Weird. If anyone can send me a link I can get it fixed. I’ve looked around the website with no luck. You guys are better at finding stuff than I!
In any case, no, the internals on the new PMG are a whole new way of dealing with the data and I am super excited for folks to get their hands on. It.
Well, good question. It is, in the same way as the MKII FPGA, at its core, a fancy upsampler and Sigma Delta converter. I assume most of you remember that’s what the MKII FPGA did: takes in PCM or DSD, upsamples it (in the case of the MKII) to 50 bit PCM and then runs that through a SDM to output DSD256. In the PureStream it is the same basic architecture but with a few important differences. Instead of generating a longer PCM "word” (which the MKII did because that’s how the volume control worked without loss), we skip all that processing and go right to a double speed upscaler that is running at double speed, 22mHz, and then into the SDM, outputting DSD512.
There’s a lot more nuance to it, but maybe it’s simplest to say that because we have taken a different course with the PMG—installing an analog stepped attenuator at the output instead of digital manipulation of the bits to achieve the various levels desired—we get a cleaner, lower noise signal.
Also, another feature of the PureStream is isolation and reclocking. When you look inside the PMG you’ll see the PureStream engine is housed in its own metal shield instead of sitting out in the open. This module approach helps maintain the necessary isolation between the galvanically isolated inputs and the last stage, the 64 step sequencer. Think of it as the PureStream module sitting all by itself with its own isolated power supply and isolated clock inside a quiet room.
The output of the PureStream Engine is also isolated so once it leaves its little quiet den, it’s in pretty beautiful shape and connects to the analog board where it is then recolcked again (that’s the Digital Lens part). At this point it is really clean, pure DSD running at 22mHz (DSD512).
BTW, on the PMG front panel touch screen, you can select the speed: DSD512, DSD256, and DSD 128. Each sound different and that’s where you can adjust sonics for synergy in your system. I prefer 512, but Brunhaver, in his system, goes with 256. In any case, you have a choice to fool around with. Also, there is a choice in the sequencer section two: Round Robin or or full 64 step sequencing: each sounding a bit different too. I prefer the RR setting as in my system it gives an amazing improvement to layering of the image. On the FR30s, it’s just orgasmic when you click that in. Stunning, actually (though the S/N ratio gets a bit worse by a few dB but who cares?).
Hope this gives you some insight.
@Paul I noticed the text that mentioned ESS when I copied the PMG DAC URL and pasted it into my email client. When the email client generates the little preview of the page, that text came up. this happened back on release day.
I just tried it again and that text was no longer there in the preview. Maybe it’s been fixed.
Huh. Weirdness. Well, thanks for letting me know. Much appreciated.
Can you get the designer to really go into the detail? A Youtube video?
Must have been early “copy”. I can’t find this page on the site.
FWIW.
Cheers
It would indeed be very helpful for prospective buyers if we could get more detail on the design. Maybe the new designer could share some info here like Ted used to on his DS DACs. 
It seems like between Pauls comments above and the website, there is quite a bit of data available… not sure what would be missing from a NPI point of view.
Hopefully a few reviews forthcoming…
Well you have a disgruntled customer on this forum with very efficient speakers. Lower noise and distortion of his DirectStream DAC would eliminate his audible noise.
Lower noise and distortion is a good thing in many scenarios
I confirm
Great details and congrats for the product. I see you revised your early spec sheet, that stated THD+N of 0.1%. But the revised number (0.002%) also does not relate to a SINAD better then -100dB (in fact, it relates to a SINAD of about -94dB). It would be good to double check.
As for the architecture, this reminds me a lot the dCS approach. Upsample the signal (dcs goes up to 128, and in their top lines using a separate box, called the upsampler, that is fpga based). The signal is then fed to a thermometer coded resistor dac (to avoid the shortcomings of pure R2R), with some mapping options to the user, regarding the assignment of the bits to the resistors.m, that also feels kind of similar to your RR or step sequencer.
If this is right, Inwould have to say dCS is a proven approach and if you could improve on this, this should be a spectacular product.
Less noise is always good. My new Marantz Link 10n is said to have very good measured values. The Ansus sortz and Phonosophie noise suppression products, as well as the Grid Protector, the Furutech Vault power strip, and the Furutech Projekt V1 power cables, are all very quiet.
Do not forget the NA muon pro Actually.
The Audioquest Dragon Source (McDonald’s😂 ) cables also have a quiet background. Compare the measured values of the Marantz Link 10n. Marantz claims they’ve never produced a component with better values.
Happy cake day, Andreas! 
After the mk1 experience, people bought mk2 expecting a long update path. I don’t think it’s right to cut it off suddenly. However, since this is the result, psaudio should open the entire library to users. People should be able to choose their own version. The decision of which one sounds better is a personal decision, and it also varies from system to system. You owe people at least that.
Don’t understand what you are referring to?
Care to clarify?
I believe he is talking about the various compiles of each release load.
all the versions are available under Firmware Archive I believe…they’re all listed
You say we chose one of the 50 versions, what I mean is the remaining 49 versions.
Understood.
I thought the latest firmware only had 20 candidates.