing out more.
Or otherwise apply tweaks to the board so the "error" comes forward better (but I don't like this much because actually this implies error and I am not really used to such analysis).

Point of course is, that while this is without any good regulation, how can I check for the regulation to work out for the better if the figures stay the same.

Lastly, when this works for nice measurement figures it is still not a finished job because I am using a chip in there which I actually don't like for sound. So you may remember that I talked about the 1950V/us slew rate this amplifier can do "for a part" but for the net result this is really related / dependent on another ship and the one in there now is 50V/us, but is is low noise (so this is how the measurement looks fine). And now it gets more complicated becase the chip I have in mind  exhibits more noise BUT is highly dependent on the power supply because its noise rejection is worse. And the power supply can not be done as should because that requires a new PCB first.
Got it ? haha
Anyway with that chip the THD+N figure is 0.00110% at this moment and although also quite good for a power amp it is not what I had in mind. On the other hand all may be better when the board is fully populated so that too is to be tested. But still quite moot because it just needs a fine power regulation first.

Summarized it is all a bit of iterations and (much) throughput time but at least it looks (very) good for proceeding.

Peter

[PeterSt]

Okay ...

First plot shows 11.25VRMS (31Vp-p) into 6.5 Ohm.
Second shows 8.4VRMS (24Vp-p) into (speaker driver realistic) 3.9 Ohm.

The latter just exceeds the power supply which is 6A so there it starts to go wrong because of that, and no power resistors at hand for a little less load. So this is how the first plot shows the 31Vp-p which implies 4.8A which is still fine for the power supply.

This is going to work ...

[christoffe]

Here we can hear what is possible with an amazing amplifier.

http://www.avshowrooms.com/Audionet_Amplifiers.html

I don’t hear this performance in my room.

Joachim

[PeterSt]

And for later reference perhaps, this is how the noise shows up to 96KHz (into 3.9 Ohm). Only a bit of 50Hz but the wires are all over the place.

So THIS is how I like it for a power amp.

[PeterSt]

All,

In between whatever jobs, a small report about the power amp testing.

What you see in the picture below, is the means of testing the amplifier audibly while using a "best reference" (which is my NOS1a);

I did something naughty as I removed the normal gain stage of the NOS1a and let the poweramp now act as gain stage instead. And if now the sound is worse than it was, something is not right to begin with.

At some stage I gave up ...
But what you see below just passed its 4th day of listening and I now reckon that I leave my NOS1a be like that (haha).

I didn't look really, but I estimate that this "little part" consumed something like 5-6 weeks of throughput time, but of course 2 or so lost because of the regulation not working out. Otherwise I can see that the last post was on May 8 and this was about testing the power (current) and since then I've only be changing configurations and listen. Reject, change, listen again. Learn from the both situations, set up something else and check for an hour or 5 of listening. Still good ? then extend more towards the intended end situation. Still all fine ? ehm, no. Back to base and start over again with another config.

And so it is now May 23 (still 2015) and all what I see is that when I extend the configuration further, sound is going to be better again. But I don't like to be spoiled so I just won't.
Yes, I am even serious.

I now wonder whether people ever heard serious slam together with super accuracy; I mean, the slam is not so difiicult but all the failure came from the accuracy going away and things really sounding like a drag. But now ?
Again I don't recognize any of the (dozens) of albums I played.

So it took a while, but this is the first real success. Now a new PCB is going to be ordered, with a new regulation which hopefully does work; dimensions have been changed as well, to even better fit the 8 channel cabinet but we'll see about that later.

Oh, almost forgot : The set up is not differential any more per se. I thought that in the end this is a waste of space and money because all what the differential really does is providing more voltage, while this will only be necessary for channels which require high power, or speakers which are low efficiency. So, one board now actualy carries two single ended normal amplifiers, but it can be used as a "bridged" amplifier just the same.

What I could do now finally as well - and I just did this morning - was write out (on 7 pages) how the connections are to be from the ladders to the outputs or poweramps (and then outputs). What I now could reason out is how no single active element is needed to be in there in any situation. Of course the power amps themselves are active elements, so disregard those.
The focus is now also more on Single Ended with the same "quality" which is very very crucial because it saves about half of the money. So all can still be differentially setup (like the power amp outlay above) but when it is not necessary for the connection I tend to go the route of having the whole chain single ended. Thus for example, when for certain channels the power (amplification) requirement is not high enough to bridge the amps on one board, the ladder feeding it will also be "one".

The design for this all is ready for a longer while, but as I saw myself this morning, it still needs 7 pages of reasoning about how the flexibility in all is to be utilized best. Btw similar to me spending 4 weeks (throughput) on trying different configurations of the power amp board. In other words, nice that I made all so flexible, but it really needs some stiff elementary thinking on how all now actually will or should work. Practice example ?

Envision you order a NOS2. So what do you want ?
- Two channels with ~60W RMS into 8 ohms.
- Two channels with probably 10W or so into 4 Ohms.
- Two channels with XLR output (and RCA/BNC) for the nice tweeter stereo amp laying around.
- One channel for the subwoofer which is active and thus RCA.

This looks easy, but it won't be. This is because all costs money. So who determines the 60W of the first line ?
Ah, we do that so we are on te safe side. Very nice. But those two channels will be two times more expensive if 30W also is fine (actually this now comes down to 15W but alas).

Regards,
Peter

[PeterSt]

Nice gag :

Because I don't know much of power amplification as such for practice, and I mean how the power required is to be mapped on to speaker situations while I actually never wondered about the real work out ...

I made a few mistakes. Or mistakes ... maybe that's a too string description, but look :

What I did was relating it a bit to what I use at the moment for power amplification and this is 30W GainClones. I always say 30W while they are over 50, but say that this is related to the better specs of them and how they must be used then.

Without really paying attention, I started to measure as you can see in posts above. But what do I measure ? ehm, 31V into 4 Ohms = 7.75A, which 31V I had to drop a bit because of the transformers I had laying around not being capable. Besides that 7A is the max for good specs - which is theoretical. From this follows 28Vp-p. Beautiful.
Also beautiful is that this is 196W and who cares (98W into 8 Ohm).

But the big mistake I made is that I never saw that this is continuous while in pratice this is never ever used. Still I measured it like it to be continuous. In other words - and this is the crux :
With this continuous power the figures are what you saw above. Thus completely crazy. Outrageous actually. And nothing runs really hot either (with the heatsink on I think I saw 50 celcius or so after a minute or 10 of this power).

The peak power will be over twice as much, but then THD will vastly increase. On the other hand it seems impossible that this power is needed by anyone in normal circumstances.
So point is : what I did was anticipating peak power for normal situations and took my 30W as a base, while without really realizing it I turned that into a theoretical 72W which ends up at 98W (8 Ohm) but which is not peak but continuous RMS.

Now you'll understand better why the power amp boards are more focused on being single ended now and how one board can carry two amps, and how for instance me with my 115dB sensitive speakers will only use SE amps for sure.
The greatest gag is almost that all was so much squeezed and squeezed and squeezed to let fit in 8 power amp channels (quite easily in the end), that now this is 16 ...
ahum
But it would require 16 ladders as well and I didn't even start looking at that - and I won't go that route because where to put al the output terminals (for 8 channels it is 40 already).

Peter

[acg]

And for later reference perhaps, this is how the noise shows up to 96KHz (into 3.9 Ohm). Only a bit of 50Hz but the wires are all over the place.

So THIS is how I like it for a power amp.

Peter,

Would you expect noise to change when pushing say a 16ohm load?  Would be interesting to see the corresponding graph.

Anthony