Phasure NOS1 vs. Pacific Microsonics Model Two

[Miska]

I'm curious which ESI card and driver supports this sample rate?

 

My guess is none, but at least one common trick is to use a dedicated I2S per DAC, thus one can get 384/32 for example. With a bit of FPGA it is possible to take it to 768/32 by combining four 192 kHz I2S channels into one.

 

I'm not enthusiastic about PCI express because its a serial interface with a serdes at each end and a not too synchronous communications unlike PCI

 

I don't see that as much of an issue, since PCI is a shared bus operating in transaction based burst mode. Both will need some FIFO on the board anyway.

 

Good side on serial systems is better tolerance for length of the bus at high clock speeds. Due to parallel nature, physical length of the PCI bus is limited by the timing requirements. If you look at size of a 64-bit PCI-X (not to be confused with PCI-Express) slot and maximum length for the max 533 MHz clock speed, it is not very practical...

 

 

[Miska]

A 32 bit dac is pure marketing, as the designer of one explained to me, thermal noise will prevent any meaningful output at bit depths much less than 32 bits.

 

Yes and no, the thing is more complex. It reduces need for lowering the precision between DAC and prior processing stages. It also gives more flexibility to deal with noise shaping inside DAC's modulator stages. So the digital domain performance can get better.

 

At least it removes all the "losing the bits" arguments against digital volume control for RedBook formats...

 

And if you look for example at datasheet of BB PCM1795 and the -144 dB and -150 dB output spectrum figures on page 14 you'll see that things are getting scaringly close to 24-bit at certain frequencies when FFT is long enough and there's enough averaging. Just zoom enough on the PDF.

 

So it depends how you look at things, since I'm pretty much specialized in software digging out weak signals from noise I can assure you that there are ways to detect really weak non-random signals out of random noise. And any crypt-analyst could tell the same...

 

My favorite demonstration is a software signal generator mixing different types of noise and sine waves at varying levels and frequencies for people to listen at which levels the signal completely disappears in noise from human auditory system. Audacity can be also used for the purpose, generating noise and sine and mixing the two together.

 

 

[Miska]

I am sure PeterSt has spent many hours/days/months/years thinking up and designing this DAC, all of which should be factored into the 'Retail Sales Price'. Does gkoones consider this time and intellectual property to be totally without value and that we should all give our time for nothing?

 

Not exactly related, but the quite interesting part is that many tend to consider this only in scope of hardware, but not in scope of software. However, in many cases development time for software is much higher than for hardware.

 

It is good for everybody to calculate how many units of price X would have to be sold to cover several man years of development time...

 

 

[Miska]

In the case of hardware products that need drivers or firmware, separating the software development costs

 

No, I was talking about pure software products without related hardware.

 

Its just part of the fixed cost that has to be amortized over the sales volume.

 

It would be interesting to know what kind of volumes you think we are talking about here?

 

I dont know how long it takes to assemble NOS1, but for my own DCA1 it takes me roughly two full working days to put a single unit together, but well, it's not only a DAC.

 

Also component costs tend to heavily depend on volumes.

 

 

[Miska]

I still don't understand how Peter squeezes 384/32 through an I2S port that stops at 24/192 but then he is a magician.

 

At least how I've been doing it is to use one stereo I2S port as a mono port. Then having two I2S ports gives you 384/24. Four I2S ports at 192/24 gives you stereo 768/24.

 

 

 

[Miska]

Is this purely theoretical or have you managed to get this to work? With an ESI card?

 

It was just a generic statement. I've managed to get it work, but not with ESI. I don't even own any piece of the ESI card in question here.