Jump to content
IGNORED

Chords New M -Scaler


Recommended Posts

On 10/10/2018 at 7:50 AM, Miska said:

 

This would have been much less eyebrow rising if you would have immediately quoted the original text... So it is virtual intermediate like I was speculating before. So there are no samples produced at 250 GHz rate ever. You said first that it upsamples to 250 GHz. But from what you quoted it specifically says it doesn't do that. What you quoted describes a polyphase filter.  The output rate is just 211 kHz which is the rate it really operates at. HQPlayer has quite a bunch of such, but operates and outputs for example at 24.576 MHz output rate. But it never crossed my mind to brag about the "virtual intermediate" rate (which can be easily in GHz range) or claim that such way to process a filter would be special. For me, more interesting is the filter design itself, instead of how it is applied.

 

Even Windows and macOS have such resampling filters built-in, just for small change ratios and poor filter designs to keep computational cost as low as possible.

 

 

I don’t think it was bragging. It was an application note on their website. John Siau writes many of these notes to explain about the hardware they design and the choices made. The 250 GHz sample rate is used to make 4 picosec timing adjustments so it is integral to the way the Benchmark DAC 3 stays phase locked while using the asynchronous clock in the DAC. For sure there is a marketing spin to everything they write - John Siau very likely  believes he is making wise and rational choices to provide optimal performance at a price point - this is no different from any technical company trying to win customers. These technical notes make it quite clear that Benchmark are not simply using the ESS 9028 chip straight out of the box - they are customizing and I guess that is a marketing differentiating point vs competitors that they are trying to make.

 

Anyway thanks for your effort and time to explain things. I take it that HQPlayer can’t really do anything for any of the Benchmark products because they are highly customized and “hardwired” unlike most DACs which just connect a power supply and inputs and outputs to the DAC chip and use the chip as it is.

Benchmark DAC2, Active speakers: ATC 150's, 100's, 20's, C6CA, C6 Subwoofer.

 

Headphones: Only for playing drums. I don't like sounds in my head. The best headphones suck. Nothing can replace good speakers played loudly. And nothing absolutely nothing is a substitute for live music!

Link to post
Share on other sites
3 hours ago, Shadorne said:

Anyway thanks for your effort and time to explain things. I take it that HQPlayer can’t really do anything for any of the Benchmark products because they are highly customized and “hardwired” unlike most DACs which just connect a power supply and inputs and outputs to the DAC chip and use the chip as it is.

 

I wouldn't say it is "just connects", there's quite a bit more involved. However I'm most interested on DACs that don't use any DAC chips, but instead have more original, discrete, design. Like Chord which this thread is about. Or Holo Audio, Denafrips, dCS, Playback Designs, etc.

 

Based on my measurements, ESS chips perform the best when running at DSD512. But Benchmark DACs don't support that. Or even PCM inputs higher than 192 kHz. So it is not really interesting to me. There are other ESS based products that offer more.

 

I also like more DACs that offer capabilities of the AK49xx chips.

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to post
Share on other sites
  • 2 months later...
46 minutes ago, ecwl said:

Haha... I just flipped through this thread and realized despite all the technical discussions, only @romaz and myself actually own M-scalers in the form of Blu2 and DAVE (and we have both tried HQPlayer) and we barely talked about our subjective sound quality impressions of M-Scaler. And @romaz and I basically owned both Chord products around the time when they came out.

The best way to describe M-Scaler over DAVE is to describe DAVE over other DACs you might have had before. I don't know if you've noticed DAVE, with some instruments or vocals, the timbre of the instruments and the voice just sound more realistic than what you're used to with other DACs. Or when you're listening to percussion instruments, going from guitar plucks to finger snapping or hand clapping all the way to strikes of drums, DAVE sounds more realistic and dynamic compared to most DACs you know. Well, whatever DAVE to the other DACs is M-Scaler to DAVE.

I almost never listen to DAVE alone but ironically in the past two days I did. My dealer had a new setup so I dropped by to listen to some unfamiliar pieces through his DAVE (direct to PS Audio BHK 300 mono amps) and things sound great except the instruments just didn't sound as realistic as I'm used to so when I got home, I listened to the same songs through Blu2 and the transients and timbre of instruments were at another level that I really enjoyed.

Because I disconnected my video system from my Blu2, I decided to watch TV using DAVE alone (instead of the 0.6M taps of M-Scaler) and within 2 minutes, I just didn't appreciate how the sounds of explosions or people fighting didn't seem quite right for me, in addition to people's voices not sounding as realistic as I'm used to. Plugging Blu2 back into the video system had a nice improvement.

