CHORD M-Scaler Review (Upsampler)

[voodooless]

Rob Watts posted this answer :



That was so vague!! Definitely a non-answer for me.

Let’s see here, the fpga does almost 1 TMAC/sec peak. That quite a lot. You’ll need half decent GPU level performance to match that. Now question is: do we need this for 1 million taps at 768 kHz? I mean just technically, not sonically

That FPGA is a damn expensive part btw, around $200~$300 for just the chip!

 

[voodooless]

He should have at least talked about the numerical resolution of his resampling algorithm. It is entirely possible that it has far less depth than a computer software version making his claim completely invalid and backward.

It has an 48 bit accumulator. I think it’s fixed point.

 

[mansr]

Something about filter lengths: https://www.audiosciencereview.com/...-rme-adi-2-dac-fs-tap-count.22124/post-734918

 

[amirm]

Now question is: do we need this for 1 million taps at 768 kHz? I mean just technically, not sonically

A million taps would need 768 million MACs. For stereo, this get doubled to 1.5 giga MACs. I have not looked at FMAC speed on Intel CPUs lately but I suspect a single modern core should be able to do this.

Edit: I was off by a factor of 1000.

 

[chris719]

Yes, the isolation of the power supply does improve stuff although it can be put into the main case with a little thought put into the effort.

It’s not just a little thought. Putting a power supply inside means it’s going to end up a Class I appliance since no one in small volumes will do the testing or use plastic for Class II double insulated rating. Being Class I means your chassis must be bonded to PE. Potentially not ideal for a little box that goes between other boxes and uses unbalanced connections. An external brick makes much more sense for a product like this.

 

[amirm]

It’s not just a little thought. Putting a power supply inside means it’s going to end up a Class I appliance since no one in small volumes will do the testing or use plastic for Class II double insulated rating. Being Class I means your chassis must be bonded to PE. Potentially not ideal for a little box that goes between other boxes and uses unbalanced connections. An external brick makes much more sense for a product like this.

Given the fact that you have to mess with the controls, I am not seeing any stacking opportunity with M-scaler. Tons and tons of low volume high-end gear has internal power supplies (think amplifiers). So see no reason why this didn't go that route. It just looks cheesy and completely out of the norm for target audience to ship an external laptop supply with it.

 

[voodooless]

A million taps would need 768 million MACs. For stereo, this get doubled to 1.5 giga MACs. I have not looked at FMAC speed on Intel CPUs lately but I suspect a single modern core should be able to do this.

Aren’t you off by a factor it 1k?

 

[chris719]

Given the fact that you have to mess with the controls, I am not seeing any stacking opportunity with M-scaler. Tons and tons of low volume high-end gear has internal power supplies (think amplifiers). So see no reason why this didn't go that route. It just looks cheesy and completely out of the norm for target audience to ship an external laptop supply with it.

It is a recipe for ground loops with unbalanced connections. If you have multiple Class I appliances you will have multiple connections to PE. You can get away with this sometimes, and no problem if you use balanced connections, but adding more is not better. Those amps you mention are Class I, not II. The only Class II you see tend to be AVRs, TVs, etc. but there is a reason they exist.

I guess he could have used an internal power supply for this since this is a digital-in digital-out device and could be transformer isolated on both sides, but still. Even has a USB connection, so one more potential PE connection depending on how it's wired.

 

[amirm]

A

Aren’t you off by a factor it 1k?

Ah yes. You are right.

 

[Jomungur]

Given the fact that you have to mess with the controls, I am not seeing any stacking opportunity with M-scaler. Tons and tons of low volume high-end gear has internal power supplies (think amplifiers). So see no reason why this didn't go that route. It just looks cheesy and completely out of the norm for target audience to ship an external laptop supply with it.

Since this is a review, I should mention it has a remote. It works fine. Doesn't help the pauses when switching modes, but it does help if you don't want to reach for the buttons. I never really use it since I kept it in 16x mode, so I forgot to send it, sorry!

And now that I am looking at the manual, I remember the DX mode thing. Those are the buttons on the right side. There's this cryptic reference:

"The last two buttons on the Hugo M Scaler unit (DX OP, DX down, DX up) are not to be used as of yet, they have been implemented for a future product that will pair with the Hugo M Scaler. Please leave them off."

