So, this thread is a long time coming. Thought I'd make a place for people to lay out their thoughts on resampling of all kinds. I know Marv(ey) generally uses resampling all the time. Some people say it's the devil, some haven't tried, and some don't care.
Until this week, I was a combination of the last two, hadn't tried and didn't really care to. Having quite the bevy of DACs in the house at the moment (all Matrix) I decided to give it a try on the X-Sabre and both Mini-I units. FYI X-Sabre - ESS9018, Mini-I Pro - ESS9016, Mini-I - 2x AD1955. The Sabres are capable of all sample rates including DSD, the Analog Devices chip supports DSD64 (not 128), but is only allowed to operate in PCM mode here, up to 384kHz. To add to this, I decided to try all of these on my speaker rig via M-Stage preamp and the HD800, HD600, as well as UERM. Lots of combinations so this has been keeping me quite busy the last several days. To save time, I focused mostly on CD quality files at 44.1kHz and then used the maximum PCM multiple, or DSD rate for each DAC, so DSD128 for the Sabres and 24/352.8 (DXD) on the AD. All resampling was done on the fly by JRiver MC19. To make matters more complicated the Mini-I Pro has DSD filters at 50, 60, and 70kHz I had to test.
Several questions I had in mind (in no particular order): Did the Sabres end up with the same results? If so is it likely to be because they are Sabres? How does a speaker rig respond in comparison to the headphones? IEMs with their special features... how would that work? Angled drivers in HD800 vs flat drivers in HD600, different result? Would the sample rates sound different at all? Would the AD's sample rates (PCM to PCM) sound as different as the Sabres going from PCM to DSD? What am I expecting to hear different with the sample rates? What will I actually hear? Are Sabre DACs always going to sound better resampling to DSD? (something hinted at here on CS) Is resampling preference DAC specific, or does it transcend? Is the difference between 44.1 and the top end similar between DACs or do they change in different ways? Will the Mini-I Pro's DSD filters do anything?
Starting with my expectations, I thought I would hear basically no difference. I decided that if I heard a difference it was most likely to be not soundstage, tone, frequency balance, macro detail, or dynamics that was affected, but rather texture, micro dynamics and micro detail that would change the most. I thought those things would change to some degree across the board, but I expected voices to change the most as they are texture and micro detail rich sounds. For the most part I expected the tiniest things to change the most.
What I actually heard was that going from 44.1 to the top end on each DAC shared many aspects, and that the differences are fairly pronounced. The particulars of the sound that actually changed were soundstage most of all (completely wasn't listening for it at first), macro detail, followed by micro dynamics and texture in a dead heat with micro detail mostly unchanged. Thinking about it after the experience it made sense why two aspects I wasn't expecting to change ended up changing the most. I liken these aspects being resampled to blade sharpening. A kitchen knife will still cut a wedge of cheese, but a razor will slice through cleaner and easier due to it's sharpening being taken to another level. With macro detail this is easier to see, a sharper leading edge will cut through cleaner, with soundstage the metaphor doesn't work as well, but think about the sound being cut into chunks shot from the driver. With the sharper blade you would cut more and finer pieces in the same time as you would the more dull blade. This spreads the sound while still retaining a sense of cohesiveness.
This isn't all rainbows and adorable puppies, though. In my case, the resamples all tended to have too wide of soundstage on all of my headphones, but
it sounded great on my speaker rig. Quick note about my speaker rig - It's basically the LCD-2 of speakers. Somewhat bassy with incredible mids and too-relaxed highs with moderate detail and fairly narrow soundstage. Hence why I'm so interested in Marv's OB speakers. Back to DACs - in all cases, the resampled version was more dulled on macro detail and slightly dulled on texture, micro dynamics and micro detail. In the end I decided I didn't like resampling very much for headphones but it was great with my speakers, a completely opposite result from expectation. I thought the higher iteration would shine through on the more detailed headphones, and I thought I would need to retain the 'roughness' of native for my speakers to have enough bite to stay interesting. The speakers actually benefitted from the larger soundstage and managed to stay interesting with the dulled detail and the headphones merely received negatively on all fronts.