I would say that if you own DAVE and can afford it, M-Scaler is a no-brainer. The only caveat is that there is a possibility that your downstream devices after the DAVE, e.g. preamplifier (if you have one) or amplifier might be a limiting factor in hearing all the additional improvements. I think that's why the Head-Fi forums tend to rave about Chord products more because there are fewer intermediate factors that can reduce the performance of the DACs or cause synergy issues.

Anyway, that's my personal subjective (and obviously biased) opinion. Hope it's helpful to you.

 

That is helpful thanks for the response. I split my DAVE use between headphones and active speakers so not much downstream implications. The sound is so pure on it's own, I think I want to keep it that way. :) I will probably go for the Mscaler, my curiosity is killin' me! 

 

Link to post
Share on other sites
On 8/2/2018 at 11:05 AM, BigAlMc said:

Just read all ten pages of this thread and not one of you bozo's answered the all important question! Will this thing improve @BigAlMcs system! ?

 

Short answer I suspect is not without significant changes, as I'm not even close to being willing to consider parting with my Directstream DAC and because I have a heavily optimized USB flow.

 

But would appreciate some more expert input on the following. The M-Scaler sounds very impressive but it has USB input but no USB output. Therefore I could potentially put it after my TX-USBultra but not before.

 

So if I have a TX-USBultra (clocked by an SoTM OCX-10 reference clock) providing a very, very nice USB signal to my DAC. In theory would feeding that USB signal into an M-Scaler have potential benefits or be a dumbass idea?

 

Would the M-Scaler take that very nice USB signal and scale it to 386Khz or whatever making it even nicer? Or would the M-Scaler completely reclock/rejig the signal to the extent that the efforts (money spent!) on the TX-USB-Ultra/OCX-10 were rendered pointless or lost?

 

Cheers,

Alan

 

Did you ever get a response to your question? I have the same question. I have the SOtM master clock and SMS200 ultra neo and an M scaler on order. Trying to figure out if anyone has used and M scaler with the SOtM USBultra and master clock.

Link to post
Share on other sites
39 minutes ago, EnjoyTheMusicNow said:

 

Did you ever get a response to your question? I have the same question. I have the SOtM master clock and SMS200 ultra neo and an M scaler on order. Trying to figure out if anyone has used and M scaler with the SOtM USBultra and master clock.

 

Not really. There was a suggestion that the M scaler makes the most sense with a Chord dac. But the combo of super clocking and super scaling wasnt really opined on.

 

Be interesting to hear how you get on.

 

Cheers,

Alan

Synergistic Research Powercell UEF SE > Sonore OpticalModule (LPS-1.2 & DXP-1A5DSC) > EtherRegen (SR4T & DXP-1A5DSC) > (Sablon 2020 LAN) Innuos Zenith SE server > (Sablon 2020 USB) Innuos Phoenix > (Sablon 2020 USB) PS Audio Directstream DAC > PS Audio M1200 monoblocks > Salk Sound Supercharged Songtowers

Link to post
Share on other sites
  • 2 weeks later...
On 10/12/2018 at 5:19 PM, Miska said:

Based on my measurements, ESS chips perform the best when running at DSD512.

 

Do you think the ESS chips noise shaping operate at 256fs or 512fs? In this old 2013 article they said the ES9018 could be operating at 256fs but also possible at 512fs (guessing).

 

From your recent measurements and looking at ultrasonic noise patterns does it look more like newer ESS chips (like in the Pro-Ject S2 DAC for example) are running noise shaping at 512fs? Or likely 256fs from your measurements/guess?

 

Or still too hard to guesstimate?

 

"The requirement for feedback-based noise shaping is still there, though: the ESS Sabre 9018, which is probably the most advanced delta-sigma converter in current production, most likely operates at 11.2896MHz, or 256fs. (Published data on the internals of the 9018 is not readily available; the 256fs speed is a best-guess. It might go all the way up to 45.1584MHz.) "

 

https://positive-feedback.com/Issue65/dac.htm

 

This ES9023 spec sheet says:

 

"For best performance. 256fs or greater is recommended for 32kHz to 96kHz sampling."

 

But I don't know if this has anything to do with the noise shaping sample rate?

 

98264182_ScreenShot2019-01-13at8_46_01pm.thumb.png.a33ed50636e64a7d3266af1c771ea3c7.png

 

https://www.computeraudiophile.com/applications/core/interface/file/attachment.php?id=6757

 

 

Link to post
Share on other sites
7 hours ago, Em2016 said:

 

Do you think the ESS chips noise shaping operate at 256fs or 512fs? In this old 2013 article they said the ES9018 could be operating at 256fs but also possible at 512fs (guessing).

 

From your recent measurements and looking at ultrasonic noise patterns does it look more like newer ESS chips (like in the Pro-Ject S2 DAC for example) are running noise shaping at 512fs? Or likely 256fs from your measurements/guess?