(Although it refers to the DX OP button here, that is supposed to work and indicates the input sample rate. So that reference is just a nit in the manual, that sentence is just referring to DX down and DX up).

 

[AlephAlpha001]

Given the fact that you have to mess with the controls, I am not seeing any stacking opportunity with M-scaler. Tons and tons of low volume high-end gear has internal power supplies (think amplifiers). So see no reason why this didn't go that route. It just looks cheesy and completely out of the norm for target audience to ship an external laptop supply with it.

Ah... you need to get out more over at some other forums and read the endless esoteric debates about third party power supplies for Chord gear -- at Chord prices, of course.

You see, it's just a Starter Power Supply Usual disclaimers apply if your system or ears are insufficiently 'Revealing'.

 

[earlevel]

Aren’t you off by a factor it 1k?

Yes. And at a million taps...aren't we well past the point where it should be done in the frequency domain (where the convolution becomes a multiply)?

But then, I'm still confused why one would want to upsample and yet keep a super steep filter ~20k. I guess it's just to have control over that filter, so the DAC can have a relaxed filter, but I sitll don't see a win unless the DAC of choice simply sucks at the lower rate. None of this makes much sense to me, the least of which being that longer explanation from Rob Watts.

 

[voodooless]

Yes. And at a million taps...aren't we well past the point where it should be done in the frequency domain (where the convolution becomes a multiply)?

Sure, if you were to emulate this on normal hardware/generic CPU hardware, FFT/iFFT would definitely be the faster way to go. It’s really a numbers game, see here how much difference it makes: https://ccrma.stanford.edu/~jos/ReviewFourier/FFT_Convolution_vs_Direct.html

Ah yes. You are right.

That would bring us almost into TFLOP territory. You’d need a high CPU count i7 or big AMD CPU to get that kind of compute power, or as I stated earlier, a decent GPU… or just see above for a shortcut

So from this point of view, Watts was almost right on his point.

 

[CN211276]

A point that needs making is that the Mscaler is designed to enhance the performance of DACs with dual BNC inputs. This is limited to the Qutest, TT2 and Dave. It does not include Chord's mobile DACs such as the Hugo 2, where it is merely compatable.

 

[voodooless]

A point that needs making is that the Mscaler is designed to enhance the performance of DACs with dual BNC inputs.

Welcome! Do you think that after 36 pages, nobody brought this up already? What enhancements would those be? Where did we see these in the measurements? The 16x dual BNC mode was measured after all...

 

[BDWoody]

A point that needs making is that the Mscaler is designed to enhance the performance of DACs with dual BNC inputs. This is limited to the Qutest, TT2 and Dave. It does not include Chord's mobile DACs such as the Hugo 2, where it is merely compatable.

So what does it do for those it is 'merely compatible' with?

Do you own one?

 

[HarmonicTHD]

A point that needs making is that the Mscaler is designed to enhance the performance of DACs with dual BNC inputs. This is limited to the Qutest, TT2 and Dave. It does not include Chord's mobile DACs such as the Hugo 2, where it is merely compatable.

First post, no proof

,.. is this BS ever gonna end?

Read the review, understand it and read the thread were it has been refuted a gazziillion times over.

Must be a fanbot.

 

[Jimi Floyd]

Something about filter lengths: https://www.audiosciencereview.com/...-rme-adi-2-dac-fs-tap-count.22124/post-734918

That is a great post!

 

[DavidEdwinAston]

It must be hell, being a "subjectivist". One is so open to the promotion of equipment at a cost, as to make all Chord products seem cheap! So very many.
Here, on the other hand, a relatively small number of choices. And as good sound as we will need to enjoy our hobby.

 

[HarmonicTHD]

It must be hell, being a "subjectivist". One is so open to the promotion of equipment at a cost, as to make all Chord products seem cheap! So very many.
Here, on the other hand, a relatively small number of choices. And as good sound as we will need to enjoy our hobby.

Yeah maybe. On the other hand sometimes people are quite content worshiping their religion. But when they set out on missionary quests and the beliefs meet the physical reality things start to spiral off course.