CHORD M-Scaler Review (Upsampler)

[aj625]

Did stereophile something do differently ? Here there is close to 130db attenuation brick wall filter at 2x and this is not even 16x. I thought m scaler is exactly for this filtering. @amirm can you please explain why your test is showing only 80db attenuation ? Can you please repeat all tests with 16x upsampling too ?

 

[DSJR]

When I went to a presentation, specifically on the M-Scaler, a Dave was used. Their rep did the before and after and had the remote controls. I definitely heard a difference, the M-Scaler in use supposedly presenting a deeper sound stage over the bypassed setting done via one of th ebuttons on the front. The thing is, I'm certain the mean volume level was different, having vivid memories of trying to compare two things audibly and getting totally confused. In this case, the 'quieter' one sounded better in an audiophile sense. The audience, all 50+ in age and most I'd say 60+ too (so with age degraded hearing I'd suggest), nodded sagely at Rob W's presentation and seemed to enjoy the dem. If we were told the upscaling amount, I can't remember the setting.

 

[goat76]

What a strange and somewhat aggressive response to a simple question. Relax buddy, there's no need to be quite so defensive.

I see no problem with your simple question, this place can be a bit toxic sometimes.

I think most people testing the M Scaler at home would have a listen to all the upscaling options.

 

[aj625]

This is from golden sound review

 

[amirm]

Did stereophile something do differently ? Here there is close to 130db attenuation brick wall filter at 2x and this is not even 16x. I thought m scaler is exactly for this filtering. @amirm can you please explain why your test is showing only 80db attenuation ? Can you please repeat all tests with 16x upsampling too ?

Yes. It is simple. The reference for the signal itself is -40 dB (due to FFT Gain):

So the difference between in-band and out of band is about 85 dB.

I take care in my measurements to zero the in-band signal level so you don't get subjected to the confusion above:

I heavily zoomed into the frequency response which shows a stepped attenuation at -79 and -89 dB which is essentially the results JA got.

It is a pet peeve of mine that folks show measurements not referenced to 0 dB causing this kind of confusion. Yes, it takes a bit of extra work to reference to 0 dB but avoids a heap of misunderstanding.

 

[amirm]

This is from golden sound review

That is a test in digital domain and not gone through a DAC as I and JA are showing.

 

[amirm]

Adding on to above, when you upsample, the DAC operates at a different sample rate so its performance/noise floor (and impact of FFT gain) changes. Testing just the S/PDIF signal in digital domain doesn't show this real-life aspect.

 

[AudioSceptic]

Yes. It is simple. The reference for the signal itself is -40 dB (due to FFT Gain):

So the difference between in-band and out of band is about 85 dB.

I take care in my measurements to zero the in-band signal level so you don't get subjected to the confusion above:

I heavily zoomed into the frequency response which shows a stepped attenuation at -79 and -89 dB which is essentially the results JA got.

It is a pet peeve of mine that folks show measurements not referenced to 0 dB causing this kind of confusion. Yes, it takes a bit of extra work to reference to 0 dB but avoids a heap of misunderstanding.

That same graph appears to show that the brickwall is at 24 kHz, not the expected 22.05. Is that correct?

 

[amirm]

The files were bypassed and 2x - not 16x or 4x and he didn't try listening with 4x or 16x, which was what I was questioning.

It wasn't what you were questioning. This is what you said:

I've not used a Hugo2. As it allows dual-BNC, why was full upsampling not used for the tests when the option was available to do so? This seems even more odd than I had thought.

You were questioning the entire test including measurements. Your statement was false and shows that you didn't pay attention to the review. I clearly tested all sample rates including 16X using dual Coax:

That is 176 kHz. Other tests also mentioned the same:

Your claim doesn't make sense anyway as the company brags about the long tap filter which is active in all upsampling modes. If all but 16X are bogus, then they should say so and they clearly don't.