I suspect ESS chips when fed DSD512 just plays it like a DSD DAC as if it's a 64-element shift register DSD design. So I don't think there's any additional noise shaping involved.

 

But when you're playing a PCM file on an ESS chip, I'm not sure if we can easily know whether it upsamples and noise shapes to 5-6 bits at 64fs, 128fs or even at higher fs. We also don't know whether it upsamples to say 16fs 24-bit first and then to say 128fs 6 bits or whether it'll be a one-step process.

 

But I think whatever the process is, it's almost a little irrelevant. First of all, I am quite certain ESS chips do not upsample from PCM and noise shape directly to DSD512. Moreover, the computational power involved in the ESS chips would be dwarfed by what HQPlayer can do when it upsamples and noise shapes to DSD512.

 

Similarly, if ESS chips upsamples to 16fs first like M-Scaler, the computational power for the upsampling filter with ESS would be dwarfed by what the M-Scaler can do. And the subsequent upsampling or noise shaping to 5-6 bits at say 128fs to be output to the 64 elements SDM would be computationally significantly less intensive than say the Chord Qutest upsampling from 16fs to 256fs and then to 104MHz then noise shaping for playback on the 10 discrete elements of the pulse array DAC.

 

I think the bottomline, whether you're going with a final DAC output of DSD512/DSD1024 or a multi-element discrete SDM design is that the more computational power (and superior algorithm) you can throw at it for upsampling and noise shaping, the better the sonic result. (Unless you believe that all DACs measure the same and all noise/distortions measured are already beyond the threshold of hearing in which case you should stick with your headphone amp/DAC/jack that comes with your cellphone...)

Roon (convolution filter using Acourate) > ultraRendu > Peachtree X1 (Toslink) > Chord Hugo M-Scaler > Chord DAVE > Chord Etude > Dynaudio Confidence C1 Signature + Sunfire TS-EQ10 subwoofers

Link to post
Share on other sites
8 hours ago, Em2016 said:

But I don't know if this has anything to do with the noise shaping sample rate?

 

98264182_ScreenShot2019-01-13at8_46_01pm.thumb.png.a33ed50636e64a7d3266af1c771ea3c7.png

 

Not much, these are just clocking requirements. The chip needs at least one clock cycle per output sample (at it's native SDM rate), but it is not unusual to need more than one clock cycle per output sample. Higher grade ESS chips can use up to 100 MHz clock. Sometimes chips can adapt and also accept lower than optimal clock rate, but then they make a shortcut at DSP side for example and quality degrades to some extent.

 

In these "traditional" implementations everything runs out of single clock (unlike in separate upsampler + DAC case). Which has challenge that when you increase clock frequency your phase noise figures get worse. And if you lower clock frequency you limit the DSP capabilities. If you run the DSP and DAC asynchronously from completely independent clocks, you can do whatever you like in DSP and you can choose optimal clock frequency for the DAC purely from the conversion stage perspective. But so far, all DAC chips run out of single MCLK.

 

For comparison, you can run the DSP stuff on CPU running at 4 GHz and DAC having oscillator of it's own running at 22.5792 MHz for DSD512 (44.1 x 512). DAC only converts samples to analog as-is. And computer only does DSP. Each optimized for their own domain of work.

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to post
Share on other sites

We know more than that about ESS.  My understanding for the 9018 is that it first makes an 8x pass, this is controllable and can be defeated (although this where the "jitter eliminator" works, as an asynchronous sample rate converter), then there is another step of up conversion, to very high MHz levels at 6-9 bits (bit rate is user selectable).

I am almost certain, that with DSD input, there is an SDM step, but I am not sure about how the sample rate is handled.  Remember that ESS volume control works with DSD, so there is processing going on.  Although ESS does seem to remain tight lipped about exactly what happens.  The ESS chip will not convert DSD if the first stage of processing (the "OSF") is turned off (true for 9018 and 9038).  It is clear the ESS chip is not operating as simply as a discrete DSC-1 type DAC, or even as simply as the recent AKM chips in pure DSD mode.

 

That said, I am very happy with the sonics of the ESS 9038 feeding it exclusively DSD 256, so I optimized my DAC build for DSD 256 and it sounds great.  I also run it synchronously, supplying Bit Clock and Master Clock from the same oscillator such that the DPLL completely drops out.

ROON: DSD 256-Sonore opticalModule-Signature Rendu optical--Bricasti M3 DAC--DIY Purifi Amplifier-Focus Audio FS888-JL E 112 sub-Nordost Tyr USB, DIY AC, Iconoclast XLR & speaker cables, Synergistic Orange Fuses, Dark Matter system clarifiers.                                                       

                                                                                           SONORE computer audio

Link to post
Share on other sites
23 minutes ago, barrows said:

It is clear the ESS chip is not operating as simply as a discrete DSC-1 type DAC, or even as simply as the recent AKM chips in pure DSD mode.

