Up/Over sampling and file formats

I’ve been thinking about my future direction in terms of upgrades from my current PC with linear PS’s & NDS and what I might end up with. I currently store all my files as WAV’s, mostly 44.1Khz/16 bit. I use the BNC input on the NDS in preference to UPNP

On my PC Flac’s don’t sound quite as good but this is almost certainly a processing issue as I run the PC underclocked and on 1 CPU core. However one thing it does do is upsample to 192khz/24 bit which is the max the sound card can deal with. This, to my ears has always sounded better than the native resolution.

I’ve been experimenting today with comparing

  1. a Qobuz stream which it upsamples on the fly (Foobar with Sox),
  2. the same file locally stored, downloaded from Qobuz and upsampled on the fly,
  3. an upsampled (via DBpoweramp) and locally stored 192/24 bit Flac
  4. and similarly a 192/24 bit WAV locally stored on the PC.

To my ears I would place the order of preference as per the last paragraph. I wasn’t totally suprised by this order of preference but the size of the difference between upsampling on the fly and stored upsampling was a little suprising. The locally stored, upsampled WAV files sounded the best.

If I were to follow the logic of this it would mean my roughly 1.2TB collection of Wav’s would become either 2.4TB of 192/24 bit Flacs or 5.3TB of 192/24 WAV’s. This clearly has serious financial implications when I start looking at Melco’s, Innuos’s etc to upgrade (assuming they will be better…)

What I would value members opinions and experience on is in what formats do they store their music on their Melco’s, Innuos’s etc and does the difference between Flacs and Wav’s matter so much. Is there a native resolution and bit depth that I should target any conversion at before I start re-processing all my music collection

I have all my music in flac, and have it transcoded into WAV on the fly for playback. All CD rips are in 44/16 obviously, and downloads in whatever resolution they came in. I’ve tried upsampling but it makes the music sound lifeless to me. I’ve never understood how it can add something that isn’t there in the first place.

It may be worth trying upnp again, you can then control it from the Naim app and untether yourself from the computer. Put the files on a nas and you can leave the computer out of it.

I’m surprised you haven’t mentioned PGGB. It’s the fad of the month for upsampling.

One thing to bear in mind is that any file change costs you not just in terms of storage, but also time when a new latest and greatest algorithm is in vogue. It’s a massively expensive and time consuming process. Conversely, hardware upsamplers like the Chord mscalar, seem expensive (GBP 4k or so) but consider this: they cost you no time, no storage, and neither do future upgrades. And in most tests I’ve read, the hardware upsamplers sonicly perform better.

FWIW, I can’t be bothered. 16/44.1 sounds superb to me and well mastered material doesn’t really sound that inferior to 24/192.

I’ve tried from my Synology NAS in the past but the PC was far superior. I might try it again with a short connection to the NDS and hi res files else I’m currently stuck with it until something better comes along.

One of my mates uses the Chord upscaler. I’m all for hardware rather than software solutions as it is clear from my little test that hardware, or at least pre-prepared software solutions are better than on the fly solutions. I’m also a bit dumb as I’ve not yet got around to visiting my nearest dealer (1 1/2hr away by the looks of it) to how this would work with a streamer and the NDS and expense to change in one go becomes a little daunting.

Well to work with a streamer, you need an Mscalar and a streamings transpirt and then demote the NDS to a DAC only.

So, it is quite a few boxes and things to plug in for a single source.

Only you can decide if it’s worth the bother. Honestly, I wouldn’t. I have a few albums not just upscaled but bought as 24/192 masters and the same again as 16/44.1 rips from the CD and to be honest, the difference isn’t large enough to make me feel like existing 16/44.1 should be converted to hi/ultra res.

I agree that any PC can upsample with relatively low load on the CPU. My point is though that the stored oversampled file is clearly better than than upsampling on the fly, at least in my system. Well understand your point about the NDS - that is why I use the BNC input. In my experience anything that lowers the amount of work a piece of hardware needs to do results in superior quality.

