To upsample to 24/192 or not

Hey peeps.

Now that Roon build 269 has sorted my physical AV chain I am curious how many people are upsampling everything to 24/192 to - possibly - reduce the processing in the KEF DSP.

I’ve been experimenting and I can’t detect a sound quality difference hence, for me, its a question of whether I want such processing done on my Intel i5 ROCK or in the KEFs. Given that my ROCK is already having to re-sample tracks (I use convolution) I’ve got it upsampling everything to 24/192 too.

I don’t know much about the Keff’s but opinion seems to support doing that on a suitably powerful PC rather than on a somewhat less powerful processor in the DAC. That said, upsampling PCM isn’t particularly taxing so it won’t matter too much. All I might do is stick to the appropriate multipliers so max 44.1 family will be 176k, 48 family will be 192.

1 Like

The LS50W’s does all of it’s DSP at 192 so no point in upsampling to 176 just to have the Kef bump it up further to 192.

Similar to @Rik_Carter I am letting Roon upsample everything to 192, but similarly haven’t noticed any difference to be honest.

1 Like

Same here, I up sample to 192 for LS50W

Upsampling to 192khz or even 384khz makes no differences other then being a heavier load on your CPU, your electrical bill, the environment and a higher wear rate of your computer. I know higher numbers makes you feel better but if you don’t hear a difference just leave it off. Besides, transporting signal higher then 96khz over spdif will put a higher demand on correct termination (say goodby to rca connectors) otherwise transmissionline refelections gets to high and timing errors (jitter) will only be worse then with lower sample rates. A bit of extra jitter is often confused as extra detail. Any difference perceived with upsampling is usually automatic conceived as better quality because higher numbers must be better. I would leave the upsampling inside the Kef and be done with it.

1 Like

I quite agree. Also if you are streaming over IP/Wifi, a 24bits/192Khz stream is 5Mbps. This means less bandwidth available for other devices, more risks of interruptions or crackles… I also use convolution and don’t do upsampling in Roon, so usually streaming at 24bits/44.1KHz 98% of the time…

My understanding is that the Kefs upsample to 192/24 across all inputs and that this is the resolution that’s transmitted to the second speaker. So surely the same costs apply irrespective of where the upsampling is done. My immature logic was that I’d rather put more ‘wear and tear’ on an external CPU which can be easily replaced rather than the internal - impossible to replace - processing units in the Kefs. By providing the Kefs with a 192/24 we have reduced the amount of processing that the Kefs are having to do. I’ve never heard a difference in SQ either so it’s more a question of what’s left to optimise and, in this case, it is very marginal differences related to lifetime wear of components.

Well I wouldn’t worry about the wear of components. The Kefs are obsolete technolgy before they wear down anyways. I don’t think the"ll last 25 years, the hardware will.

Rik I was upsampling previously to 192 via an SMS-200. Now using the direct Ethernet route I’m still upsampling and so far happy with that.

Personally I feel it does sound better

As and when in app Mqa processing appears depending on the options available I may reconsider this as at the moment MQA is making up much of my listening. Looking towards this as one of the next big developments

Yep but as you know heat build-up has a detrimental effect on the lifetime of electronic components and, hence, my logic is that minimising the processing in the KEF’s to only critical stuff should result in a lower operating temperature in the interior. Given that the components are pretty crammed in there any kind of strategy to reduce the need for use of components that generate heat can only help. I should say that this thought process has to be seen within the context of (a) maximising the longevity of these speakers, and (b) the fact that some of us (i.e. me included) have already had to have their KEFs replaced as they quickly developed a fault. That fault may be the result of overheating and I’m waiting to hear back from KEF on the failure mode in the pair I sent back to them.

Would the KEFs attempt to 'up’sample to 192khz even though Roon DSP had already sent through this exact same sample rate … do they have the smarts to recognise that and disengage any upsampling processes or do they continue to take the same approach no matter what?

I can’t think of a reason why they would need to further process a 24/192 input. However once it gets to the DACs they will oversample at some ridiculously high rate as part of the conversion process. That oversampling happens in all DACs as far as I’m aware.

I don’t have the KEF speakers, but for DACs in general it’s supposed to be best to up-sample to whatever sample rate they want, to give the DAC less work so it can focus on the important work. So if I had the KEFs, I would up-sample to 192 (unless on a really slow host).

1 Like

Its unclear to me if the LS50’s are upsampling any further internally. But it reads like the initial DSP stuff is being done at 192 - all input is converted to 192 before entering the DSP stage (analog in included). If they are then oversampling before hitting the D/A chip is unclear - and we don’t even know what Dac chip they are using!

My guess is that the LS50 Dac, in and of itself, is not terribly great. But that doesn’t matter because the active crossover/phase correction/bi-amping is so good that it overwhelms any Dac weakness.

I doubt they are oversampling before the DAC. I did some searching around for what the DAC might be and I mentioned it before in your post (?) on recommendations for a $3000 system? Anyway, the chip is apparently 7 years old. I’m not so sure how important that is when you consider the level of system synergy we are getting with the Kefs.

Right, there was one poster in HK who listed the specs and said the Dac chip used was a Wolfson WM8740E … who knows … it does only support up to 192kHz so that might be why the LS50 is spec’ed similarly. A shame if so - I can find references to the WM8740 going back to 2006! Not sure what the ‘E’ is for. The chip costs about $1.26 so that certainly helped keep the LS50W price down if so :wink:

If the LS50 Wireless truly does perform all DSP at 192 kHz, then I expect it asynchronous sample rate converts all inputs to 192 kHz.

And I seem to recall that the DAC actually is a multichannel codec – ADC and DAC all in one chipset.


1 Like

Yep so not exactly an authoritative source… at least let’s hope so…

1 Like

Interesting. So as I understand there are 2 DACs per speaker, one for each mono amp. So is this multichannel codec a separate DAC that sits just after the inputs? Or are all 4 DACs (2 per speaker) of this kind?

The subwoofer output requires another DAC channel.