Phasure NOS2

[acg]

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.

Peter, so you are going SE amps now.  How does this affect the measurements?  The theoretical THD+N for the differential amps was less than 0.00003% and some premature measurements above showed values about a factor of 10 above that for the amps in differential mode.  Now this change to SE.  Are we going to get much more distortion, respectively speaking, at these already vanishingly low levels?

What really interests me (as you know) are small signals (say 1W) into high impedances (say 16ohm) because that is what will be in my speakers.  Most of the time I will be using well less than 1W.  There are available reasonably priced SS amps with 1W THD+N of 0.0003% which rises to 0.01% at 0.5W.  SS amps (such as yours) seem the be characterized by rising THD+N as the amplifier output (Watts) falls.  I think that it will be very interesting to see how your new amplifier behave at less than 1W.

Regards,

Anthony

EDIT:  Afterall, your top two channels are 119dB sensitive, so you will be using a fraction of a watt especially if you remove the passive crossover as seems to be the logical step with the NOS2.

[PeterSt]

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.

I don't know where that 115dB came from, but that's my old speakers.

EDIT:  Afterall, your top two channels are 119dB sensitive

Don't know where that came from either.
It really is 118dB.
But Ok, who cares.

[PeterSt]

Anthony,

Below you see that same amp but now into 50K Ohm and something like +/- 1.2V (so p-p 2.4V). This is the SE (Single Ended) situation for the poweramp. Also, what I measure here is the whole chain in SE (and output is 6dB less indeed).

What you see (is actually the situation from the picture in my post from 2 days ago) is the normal output of the NOS1a but a bit better (worst harmonic is at -116 dB or so) but with the notice that the output is 6dB less and this even weighs in for THD+N (the higher the signal while the harmonics stay at the same level, the better THD+N will be).

Maybe it is hard to follow, but this is thus the exact other side of the "chain" to test as this is no load at all. However, there's infinite current available were it for headroom. In this built (config) case this is 750mA while normally we use 80mA for our NOS1(a) (and then usually into 50-100K).

So what I found, unexpectedly, is that this matters a LOT. It is a bit of the same thing as people calling you names when you come up with a 6GB bandwidth long distance video solution and name it Blaxius, which also matters a lot. Do the math with capacitance and all (and the 80mA) and you'll come up with 20m+ of cable length. But make an even less capacitance cable and it helps vastly. What I did here was increase the current possibiliy and again it helps vastly.

What did I really say ? well, that the spades of headroom, also for the poweramp situation (which really is the very same) are always useful. Remember, headroom, which is different from taking 30V into 4 Ohm and then needing the current (and the dissipation and all).


About differential vs. SE and possible (2nd harmonic) distortion :
This is a longer story while I will try to keep it short - think about differential as things being in super balance and that doing all kind of jobs. That is nice but then first the opportunities for the super balance must be there and I don't see that for the power amp. Or at least not the way it is done here and how it can be done SE by means of the exact same principle. Fact is that this is still to find out for real because with this board I can't test the real SE situation. So secret : the picture from two days agso *and* the measurements you see below are from a differentially populated board, but measured from one (SE) side only. I listen to it SE as well because behind this is my normal power amp from the moment and this has SE input only.
The new board will allow for genuine SE population (2x) as well as genuine differential population. The difference emerges with how the input signal to it is connected and how the output signals are used.

So two examles from day before yesterday's post about the connections :

If I have a SE ladder because it is only half of the price from differential, how can I make a differential power amp out of that witout active (OpAmp) elements ? I can't. So this is how the power amp is allowed to be SE because what to do else (hey, for best SQ !).

If I have a differential (voltage !) ladder because that is supposed to be better for THD+N and I like to spend the $, how to make that work with a SE power amp and no active elements (like a summing OpAmp) ? I can't. So or I now use a differential poweramp as well, or I use only half of the differential ladder.
Now laugh : the ladder is so complicated and is so much set up differentially, that that really *will* matter for THD and all when you use only half of it. This is totally crazy stuff. So you use half of it, throw thus out half of the power, but it is double because it measures better (and may sound better) and this is because of all the really balancing out stuff in there (say what's done in plus is counteracted in minus at the same time to eliminate current spikes).

And in the end all is just possible thus without active elements outside of the power amp itself.

Peter

[dcgold]

Hi Peter,

This looks like such an incredible ambitious project. Having the ability to DSP X-over and tailor the room acoustics via DSP is the golden key to audio, takes the room out of the equation. When you have a turn-key system and want to show in the US again, let me know, would love to work with you again - we are working on a promising new motor design, better efficiency, lower inductance & distortion and new VC design.

Laurie LOVED your NOS-1 and cannot wait to hear your complete system.

I do not know if it would help with DSP math, but know someone in Norway that has written some DSP for Windows that might be a marriage for your system. If you are interested in contacting, drop me a line via e-mail and I can share.

Best to you and family,

David
 

[PeterSt]

So, nothing happened eh ...