I also explained in the review why I did not attempt to make a comparison to higher sample rates. I explained that they cause a pause/glitch which invalidates any short-term memory you may have. You are welcome to ignore that and arrive at bogus comparisons but I have to follow the science here.

 

[amirm]

That same graph appears to show that the brickwall is at 24 kHz, not the expected 22.05. Is that correct?

Different sample rates were used. Mine was at 48 khz instead of the usual 44.1 kHz. Label was incorrect in my screenshot.

 

[amirm]

Has the term ever been used (not just quoted) in ASR?

Which term? Processor? If so, some of the high-end external DACs years back called themselves "processors." This fell out of favor after a while but stereophile has stayed with it.

 

[amirm]

In my experience using bypass was not the same as removing the M-Scaler from the chain, so using bypass might be flawed anyway (I believe bypass might be very slightly louder but I'm going from memory).

Your memory is wrong. Again, this was explained in the review. In all modes M-scaler reduces the volume. If you compare no M-scaler to M-scaler, you get a severe volume mismatch which invalidates any listening test. Your comparison must be bypass vs upsample with M-scaler always in the loop as I did so levels don't change.

 

[sq225917]

I just read the subjective part of that review and JA has slipped down a few more notches in my estimation.

And what's with this quaint term "D/A processor" he still uses when the rest of the planet calls it a DAC?

But the mscaler isn't a dac, it's solely an upsampler, so D/A processor is entirely correct.

 

[Chester]

It wasn't what you were questioning. This is what you said:



You were questioning the entire test including measurements. Your statement was false and shows that you didn't pay attention to the review. I clearly tested all sample rates including 16X using dual Coax:

That is 176 kHz. Other tests also mentioned the same:

Your claim doesn't make sense anyway as the company brags about the long tap filter which is active in all upsampling modes. If all but 16X are bogus, then they should say so and they clearly don't.

I also explained in the review why I did not attempt to make a comparison to higher sample rates. I explained that they cause a pause/glitch which invalidates any short-term memory you may have. You are welcome to ignore that and arrive at bogus comparisons but I have to follow the science here.

As far as I understand it, the million tap filter is only engaged at the max upsampling, I think that is the point that @spooky is trying to make. And it’s the million tap filter that is the main feature of the product.

I’m not saying it would have changed anything.

 

[sq225917]

Here's a thought, maybe the dave osnt phased by the added jitter and is able to reject it? Its not like anyone buys an mscaler to use with a none Chord dac, do they?

 

[Jimbob54]

As far as I understand it, the million tap filter is only engaged at the max upsampling, I think that is the point that @spooky is trying to make. And it’s the million tap filter that is the main feature of the product.

I’m not saying it would have changed anything.

I'm genuinely interested, would one expect any of the tests conducted to give a measurably different result had that been engaged?

 

[amirm]

Here's a thought, maybe the dave osnt phased by the added jitter and is able to reject it? Its not like anyone buys an mscaler to use with a none Chord dac, do they?

Again jitter is not a practical problem. It is just a sign of poor engineering. It may also be showing too high of quantization noise in the filter.

 

[amirm]

I'm genuinely interested, would one expect any of the tests conducted to give a measurably different result had that been engaged?

It was engaged. I just showed it in two measurements.

 

[AudioSceptic]

Which term? Processor? If so, some of the high-end external DACs years back called themselves "processors." This fell out of favor after a while but stereophile has stayed with it.

I meant, has "D/A Processor", as still used by JA, been used here at ASR instead of "DAC"? I wasn't referring to what manufacturers choose to call their own products. Digital Decoding Engine was a good one though.

 

[Chester]

I'm genuinely interested, would one expect any of the tests conducted to give a measurably different result had that been engaged?

No idea, probably not. But leaves that element of doubt for manufacturers and die hards to dismiss anything that is there. I think that’s the missed trick. Not intended as a criticism Amir, you know I think the effort you put in here is outstanding.