27bit DAC -> 162 dB dynamics...

This DAC exists for the same reasons why multipath ADC exist and are successful (like STAGETEC's TrueMatch, 158dB of dynamic range for their Mic ADC) : No need to ever worry about headroom vs noise anymore, no trade-offs, no operator errors wrt gain structure anymore.
That might be irrelevant in a simple home stereo final consumer setting, but in a pro environment this can become relevant quickly.

BTW, multipath ADC are mainstream these days and found in audio gear like digital guitar preamps, field recorders etc.
I see no reason why (affordable) multipath DACs won't gain traction as well.

For 16bit home audio playback it's overkill, sure, but home audio is just one (small) market segment for audio gear.
 
  • Like
Reactions: TNT
This DAC exists for the same reasons why multipath ADC exist and are successful

Not sure if TI's approach counts, as it makes do with a single DAC and a PGA, but TI has multiple ADC chips with that.

No need to ever worry about headroom vs noise anymore, no trade-offs, no operator errors wrt gain structure anymore.

So we remove the most basic sound engineers skill, correct level setting. Which with a 20...24 ADC is really not that hard.

I routinely use digital mixers and never found setting the input level a problem

Setting the EQ by ringing out the stage and Mic's and then gainriding to bring out the solo performer or perhaps to move the bass forward in the mix, because it's a song where the bass drives everything, now that useful.

But a 160dB ADC doesn't help neither does AI.

So even in a professional context I remain unconvinced.

Thor
 
The DAC is a real product and the Mr. L Grou is a serious designer. This thing works as advertised, I'm 100% sure. This is pro stuff, not audiophile BS.
John is the indeed the real deal and he has spent ten years working on this. I remember visiting him in his garage over thirty years ago when he was building his first microphone pre amplifiers. He was very carefully measuring and matching parts on his AP back then.
 
As others have stated, I also wonder how they managed to design and, more importantly, build an adequate power supply to those claimed specifications.
Even with multiple DAC paths and specific DAC path-tailored bit-shifts, there are crucial system blocks that operate in the analog domain and share power supply rails and ground planes with the rest of the (additional) noise-generating system blocks. PCB layout and noise isolation between these system blocks would need to be something out of this world, which, judging by the overall sentiment, the designer is capable of designing.... Or, is it all the marketing BS that is messing with common logic.... and the laws of physics.

Here's the system diagram, courtesy of Jan Didden of "AudioXpress":
https://audioxpress.com/news/imersi...analog-converter-completes-32-bit-audio-chain

I circled in red the two system blocks that would have to exhibit extraordinary precision, and be supplied with power supply rails of extraordinary noise performance (mind you: at room temperatures, using conventional parts), while existing on separate ground plane(s) from the rest of the noisy digital system blocks:


1746414643214.png
 
Last edited:
  • Like
Reactions: jan.didden
CS43131 and CS43198 use some secret algorithm to enhance the DR. This has been reported to cause audible and measurable issues (e.g. https://www.audiosciencereview.com/...gh-unbalanced-output-power.62024/post-2275689).
That was an interesting read! I bet we would see the same behavior in a proper review of the Millennia. Hysteresis and time-constants are a way to get around not having ultra precise matching of the gains of the two DACs.

I think it’s pretty obvious that they can’t do switching between the two on every sample that crosses the threshold, there has to be some delay and lookahead involved.
 
If people pay those crazy prices, then equipment priced that way will keep coming. Nice thing about that pricing is that the seller has to sell only a few of them to break even, given that the cost of boxing one up for sale is far lower than the asking price.
 
My diy dac which probably costs around $2,500 to build, if sold commercially would probably have to sell for $10,000 or more. That's a perfectly normal incremental build cost to retail price markup ratio. Typical ratio for consumer items its 3 to 5 times. For very low production volume hi-fi devices, the markup might have to be 6x or more. That's just to stay in business, not to get rich quick. 60% to manufacturer for all costs including amortized engineering costs, 10% for distributors who have to pay for regional marketing and first-line warranty service, as well as for recruiting retailers, and 30% retail markup for costs including leasing a storefront, etc.

I would just add that these are luxury items for people who can afford them. Often a case with a thick custom machined aluminum front panel is expected by customers who won't buy if the item doesn't both sound great and look great. The case can be the most expensive part of the manufacturing cost. Next most expensive cost is often the custom power transformer. And so it goes.
 
Last edited:
This thing does neither have 162dB Signal-To-Noise Ratio, nor 28bit actual resolution nor does it have lower distortion. It "only" offers an intelligent increase of dynamic range to allow to reproduce very small signal levels more accurately...
Please see posts in this thread by @KSTR. Its not about music with silly dynamic range. Its about reproducing low level signals more accurately. And, yes, the effect should be audible on a good enough system.
 
Please see posts in this thread by @KSTR. Its not about music with silly dynamic range. Its about reproducing low level signals more accurately. And, yes, the effect should be audible on a good enough system.
Mark,

Yes, this Imersiv Dac reproduces signals more accurately at low levels, but is that the reason why it seems to sounds so good ?

You described the Firdac you are working on as one of the best you ever heard, and I believe you.
But your Dac with a dynamic range in the order of 110 dB has a much worse resolution at low levels.

So one should avoid to jump to conclusions regarding low level resolution.

Hans
 
That was an interesting read! I bet we would see the same behavior in a proper review of the Millennia. Hysteresis and time-constants are a way to get around not having ultra precise matching of the gains of the two DACs.

I think it’s pretty obvious that they can’t do switching between the two on every sample that crosses the threshold, there has to be some delay and lookahead involved.
I am still waiting for a proper measurement of THD vs generator level showing no weird things at the switching point...
 
  • Like
Reactions: TNT
There is no real switching point as such. According to the patent it’s a gradual fading in and fading out between the two Dac’s of a many samples delayed signal while a separate ADC is looking at the non delayed signal what’s coming to alert the fading process in time.

Hans
 
  • Like
Reactions: mvs0
...is that the reason why it seems to sounds so good ?
Not sure, but probably that's not all there is to it. It may sound good in part from trying to make it as low noise as possible. Not just noise floor, but also audio-signal-correlated noise as seen in FFT spectral line skirts.

Since some of the effects of some types of noise are hard to measure (e.g. to separate AN from PN -- amplitude noise from phase noise, as seen in noise skirts), it can be hard to say exactly.

I would defer having some opinion on it for now. Maybe a better idea if I ever get a chance to listen to one.