High Resolution link Phasure NOS2 01

The loose boards in the left are the Linear Power Supply (the part of it which goes in the external case) and the new design of the Power Amp (black). The Power Amp is optional per (max 72W RMS) channel.

What you see on the right hand side is a 50 pieces of "bit modules" (they are all connected together per the production means). They connect to the sides of the ladder boards, in parallel.
You see 8 ladders here, for 8 channels. 56 of the bit modules go on one ladder (2x 28 bits).

High Resolution link Phasure NOS2 02

High Resolution link Phasure NOS2 03

But there is a small problem ...

[acg]

But there is a small problem ...

Need a bigger case?

Looks rather good by the way.  Thanks for the update.

[PeterSt]

Below you can see how the bit modules are supposed to attach.
Don't look at the values or "untightness". This is just experimenting with the attachment means and how/whether things fit as intended.
They do.

So here you see three bits (one module for each) with only the "through hole" resistors soldered. You can see that the size of these resistors is ~ 2mm too large to fit really, but this is intentional. So now this size fits, but a smaller size fits just the same.

Under the resistors you see pads for SMD resistors. So instead of "through hole", the ladder resistors (and it is all about that, SQ wise) it can also be SMD.

And thus we can choose the resistor we like, of the sound quality we like.

The ladder module (the long board) can hold 28 bit modules. Or less - and again it is about what we like or what is useful. This is of course all to find out.

Peter

[PeterSt]

The ladder module (the long board) can hold 28 bit modules.

And that is the problem ...
It should have been 56.

So I have a design for SE (Single Ended) ladders and I have a design for Differential ladders. The decision for which is a bit (haha) crazy and is all related to the output. For example, nice to have a diferential power amp, but with a SE source something has to make that "differential" (make two signals out of one). Or, nice to have a SE poweramp, but what to do with the differential source (combine two signals into one which is veeeeery expensive, because a waste).

So maybe 8 weeks ago the idea was "SE" again, and that's just in a folder on disk somewhere, in the midtst of all the other stuff.

Then the whole boatload was given to China for production and over there they did not understand some file. So I adjusted a few things and resent the file. Again they did not understand so I sent the original again but now with an additional description. Ah, now all was fine.
But the last time I picked the wrong folder ...

I only found out yesterday after everything arrived.

So yesterday I put the ladder board to an order again, with again the original file.
Most probably I will now get a complaint again, have to change stuff and ... hopefully I will not be sending the wrong files again.

So again two more weeks of waiting.

Peter

[acg]

And thus we can choose the resistor we like, of the sound quality we like.

This is important.  Those Vishay Dales are expensive little fellows especially in the 0.1% variant and at the number you will probably buy them they are like 40c each.  But there are a number of good low noise resistors to choose from, so it will be interesting to see what you come up with.  Var-Z Foil for about $15 each will make the dac very, very expensive, but perhaps they will be just what the doctor ordered!

[acg]

And thus we can choose the resistor we like, of the sound quality we like.

This is important.  Those Vishay Dales are expensive little fellows especially in the 0.1% variant and at the number you will probably buy them they are like 40c each.  But there are a number of good low noise resistors to choose from, so it will be interesting to see what you come up with.  Var-Z Foil for about $15 each will make the dac very, very expensive, but perhaps they will be just what the doctor ordered!

I know it is off topic but a similar things happened to me this week.  I sent some gerbers to China for some pcb's for the ATX Linear psu project and about 30 minutes after I got the email that they had been put into production I realised that I should have added a snubber in one of the power supply boards...oh well...at this stage I just want to get the thing working...I can fine-tune it later.

[PeterSt]

Anthony,

Foil for about $15 each will make the dac very, very expensive, but perhaps they will be just what the doctor ordered!

Yes. And in the past month they became 30% more expensive (for us - EUR-USD).

And you know what the biggest problem is ?
Someone has to test this and see whether it's worth it.
Not even me is going to do that.

Regards,
Peter

[acg]

Anthony,

Foil for about $15 each will make the dac very, very expensive, but perhaps they will be just what the doctor ordered!

Yes. And in the past month they became 30% more expensive (for us - EUR-USD).

And you know what the biggest problem is ?
Someone has to test this and see whether it's worth it.
Not even me is going to do that.

Regards,
Peter

28 bits x 3 resistors per bit (from photo - not sure if this is true) x $15 = $1260 per board.

[PeterSt]

Times 2 for differential (my own Gerber issue), times 2 for 2 channels.

How many channels do you like to have, you said ??

[acg]

Yowser!

[PeterSt]

Whhoow ... I just finished ordering the parts for a complete 2 channel NOS2.

It always takes the most time to source the parts (where to find them / where to really obtain them from (is also $$$ related)). This time it took me over 3 full days.

A total of 120 order lines, spread over 4 suppliers.
This means 120 different parts.

If I count right, it is 2886 parts, excluding the cabinet, wiring, input/output terminals, switches, a display, remote and possibly more unimportant stuff.
And those 2886 parts all need to be soldered.

See you in a year ?



Peter