Review and Measurements of CHORD Qutest DAC

[KevinJKim]

I liked the neutral in a way ~ I never had the opportunity to try the Qutest on my hifi system but I appreciated the way it sounded with my headphone amps the bakoon 5210 and the Allnic 5000XL. you should check out the reviews on pursuit perfect system on YouTube ~ The reviewer has some extensive reviews on the qutest and I thought it sounded fantastic with his gears~ Congrats again on your purchase ^^~

 

[RustyGates]

Just for correctness - the SINAD measurement here with the Chord is larger than the SINAD measure of the SMSL D1 which was 112.269dB Ch1 & 113.999dB @ 192Khz Balanced (lets not go into its unbalanced performance at a non-fixed vol @ 192KHz), so should it not come "first" in that bar graph.

Just my OCD twitching.

 

[deafenears]

Just for correctness - the SINAD measurement here with the Chord is larger than the SINAD measure of the SMSL D1 which was 112.269dB Ch1 & 113.999dB @ 192Khz Balanced (lets not go into its unbalanced performance at a non-fixed vol @ 192KHz), so should it not come "first" in that bar graph.

""Note that the above is with 90 kHz of bandwidth (to go with 192 kHz sampling). At 44.1 kHz, SIAND climbs to 114.5 average as reported in the original review."""

Take a look at the bottom small print of the dashboard.

 

[JohnYang1997]

SINAD is just a number. If you dig into the graph, you will see chord only has a bit more lower frequency noise. The whole spectrum is super clean.

 

[maxxevv]

Interesting to note that the D1's frequency chart shows cleaner sub-1Khz, the Qutest is much cleaner post-1Khz.

But jitter wise, its very much a wash post-2Khz, though its arguable that the D1 is actually slightly lower.

 

[RustyGates]

""Note that the above is with 90 kHz of bandwidth (to go with 192 kHz sampling). At 44.1 kHz, SIAND climbs to 114.5 average as reported in the original review."""

Take a look at the bottom small print of the dashboard.

Oh right. would've been good to actually see the 114.5 SINAD @ 44.1KHz result on scope and FFT.

 

[RustyGates]

It is possible his architecture doesn't produce balanced output as DAC chips do. He has it on higher end products like Dave but we are talking $12,000 then!

All his designs are inherently single ended; his argument is that the measured performance is good enough such that balanced is not required. To have a true balanced DAC you would need to essentially have double the hardware.

In the TT2 and DAVE, there are balanced outputs, but these are derived from the single ended signal at the output stage - but the measurements from balanced will or *should* be ever so slightly worse than the output through single ended outputs because its gone through an additional stage - the more components through the signal path, the less the quality (even if its ever so slight).

There is no space left on the Qutest PCB to add this additional circuitry for balanced output.

 

[TomHP]

Not unexpected. All my macbooks I've listen to (work or personal) had great (transparent?) sound. It's basically a well implemented Cirrus Logic DAC and headamp set-up which Apple has down to a T. No one would ever believe you though

I disagree. My MacBook Pro 2013 has very noticeable noise floor when connected to sensitive IEMs. I'm sure it sounds good with high impedance headphones, but is to be avoided with sensitive earphones.

 

[Veri]

I disagree. My MacBook Pro 2013 has very noticeable noise floor when connected to sensitive IEMs. I'm sure it sounds good with high impedance headphones, but is to be avoided with sensitive earphones.

Mm there's few devices that work that well with the most sensitive earphones though. My 2015 model also has some noise floor with IEMs yes. Not sure there are laptops that are much if any better

 

[Miska]

I disagree. My MacBook Pro 2013 has very noticeable noise floor when connected to sensitive IEMs. I'm sure it sounds good with high impedance headphones, but is to be avoided with sensitive earphones.

Same as my experience with MacBook Pro's I've had. Doesn't even need particularly sensitive headphones. IIRC, I used Sennheiser HD 280 Pro's and HDD/SSD access was clearly audible and also the power management power-down; 10 seconds after playback had stopped silence returned because OS powered down the audio hardware. It is also not unusual to hear things like display scrolling and such.

 

[SmarterthanU]

The Chord Qutest is the best sounding DAC I have heard for $2000 or even higher. Add the M Scaler and the Qutest is as good as I ever heard regardless of price. I would not waste time or money on any other DAC if you are serious about only investing in hardware of the highest value. RCA outputs are all that is needed to get the best sound out this DAC so XLR would be superfluous on the Qutest as all cable runs will be extremely short in a consumer home setup.

 

[Veri]

I would not waste time or money on any other DAC if you are serious about only investing in hardware of the highest value.

Value you say, hmm.

RCA outputs are all that is needed to get the best sound out this DAC so XLR would be superfluous on the Qutest as all cable runs will be extremely short in a consumer home setup.

You are generalising, there are for sure people who will use longer runs of cable in which balanced is simply better, objectively and practically. Not to say RCA can't suffice but it is a shame in such an expensive device.

 

[SmarterthanU]

Value you say, hmm.


