Jump to content
IGNORED

Chords New M -Scaler


Recommended Posts

Rob Watts is technically correct in that a longer filter gives a more accurate reconstruction. However, I'm certain that nobody can hear the difference between 100k and 1M taps. Conferring such great importance on the filter length falls, in my opinion, in the same category as talking about skin effect at audio frequencies: real phenomena of limited or no relevance to audio applications.

Link to comment
13 minutes ago, Jud said:

Would it be roughly correct to say:

 

(1) the number of taps is the number of times (or length of time) the filter acts on the signal; 

I wouldn't put it that way.

 

13 minutes ago, Jud said:

(2) as a consequence of #1, a filter with a greater number of taps will cut the signal further (more steeply) than one with a lower number of taps - at least to the point where the signal disappears into the noise? 

More taps allow for steeper filters, yes. More generally, the greater the number of taps, the more precisely you can shape the frequency response.

Link to comment
4 minutes ago, Jud said:

If Chord actually has hardware that will do tasks billions of times larger than can be done now in CPUs or GPUs, and/or do them billions of times faster, I suspect the UK military, Intel, Apple, Microsoft, etc., would all long since have been knocking at his door. 

They use FPGAs from either Altera (now Intel) or Xilinx, both of which have much bigger chips for sale.

Link to comment
15 hours ago, ecwl said:

The issue is that to compute each digital filtered sample, you would need to store in memory the 62500 original digital samples, do the 62500 multiplications and add up the sum, and then do it again and again using 16 different sets of coefficients a million times to get to 1 million samples from the original 62500.

Actually, you don't have to do all those calculations.

Link to comment
40 minutes ago, audio.bill said:

Here's a link to a presentation by M Scaler designer Rob Watts detailing its objectives and technology. He explains how his WTA filter implemented with 1M taps achieves better than 16 bit accuracy while claiming that even the best other DACs today are only 2 or 3 bits accurate with reconstruction. The M Scaler uses 528 DSP cores running at 4096 FS with a total bit depth of 56 bits. Many more details are in the linked presentation. I expect this to generate some further discussion which I look forward to... ?

2-3 bits? That's crazy talk. The whole presentation looks rather hand-wavy to me.

Link to comment
3 minutes ago, ecwl said:

Can you elaborate so I can understand better? I understand that if you're using a million taps to do 16fs, since you only start with 62,500 samples, you'll end up with with 62,500 computations and then sum them up. The only way I can see not doing all those calculations would be to set more coefficients to zero. Except isn't that cheating a little? Then I can claim a 100-tap filter is 1 million taps because the other 999,900 taps have coefficient of zero. Am I missing something here?

With a large number of taps, the calculation can be done more efficiently through the use of FFTs. The end result is the same even though fewer multiplications are performed.

 

3 minutes ago, ecwl said:

I also found the presentation a little too hand-wavy. But that's why I find @rayl1234 diagram (2nd post of this thread) to be helpful, showing how tap length affects amplitude accuracy in the time domain. Can you elaborate more how this is wrong?

How what is wrong? rayl1234's graphs look correct.

 

3 minutes ago, ecwl said:

I feel like I'm hearing lots of people say this whole Chord approach is just marketing or plain wrong (which is a fair enough opinion). And I don't mind finding out that I'm wrong. I am just wondering about the why so that I can learn more about this.

The approach of using a long filter is technically sound. The marketing talk about why they need to be that long is sketchy and over the top.

Link to comment
2 minutes ago, rayl1234 said:

I am not going to play RW defender beyond saying this bit.

 

While I sometimes find his statements imprecise, he has almost always clarified when asked at a show or even on his thread at headfi. He is an earnest artisan and enthusiast more than a commercial marketer. But he does have certain philosophies so to speak. Like against dsd...

I believe him to be sincere in his desire to build the best possible DAC. Some of his engineering decisions in pursuit of that goal also make perfect sense. Some others seem, not exactly bad, but definitely influenced by what you call his philosophies. In marketing oriented material, such as the presentation slides linked above, he tends to make some rather large leaps, such as the "2-3 bits" claim. That just doesn't make sense at all. Oh, and those "transient" waveforms he shows are not band limited as he claims them to be.

Link to comment
46 minutes ago, Albrecht said:

Your speculation here is of course as irrelevant as any "claim" of pet features. Differences in phase filter implementation may play an important, or less important role in the final end result of what one may hear coming out of their speakers. The ONLY way to know is to actually conduct listening tests. ANYTHING else is speculation.

 

"" If two men say they're Jesus, one of them must be wrong.""

A bad analogy: of course there is no right or wrong as what makes Chord or Ayre DACs both very good DACs may or may not be based exclusively on their implementation of phase filters.

 

""I still think he's overdone it to an extent that is hard to justify."

It is really great that very few people would agree with you, and form an opinion about a product on the basis of such unreasonable speculation. Also, a somewhat indirect insult to not only Chord engineers, but to others as well.

Did that outburst make you feel better?

Link to comment
8 minutes ago, mav52 said:

Very true. But, for some the recipe is as exciting as the actual listening. I for one go the listening route but I do appreciate knowledge base of others in helping to explain how something works.   Like my wifes mother ( who is an award winning baker) tells her , there is more ways to bake a cake, its all in the recipe and the methods used.