 

AKM also goes a bit further and has always been recommending a different (reference) analog filter design for DSD usage than for PCM.

 

Not many DACs (end-user products) really do something like that, most just use analog filters optimized for PCM. Still they tend to perform better when running at DSD. And could be even better if optimized for DSD.

 

For example T+A DAC8 DSD is quite unique in a way that it really has two different analog filter settings you can select.

 

23 minutes ago, barrows said:

My understanding for the 9018 is that it first makes an 8x pass, this is controllable and can be defeated (although this where the "jitter eliminator" works, as an asynchronous sample rate converter), then there is another step of up conversion, to very high MHz levels at 6-9 bits (bit rate is user selectable).

 

It has two cascaded programmable FIR filters. First stage is 64 taps and second stage is 16 taps. This is also what is used for "MQA rendering" with MQA's own filters.

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to post
Share on other sites
7 minutes ago, Miska said:

Not many DACs (end-user products) really do something like that, most just use analog filters optimized for PCM. Still they tend to perform better when running at DSD. And could be even better if optimized for DSD.

That is pretty cool if people take advantage of it.  I changed the analog filter in my DAC a bit (relaxed it for DSD 256 input).

Jussi, are the recommended analog filters in the AKM data sheet?

ROON: DSD 256-Sonore opticalModule-Signature Rendu optical--Bricasti M3 DAC--DIY Purifi Amplifier-Focus Audio FS888-JL E 112 sub-Nordost Tyr USB, DIY AC, Iconoclast XLR & speaker cables, Synergistic Orange Fuses, Dark Matter system clarifiers.                                                       

                                                                                           SONORE computer audio

Link to post
Share on other sites
4 minutes ago, Miska said:

It has two cascaded programmable FIR filters. First stage is 64 taps and second stage is 16 taps. This is also what is used for "MQA rendering" with MQA's own filters.

The second stage is symmetrical, so there are twice as many taps as programmable values. The 9018 datasheet gives this number as 16 in one place and 14 in another. Datasheets for more recent chips say 16 values can be written but only 14 are used.

Link to post
Share on other sites
1 hour ago, mansr said:

The second stage is symmetrical, so there are twice as many taps as programmable values. The 9018 datasheet gives this number as 16 in one place and 14 in another. Datasheets for more recent chips say 16 values can be written but only 14 are used.

 

IOW, the second stage must be a linear phase filter... Which is a sensible choice.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to post
Share on other sites
3 hours ago, barrows said:

We know more than that about ESS.  My understanding for the 9018 is that it first makes an 8x pass, this is controllable and can be defeated (although this where the "jitter eliminator" works, as an asynchronous sample rate converter), then there is another step of up conversion, to very high MHz levels at 6-9 bits (bit rate is user selectable).

 

So we still don’t know to what MHz sample rate?

 

Do you have an idea/guess with your 9038 Pro? If so, how did you come to this number?

 

Link to post
Share on other sites
43 minutes ago, Em2016 said:

So we still don’t know to what MHz sample rate?

I used to know for the 9018, cannot remember now.  There was a huge thread )probably still there) at diyaudio.com about the ESS 9018, and designer Dustin Foremen talked a lot there about how the chip operates.  My understanding is that the final rate is related to the master clock rate (which has a wide possible range).

ROON: DSD 256-Sonore opticalModule-Signature Rendu optical--Bricasti M3 DAC--DIY Purifi Amplifier-Focus Audio FS888-JL E 112 sub-Nordost Tyr USB, DIY AC, Iconoclast XLR & speaker cables, Synergistic Orange Fuses, Dark Matter system clarifiers.                                                       

                                                                                           SONORE computer audio

Link to post
Share on other sites
16 minutes ago, mansr said:

Most likely some fixed fraction of the master clock rate, which due to the ASRC doesn't need to be related to the audio sample rate.

Yes, I think this is correct.  But, the ESS chips, IMO, do sound better when run synchronously (as long as the source/masterclock is low jitter) where the masterclcok and bit clock are derived form the same clock.  For example, I run the ESS 9038 with a 45.1584 masterclock, and this same clock is the clock input to the Amanero interface, and re-clocks the data lines just before the DAC.  Synchronous operation like this drops out (or one can turn it off) the DPLL as it has no need to make changes as the masterclock and bit clock are already synchronous.

ROON: DSD 256-Sonore opticalModule-Signature Rendu optical--Bricasti M3 DAC--DIY Purifi Amplifier-Focus Audio FS888-JL E 112 sub-Nordost Tyr USB, DIY AC, Iconoclast XLR & speaker cables, Synergistic Orange Fuses, Dark Matter system clarifiers.                                                       

                                                                                           SONORE computer audio

Link to post
Share on other sites
  • 1 month later...

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...