You are generalising, there are for sure people who will use longer runs of cable in which balanced is simply better, objectively and practically. Not to say RCA can't suffice but it is a shame in such an expensive device.

Value you say, hmm.


You are generalising, there are for sure people who will use longer runs of cable in which balanced is simply better, objectively and practically. Not to say RCA can't suffice but it is a shame in such an expensive device.

The DAC is not of a balanced design. Would be interesting to use two Qutests in a balanced setup to see the result.

 

[shanecoughlan]

It is possible his architecture doesn't produce balanced output as DAC chips do. He has it on higher end products like Dave but we are talking $12,000 then!

In my space FPGA architectures are increasingly common. One thing I noted is that there is very little information from the company about what FPGA they are using and how they are using it. This would provide much better insight into whether they were accomplishing something unique or reinventing the wheel. After going through several of your datasets on their products, there is some cause to consider it may be closer to the latter.

 

[JohnYang1997]

In my space FPGA architectures are increasingly common. One thing I noted is that there is very little information from the company about what FPGA they are using and how they are using it. This would provide much better insight into whether they were accomplishing something unique or reinventing the wheel. After going through several of your datasets on their products, there is some cause to consider it may be closer to the latter.

As far as I know.
It's just "DSD" + Digital filtering. The power of the fpga chip determines the speed of modulation carrier signal and the accuracy of the digital filter. Digital filter is the core part of this type of design.

 

[amirm]

One thing I noted is that there is very little information from the company about what FPGA they are using and how they are using it.

He uses different versions of Xilinx FPGAs (Spartan-6, Atrix, etc.).

 

[Music1969]

One thing I noted is that there is very little information from the company about what FPGA they are using and how they are using it.

The exact model is literally right there on the Chord Qutest product website....

 

[shanecoughlan]

The exact model is literally right there on the Chord Qutest product website....

View attachment 26481

@Music1969, that gets us into the ballpark for this product, but it is not specific. See for example three variants of the XC7A1ST:
https://www.digikey.jp/product-detail/en/xilinx-inc/XC7A15T-1CSG324C/122-1927-ND/5248114
https://www.digikey.si/product-detail/en/xilinx-inc/XC7A15T-2FGG484C/122-2204-ND/5247286
https://www.digikey.jp/product-detail/en/xilinx-inc/XC7A15T-3FTG256E/122-1932-ND/5248119
These have different input/output (I/O) capabilities.

When it came to the Mojo I didn’t see any details on the site:

While it made claims to be trickle down technology from the Dave line, which gives us a ballpark of using the same line of FPGA as the Qutest:

I raised this because there is heavy emphasis by the company on their use of this technology with phrases such as “based on our award-winning proprietary FPGA technology developed for the class-leading Hugo 2 DAC/headphone amp”. Such references make me think “ok, what precisely are you using and how exactly does your specific use methodology accomplish X rather than Y.”

The former item has a ballpark. The latter item is unaddressed except in very general terms unless I missed something. This drifts us a tad too close to a type of phrasing that can illustrated by grabbing another sector. Let’s pick cloud technology: “we use AWS and our special AI to provide...” or “our Azure hosted self-healing system...” Using AWS or Azure won’t make your AI or self-healing service operate better in and of itself, and an additional proviso pops up about what specifically is great about that cloud choice (is the AI using AWS regions for high speed, does it have priority in XYZ service for computational advantages?), and what precisely *is* the AI or self-healing that makes it better than SolutionB from CompanyC.

It is a little like their reference to using “aircraft-grade aluminium” in the Qutest. Pretty much everyone else does this too...but there is no such thing as this material. We actually have a bunch of aluminum grades used in that sector. 7000 series is common, with 7075 and 7005 being two frequent examples. You can see a comparison between 7075-T6 and 7005 here. They are different

All of this is fine and it does not mean Chord is doing any false advertising or leading people up any garden paths. The Qutest obviously performs well as evidenced by the excellent ASR analysis, though the Mojo did substantially worse when it was covered, and included a heat issue that suggests an overloaded FPGA or unoptimized programming. Which is fine too, except in the context of battery life, and can help on cold winter nights As @amirm noted for that specific case “The issue with it is so much technical hype about its superiority that one is left empty after seeing performance that is well below state-of-the-art.”

I personally would love more insight from the company to satisfy my curiosity about these tools. It would be the same if they mentioned using a bespoke memory system or custom timing chips. I like this stuff.

 

[JohnYang1997]

@Music1969, that gets us into the ballpark for this product, but it is not specific. See for example three variants of the XC7A1ST:
https://www.digikey.jp/product-detail/en/xilinx-inc/XC7A15T-1CSG324C/122-1927-ND/5248114
https://www.digikey.si/product-detail/en/xilinx-inc/XC7A15T-2FGG484C/122-2204-ND/5247286
https://www.digikey.jp/product-detail/en/xilinx-inc/XC7A15T-3FTG256E/122-1932-ND/5248119
These have different input/output (I/O) capabilities.