Really? I think you’ll find the performant upsamplers absolutely thrash even modern CPUs during upscaling. Upscaling is a hugely computationally expensive operation. Anything that’s doing it fast and lightweight is almost certainly just doing basic interlopation oversampling.

Yet it still sounds better, so why not convert to hi res before you use the file?

I haven’t looked if one software converter is better than others, life is probably too short.

Set a program running and come back in the morning when it’s all converted (might be a bit hopeful there on timing - it took 1min 37 secs to do an album).

Question still remains what the best resolution and format is to minimise the load on whatever source component you are using. Does it matter?

Other than the Hard drive it’s free for me to convert to hi res. 4TB SSD hard drives are about £400 so in hi fi terms it’s still cheap for me to do. 12TB HDD’s about the same. Its only when you start looking at Melco’s, Innuos etc that price of storage is more of an issue hence why I’m asking if the differences are as noticable. If I convert now then I will save on the price of an upscaler but spend a bit more on the streamer (probably)

Changing sample size and bit depth won’t increase SQ just like it’s not possible to convert lossy files into lossless files.

Only way to get true hi-rez files is to buy/stream them or record your own at the higher sample rates. 48hz and 24 bit is optimum.


If that is the case then why do so many of us upsample/oversample. There wouldn’t be any point. I’ve read but I don’t fully understand some of the more technical explanations re filtering in the DAC etc that do seem to suggest it is better to upsample and points to why we(some of us at least) hear a difference in our systems. If that is true then it suggests to me that upsampling the original file and then playing back without software intervention is preferable to upsampling on the fly. On my current system this seems to be the case. If it isn’t on more expensive server/music library systems then I can save some dosh. Nothing would give me greater pleasure, but it is an itch I am trying to understand.

By this logic software should be able to turn MP3s into 24/192…

Naim streamers do some oversampling in the DAC stage but that’s something different.

Converting files normally leads to loss of fidelity in my experience.

You, of course, can but it’s like watching an old 70’s sitcom on your 4kTV (And yes I know the oversampling is totally different)

Hifi people do many things for which there is no point, so this doesn’t say much either way :slight_smile:

I figure that with certain (probably poor) player implementations there could be an advantage in removing the need for the sharp low-pass filter that has to cut off the data above 20 kHz when using 44.1 sampling rate. But that’s a wild guess on my part

1 Like

I think we need to seperate upsampling and oversampling. I can quite see how upsampling can cause degradation but not oversampling ie to double/quadruple rates as the original data is still there but presented to the DAC at something nearer it’s theoretically capable sample rates

Indeed. And if you start with 16bit 44.1 you can’t make it sound better by converting it. It would make the files larger but not with info just zeros.

I have about 20 songs that Nick (four legs) gargle blasted for me using PGGB software. Each song is now around 1.5 GB in size. They are now 24 bit 705 KHz,they started off as 16/44.1 CD quality files.
They sounded significantly better using the USB input on my old Chord TT 2 dac,rather than letting my old Mscaler do the work.
I traded the Chord gear in,so my current Dac can not handle these files.
Believe me,it works,but albums take up about 15-20 gigs using PGGB.
Your dac no longer has to process anything this way,just convert it to analog.
Don’t dismiss it as pointless,unless you have actually tried it…I have.:slight_smile:

1 Like

This is a point for consideration though. As mentioned, given enough albums, the running cost of a NAS for that (large disks, spare disks, backup disks, eventual replacement of the lot) will cost more than a hardware upscaler. And if PGGB or whatever you use comes out with the latest and greatest algorithm that they claim is a game changer (as they always do), you have a lot of work cut out.

It’s not for the faint of heart - regardless of how good it sounds.

@mikebu You asked about the theoretical best sampling rate to reduce workload and therefore processing noise on a DAC. An educated guess would tell me this is going to be the maximum supported bit depth and sampling rate of the DAC. Of course if you buy a new DAC you may have to do it all over again.

I hear ya,I never ended up even getting the free trial to work,my rickety old laptop was not powerful enough. I would probably only blast my 100 favorite albums if I did go there.
I went in a different direction anyway,by trading in my Chord gear,which PGGB is really meant for.