If a cake recipe calls for vinegar, I suspect we can all say something about how it is likely to taste without actually eating it.

Link to comment
10 minutes ago, firedog said:

I guess you don't bake much. Lots of good cake recipes use vinegar or other similar tasting ingredients. It's all in the taste you get in the end from the combination of ingredients.

You can still predict something about the result just looking at the recipe. Things have known tastes that combine in known ways. In electronics, components behave in predictable ways. It is not necessary to listen to everything in order get some idea of how it might sound, especially when the question is simply whether or not a specific change will make an audible difference. It's how things are designed. Or did you think people like Rob Watts wire together a bunch of parts at random until they chance upon something that works?

Link to comment
1 hour ago, austinpop said:

So bottom line, while you could get HMS upsampling to 192kHz to most DACs, that only gives you access to 1/4 million taps.

Why can't the full filter length be used with lower output rates? That should be easier, not harder.

Link to comment
9 hours ago, barrows said:

Higher sample rates also allow for a very simple, discrete component, D/A conversion stage.  Such as Chord's Pulse Array, or Jussi's DSC-1 approach.

Many people feel there are sonic advantages to these simple, discrete component, conversion stages.

Now you're talking about rates of a few MHz and up together with sigma-delta modulation to reduce the bit depth. That has very little to do with (sinc) interpolation.

 

Most playback systems, whether contained in one or several boxes, use both interpolation and sigma-delta modulation. The input is first resampled to a higher rate, typically in the general range of 400-1500 kHz. This step (mostly) removes images up to half the new sample rate, thus reducing the demands of the analogue filter. The conversion from this higher rate digital to analogue can be done either with a direct multi-bit converter, such as an R-2R ladder, or via a sigma-delta modulator to a reduced bit depth and a simpler D/A stage. In the latter case, the signal is further oversampled (by zero-order hold) to a still higher rate, typically 5 MHz or more. Software DSD converters, such as HQPlayer and Audirvana+, may use a proper interpolation filter all the way to the final rate, effectively eliminating images entirely.

 

The issue of discrete vs IC conversion stage is a different one again. While there is no question of IC superiority for full-resolution multi-bit DACs, both types of circuit can perform well at low bit depths.

Link to comment
22 minutes ago, Miska said:

That's also what DAC chips do, but since they don't have enough DSP processing power, they can run digital filters only up to 352.8/384k rates and from there onwards use zero-order-hold, so they have images at multiples of the 352.8/384k sampling rate.

The Burr-Brown (TI) DSD1793 and, presumably, others with the same architecture resample to a maximum of 1536 kHz depending on the input rate.

Link to comment
27 minutes ago, Miska said:

Depends, but since most of the source content is 44.1k RedBook, it makes sense to consider 1x rates as baseline when discussing digital filters.

I think it makes sense to not misrepresent facts in order to further ones business interests. Get a sponsored section if you want to promote your products.

 

27 minutes ago, Miska said:

The example plots I posted earlier are from TI/BB DSD1793. For 44.1k inputs, also Analog Devices, AKM, Cirrus Logic and ESS have images around 352.8 kHz.

 

Example from another DAC with TI/BB PCM1795 at 44.1 kHz input, digital filter set to "sharp":

TEAC_UD501-sweep441-sharp.thumb.png.b37ed4d7f9a60e62fb56304a7cbd8cbc.png

 

Here's AK4399 chip, 44.1k input:

AK4399-sweep441-wide.thumb.png.7c4b80409001f69c5e7e3ec8f601c8c7.png

 

Here's AK4399 chip, 192k input:

AK4399-sweep441-192-wide.thumb.png.2472844b891c2162794cf6a93da6613f.png

What about more recent AKM chips? The AK4399 is 10 years old.

Link to comment
2 minutes ago, barrows said:

Yes, perhaps.  But it is not going to be very informative to discuss the M Scaler in a vacuum, with no comparisons to other approaches.  The M Scaler is specifically one way to do oversampling, as there are other ways, it makes sense to make comparisons.

The M Scaler implements the first of the three stages in a typical DAC design, the other two being sigma-delta modulation and D/A conversion. Software performing the same function has been widespread for many years. Some software will also do the sigma-delta modulation. What makes the M Scaler unusual is that it is a standalone box, as opposed to a computer program, doing only the upsampling stage.

Link to comment
23 minutes ago, Jud said:

@Miska does post plenty of measurements of HQPlayer with various DACs.  The DSC1 is not something he is selling, but an Open Hardware design that anyone can build and/or modify as they like.  @mansr has posted measurements of Audirvana Plus using the SoX upsampling and its delta-sigma modulator (which he contributed to the open source SoX software).  And I believe he has his own DAC design available for people to build or modify as they like (again, not something he is selling).

To be clear, Miska is the one with a DAC design.

Link to comment
9 minutes ago, Jud said:

Thanks. My impression is you've built one and provided at least some information about it, though I haven't seen much of what you may have said. I thought you may have provided enough information for someone else to build a duplicate or modification, but I stand corrected.

I once challenged some know-it-all to a DAC building "duel." He didn't accept, so I still haven't done it.

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...