19-Jul-2017, 21:26
After trying 'Roon AIR', then briefly 'old' AIR 3.04 at the weekend, I have found myself thinking (danger!!) rather a lot about the slightly unexpected result. As I posted at the time, 'Roon AIR' was rather good, and the difference to my normal HQPlayer/mR/Mutec was very much smaller than I was expecting.
So lets go back in time, I used to use AIR regularly, I wanted a better 'front end', I tried many things with the D800. The Aurender N10/W20 and others sounded great, but for reasons posted elsewhere in this forum I ended up with a microRendu and Mutec MC3+USB. Here is the thing though, thinking back I can recall trying the mR/Mutec combo vs AIR2, and although I found AIR to be very good, it was relatively easy to discern the sound quality improvements offered by the mR/Mutec combo.
So back to today, the difference between AIR and the mR/Mutec combo has shrunk dramatically, so what has changed? OK, maybe I've gone deaf or maybe I'm imagining it all, but lets assume not for now, I have some theories.
Firstly the easy one, many have reported that AIR3 sounds better than AIR2, so that is one small step closing the gap.
Then we have the fact that AIR is an asynchronous protocol, so the computer is sending packets of 'music' data to the Devialet, the Devialet buffers the data, and somewhere in the land of 'magic wire' and ADH the data is clocked by the Devialet. This differs a little the XLR/AES3 input of the Devialet used by the Mutec, where there is a clock signal in the AES3 feed, and the Devialet needs to 'extract' this signal. Here my knowledge of exactly how the Devialet does this is a little lacking, I'm not sure the Devialet relies 100% on the clock in the feed, and once talking to someone from Devialet I was advised the quality of the feed clock is not that important, subsequently I and others have found that the clock quality of the feed is vey important, but exactly what the relation between the feed clock and the Devialet clock actually is, I am not sure. (If anyone knows for sure, please let me know!)
Anyway, so much waffle above, here is another theory. I suspect the clock in the 'Pro' is higher performance than in the Expert. There is no reason why not in fact, audio clocks have been getting both better and cheaper recently, so you would expect the Pro clock to be better, later, tech. So having a better clock means that the 'asynchronous' performance of AIR will be better, and maybe the AES3 feed clock to Devialet clock extraction does not benefit as much as AIR can from the Pro improved clock, another little step closing the gap for AIR. In a related observation, I and others have posted subjective observations that the USB input on the Pro is relatively better than that in the Expert. Other have stated that the USB board is unchanged in the Pro. Maybe both are true, and the observed improvement in the USB input is due to improved asynchronous clocking. Personally I have observed that bass accuracy in the Pro via USB is improved, a classic result of improved clocking perhaps?
The final theory for now (I might just think of another one later!), is the benefit of AIR integration into Roon. The 'virtual soundcard' part of the process is removed, Roon can engineer the their software to talk directly to Devialet AIR protocol. I do not know enough about this process to state exactly why this should be better, but I am speculating that it is at least possible that this can offer an improvement. Maybe Roon have made other tweaks related to clocks, timing in the computer OS or something else, who knows. Surly there is some scope for optimisation here? Anyway, all of this offers yet another small step in the sound quality of AIR versus the rivals like the Mutec.
If you then take all these little steps and add them together, one Saturday morning some bod tries 'Roon AIR' versus something else, thinking the differences will be fairly easy to spot, but those little steps have added up, and suddenly those easy to spot differences have become very tiny little things that are surprisingly hard to find.
I will be honest, there is much speculation and subjective observation in the above, but I am entirely serious in what I am trying to say and would be interested if anyone can offer any thoughts, observations or technical know how to add to, or indeed debunk, the above.
So lets go back in time, I used to use AIR regularly, I wanted a better 'front end', I tried many things with the D800. The Aurender N10/W20 and others sounded great, but for reasons posted elsewhere in this forum I ended up with a microRendu and Mutec MC3+USB. Here is the thing though, thinking back I can recall trying the mR/Mutec combo vs AIR2, and although I found AIR to be very good, it was relatively easy to discern the sound quality improvements offered by the mR/Mutec combo.
So back to today, the difference between AIR and the mR/Mutec combo has shrunk dramatically, so what has changed? OK, maybe I've gone deaf or maybe I'm imagining it all, but lets assume not for now, I have some theories.
Firstly the easy one, many have reported that AIR3 sounds better than AIR2, so that is one small step closing the gap.
Then we have the fact that AIR is an asynchronous protocol, so the computer is sending packets of 'music' data to the Devialet, the Devialet buffers the data, and somewhere in the land of 'magic wire' and ADH the data is clocked by the Devialet. This differs a little the XLR/AES3 input of the Devialet used by the Mutec, where there is a clock signal in the AES3 feed, and the Devialet needs to 'extract' this signal. Here my knowledge of exactly how the Devialet does this is a little lacking, I'm not sure the Devialet relies 100% on the clock in the feed, and once talking to someone from Devialet I was advised the quality of the feed clock is not that important, subsequently I and others have found that the clock quality of the feed is vey important, but exactly what the relation between the feed clock and the Devialet clock actually is, I am not sure. (If anyone knows for sure, please let me know!)
Anyway, so much waffle above, here is another theory. I suspect the clock in the 'Pro' is higher performance than in the Expert. There is no reason why not in fact, audio clocks have been getting both better and cheaper recently, so you would expect the Pro clock to be better, later, tech. So having a better clock means that the 'asynchronous' performance of AIR will be better, and maybe the AES3 feed clock to Devialet clock extraction does not benefit as much as AIR can from the Pro improved clock, another little step closing the gap for AIR. In a related observation, I and others have posted subjective observations that the USB input on the Pro is relatively better than that in the Expert. Other have stated that the USB board is unchanged in the Pro. Maybe both are true, and the observed improvement in the USB input is due to improved asynchronous clocking. Personally I have observed that bass accuracy in the Pro via USB is improved, a classic result of improved clocking perhaps?
The final theory for now (I might just think of another one later!), is the benefit of AIR integration into Roon. The 'virtual soundcard' part of the process is removed, Roon can engineer the their software to talk directly to Devialet AIR protocol. I do not know enough about this process to state exactly why this should be better, but I am speculating that it is at least possible that this can offer an improvement. Maybe Roon have made other tweaks related to clocks, timing in the computer OS or something else, who knows. Surly there is some scope for optimisation here? Anyway, all of this offers yet another small step in the sound quality of AIR versus the rivals like the Mutec.
If you then take all these little steps and add them together, one Saturday morning some bod tries 'Roon AIR' versus something else, thinking the differences will be fairly easy to spot, but those little steps have added up, and suddenly those easy to spot differences have become very tiny little things that are surprisingly hard to find.
I will be honest, there is much speculation and subjective observation in the above, but I am entirely serious in what I am trying to say and would be interested if anyone can offer any thoughts, observations or technical know how to add to, or indeed debunk, the above.
