# Upsampling

Does the NSC222 have sample rate conversion built in? Also any interesting comments on upsampling?

The sample rate is the part measured in kHz rather than bit depth of typically 16 or 24. Upsampling is where 44kHz is converted to 192kHz for example.

To answer the OP Naim use 8x oversampling up to 352kHz.
Oversamples are done in powers of 2 so is still theoretically but perfect unlike 44 to 192 where the resampled stream has changed and cannot be mathematically put back.

I’m crap at explaining things so hope this makes some sense.

I’ve always understood that upsampling is where one sample rate is converted to another, which involves an estimation of what was in the original sample. This is possibly why high res recordings taken from a lower sample rate original recording don’t always sound great.

Over sampling is used in DACs to move the conversion noise, as it converts from digital to analogue, out of the audible range. Or something like that. There is loads online about it. Just search for ‘upsampling vs oversampling’.

Not quite true. Upsampling is done in integer multiples of the original signals. When upsampling the signal you add additional samples between two existing samples on the sine wave. Which does not actually change the original, just adds more samples.

Adding additional samples is ‘oversampling’. Up sampling is where it’s not done as an integer multiple e.g. 24/44 to 24/192

Could you explain what you mean?

HH, you may be confusing upsampling with sample rate conversion. Apparently the latter can be potentially problematic where sample rates are not easily divided/multiplied. As an example, the sample rate conversion from 16bit 48kHz as used by DAT to 16bit 44.1kHz as used by CD can give you a CD that sounds rather less good than the DAT. Whereas for example, making a recording at 24bit 88.2kHz or 24bit 176.4kHz is attractive as you can get a very nice hi resolution file from which you can also derive a file suitable for CD by down converting very neatly to 16bit 44.1kHz, which will still sound very good.

2 Likes

From the Naim nDAC white paper.

“To increase the sample rate we have to insert additional
samples between the original samples. If we want to
double the sample rate then we need to insert one extra
sample between each two original samples, if we want
to quadruple the sample rate then we need to insert
three extra samples between each two original samples,
and so on. But what sample values do we put there?
If we oversample by a factor of two we could perform
linear interpolation, so that each additional sample has
The same 1kHz signal as before (upper) and its spectrum
(lower), sampled originally at 48kHz but now zero-stuffed
with 15 zero-valued samples between each adjacent
pair of original samples. The sample rate is now (16 × 48 =)
768kHz
Copyright Naim Audio 2009 Page 6 of 12
Here we have oversampled the data 16 times: why not
8, 10 or any other convenient multiple? The 6th-order
analogue filter used in the Naim DAC provides 36dB
attenuation per octave. Working backwards, we see that
with 8× oversampling applied to CD data the first image
will appear at (8 × 44.1 – 22.05 =) 330.75kHz. At this
frequency the analogue filter (see simulation below)
provides about 125dB attenuation. Given that 24-bit data
has a dynamic range of about 144dB (20log10(224)),
if we use only 8× oversampling we won’t exploit the full
potential of high-resolution music. So the Naim DAC
uses 16× oversampling, where the first image frequency
appears at (16 × 44.1 – 22.05 =) 683.55kHz. At this
frequency the analogue filter provides 191dB attenuation,
exceeding 36dB per octave overall as a result of additional
output stage”

So the nDAC does 16x oversampling for 44.1 recordings, fewer oversamples will be needed if it was 88.2.

So you just replace upsample x times with oversampling?

Thanks Richard. I still don’t understand, but it’s knowledge I feel I can live without.

2 Likes

Essentially, upsampling is a form of sample rate conversion, so I can understand the confusion. However, upsampling is something only done where there’s a theoretical or perceived sonic benefit to doing so, which is not always the case with sample rate conversion (see my post re. converse. of DAT to CD for example).

If you want to play it safe many also reduce the amplitude -6dB to -3dB. This is to avoid the interpolated sample values causing clipping. I think this may be overkill as mastering/mixing today most often involves using metering emulating the interpolation when 8x oversampling so the levels are already reduced (at least in the environments I know). This started happening about 15-16 years ago so older recordings do not have this headroom.

This is how it is described in the Roon documentation (in Roon, like many software doing upsampling, you can set this headroom):

It’s possible for a totally “in-range” piece of source material to have “out-of-range” samples after upsampling. The very simple explanation is: when the interpolator “connects the dots” to form a signal at a higher sampling rate, sometimes the new “dots” fall out of range - causing clipping. This is essentially a defect introduced in the mastering process, and is not a problem with the upsampler. Unfortunately, most software silently ignores these defects, and provide no means to track or manage them.

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.