When it came to the Mojo I didn’t see any details on the site:
View attachment 26489
While it made claims to be trickle down technology from the Dave line, which gives us a ballpark of using the same line of FPGA as the Qutest:
View attachment 26490
I raised this because there is heavy emphasis by the company on their use of this technology with phrases such as “based on our award-winning proprietary FPGA technology developed for the class-leading Hugo 2 DAC/headphone amp”. Such references make me think “ok, what precisely are you using and how exactly does your specific use methodology accomplish X rather than Y.”

The former item has a ballpark. The latter item is unaddressed except in very general terms unless I missed something. This drifts us a tad too close to a type of phrasing that can illustrated by grabbing another sector. Let’s pick cloud technology: “we use AWS and our special AI to provide...” or “our Azure hosted self-healing system...” Using AWS or Azure won’t make your AI or self-healing service operate better in and of itself, and an additional proviso pops up about what specifically is great about that cloud choice (is the AI using AWS regions for high speed, does it have priority in XYZ service for computational advantages?), and what precisely *is* the AI or self-healing that makes it better than SolutionB from CompanyC.

It is a little like their reference to using “aircraft-grade aluminium” in the Qutest. Pretty much everyone else does this too...but there is no such thing as this material. We actually have a bunch of aluminum grades used in that sector. 7000 series is common, with 7075 and 7005 being two frequent examples. You can see a comparison between 7075-T6 and 7005 here. They are different

All of this is fine and it does not mean Chord is doing any false advertising or leading people up any garden paths. The Qutest obviously performs well as evidenced by the excellent ASR analysis, though the Mojo did substantially worse when it was covered, and included a heat issue that suggests an overloaded FPGA or unoptimized programming. Which is fine too, except in the context of battery life, and can help on cold winter nights As @amirm noted for that specific case “The issue with it is so much technical hype about its superiority that one is left empty after seeing performance that is well below state-of-the-art.”

I personally would love more insight from the company to satisfy my curiosity about these tools. It would be the same if they mentioned using a bespoke memory system or custom timing chips. I like this stuff.

Why not just search on YouTube. I remember there are videos of a little bit more detailed explanation of the technology and how they do it.

 

[shanecoughlan]

All, skip this unless you are interested in FPGA rabbit holes.

Why not just search on YouTube. I remember there are videos of a little bit more detailed explanation of the technology and how they do it.

Thanks for the pointer. I found the video you referenced:

It is very interesting but it still only covers the FPGA advantage is general terms. The bias is towards having more taps, which they cite as a product of their fast crystal oscillator (separate component) and the capability of the FPGA, delivering their WTA Filter algorithm. They point out that a normal DAC only operates at a frequency of 6MHz, giving "poor resolution" while their Chord FPGA operates "around 20 times faster" at 104MHz. This is done to reproduce "incredibly small micro-details from the original recording."

Maybe kinda. An FPGA running at 20x the speed of a fixed silicon chip does not indicate being 20 times better or 20 times faster. For example, if the WTA Filter involves any floating point math then a regular chip would potentially be more efficient. A good overview of the relative nature of FPGA advantages/disadvantages is here. This overview is useful because it shows how relative the advantages or disadvantages are depending on what FPGA is being used, how it is being used, and for which purposes.

I am most interested in the first two items right now, but I can't find any detailed information, especially around the how part. If you have information please pass along a link. As mentioned above, I'm curious.

Anyway, let's move away from 'how does the WTA Filter work' and into 'what does the WTA Filter accomplish.' This is the which purpose bit. Chord appear to focus on taps (a sampling rate) because it "solves the question as to why higher sampling rates sound better.” So let's zoom into why more taps are better. As per the link earlier in this paragraph, Chord say it "is well known that 96 kHz (DVD Audio) recordings sound better than 44.1 kHz (CD) recordings [...] What is not well known is that 768 kHz recordings sound better than 384 kHz and that the sound quality limit for sampling lies in the MHz region." They slam mass produced DACs as not being able to provide the same capability.

Hm...their numbers in the video cited above contradict this statement (a normal DAC "operates at a frequency of 6MHz"). But perhaps we should read "the sound quality limit for sampling lies in the MHz region" as "the sound quality limit for sampling lies in the [high] MHz region." Unknown. That information is not presented. But I digress. Just the claim that sound quality limit for sampling lies in the MHz region takes us into a whole new territory beyond the scope of the FPGA per se. Let's just say Chord thinks their WTA filter is useful because it allows lots more sampling. Sometimes this seems to provide good outcomes (Qutest) and sometimes not (Mojo). It does not appear to be a panacea.

In the end, what I can conclude from the Chord information is that their statements ultimately mean "mass-produced DACs do not run our WTA Filter" and "we propose that the WTA Filter is a clear advantage in making better sound" rather than"the use of X FPGA is objectively better because of XYZ factors." There is a subtle difference between these two propositions, and my take from the majority of their material is that they do put some emphasis on the latter. Yet, from the information at-hand, if they had the distribution scale, there is a real possibility that a mass-produced DAC with the WTA Filter would be faster, more effective and less power hungry. This does not mean I am correct. It just means this is where I got with the material I have seen.