Pushing further I had to look into the DSD filter settings on the Mini-I Pro. These crossovers are at 50, 60, and 70kHz, which are well above audible range, so I absolutely expected no change, but in the back of my mind I remembered discussion here about noise shaping in DSD possibly affecting the audible range. In addition to the soundstage widening from going to DSD, going from 50k to 70k pushed it wider still and around my head as if I was being dipped nose-first into the music. This existed on both headphones and speakers. There was not a lot of change otherwise, so mostly spatial cues were affected. With the speakers I first loved 70k, it was enveloping and I could 'feel' vocals being sung at me, then over my shoulder, then to the side. It felt like I was very close to the stage at a rock concert, very invigorating. Flipping to 50k brought the vocals more into focus (less beaming around everywhere) and collapsed the soundstage a good deal. At that point I wasn't entirely happy with either and then I remembered I had 60k yet to test. This ended up being the best compromise, larger more enveloping soundstage, but with still well focused vocals that didn't sing past me.
Moving to headphones I expected the HD600s to benefit similarly to my speakers as they are non-angled and not that great on soundstage to begin with. They actually reacted the most violently to 70k of anything I used. It felt, as I moved from 50 to 60 to 70, that the singer was being implanted in my brain as I reached 70. It was quite uncomfortable with the background sounds still in the background with the singer in my brain. I easily preferred the 50kHz setting. The next most changed was the UERM where I again preferred 50kHz, but not as strongly. The one I expected to benefit was the HD800, but I once again preferred the 50kHz setting, but the differences between settings were fairly small. The HD800's already large soundstage shrugged off attempts to modify it.
OK, so what about the DACs themselves? The best headphone performance for each unit was: X-Sabre - native on all, Mini-I Pro: DSD - 50k on all, Mini-I: native on all. Now wait, I just said DSD was not ideal for headphones. Yes and no. The difference here is that the Mini-I Pro is a bit aggressive by nature, so smoothing that out a little with resampling resulted in a better sound. The best speaker performance for each unit was: X-Sabre - DSD, Mini-I Pro - DSD 60k, Mini-I - native. OK, so this is odd. 2 resamples and a native. On the X-Sabre the difference is really small, like placebo type doubt margin. The Mini-I Pro sounds significantly better on DSD 60k, while the Mini-I normal just lacked excitement when resampled to DXD. The Mini-I normal is somewhat rougher at native, but more involving.
Let's revisit some of my questions.
Did the Sabres have the same results? No.
If so, was it because they were Sabres? I can't say they couldn't have had the same results if the Mini-I wasn't as aggressive and required the DSD resa
mple to calm it down. I can't answer this one, despite the condition of the question not being met in the first place.
How does a speaker rig respond in comparison to the headphones? Completely differently, at least my rigs did.
IEMs with their special features... how would that work? Not a significant difference from other traditional headphones.
Angled drivers in HD800 vs flat drivers in HD600, different result? Not with resampling. The DSD filter in the Mini-I showed strong differences, though.
Would the sample rates sound different at all? Certainly, though the effects are still subtle. My comments here are to demonstrate differences, but remember they are small even if I indicate they are drastic. I can't help getting caught in the moment of description.
Would the AD's sample rates (PCM to PCM) sound as different as the Sabres going from PCM to DSD? The AD did sound quite different, but not as different as the Sabres did. I didn't test 44.1 to 352.8 on the Sabres (mostly because I have a bad Pro unit that can't play high rate PCM without stuttering, I'm told this is my-unit-specific and that no other Pros tested by several others can duplicate my problem).
Are Sabre DACs always going to sound better resampling to DSD? (something hinted at here on CS) With my sample size of two, I can't be certain. Based on these two Sabre DACs I find that it is likely the case, as in 3 of 4 combinations (DAC and driver type) I preferred the DSD version. The outlier was the X-Sabre with headphones where I preferred native PCM.
Is resampling preference DAC specific, or does it transcend? There may be some trends (like Sabres and DSD mentioned above), but I think it is mostly DAC (as a whole unit) specific. You should walk into each DAC evaluation expecting a different result from resampling.
Is the difference between 44.1 and the top end similar between DACs or do they change in different ways? The Sabres tended to change almost exactly the same. The AD DAC went ultra smooth and laidback with DXD, a similar effect to the Sabres, but where the Sabres retained some detail, the AD went full Lebenesque. The AD DAC is naturally a softer squishier sound, so it's fall to Lebenesque was hardly surprising.
Will the Mini-I Pro's DSD filters do anything? Amazingly, yes.
--
Thanks for reading. Please add your experiences and thoughts on resampling and DAC chip trends with resampling. I'm also interested to hear your experiences specific to headphones vs speakers with resampling.
