Saturday, 8 July 2017

MEASUREMENTS: MQA Filters on Mytek Brooklyn & Thoughts on USB Doohickeys


Hey folks... It's summer and I'm in and out of town doing a bit of R&R mixed with the day job.

I sent my friend with the Mytek Brooklyn and Fireface ADC a package of the MQA "Render" tagged files for testing. As you can see above, that's what the MQA filters look like with the Mytek Brooklyn serving as "Renderer" for the impulse responses. If you look at the top row, we see that the Brooklyn's standard digital filter at 24/96 is the same as what MQA uses for playing back 96kHz "original" unfolded PCM. This is different from the Dragonfly Black DAC presented a couple weeks back where the standard filter at 24/96 is a much sharper one and would shift to the weaker MQA filter only when it detects the MQA Render tagged data.

Here's the previously published Dragonfly Black table of impulses equivalent to the Brooklyn for comparison:

As you can see, the impulse responses are very similar. Slight differences may be attributed to the ADC used to capture them to some extent - Focusrite Forte vs. RME Fireface; both at 24/192. The overall morphology suggests very similar parameters being fed to the ESS Sabre DAC's programmable FIR filter used in both the Dragonfly and Brooklyn (as expected). We know the Meridian Explorer 2 uses the TI PCM5102 DAC chip but unfortunately that DAC does not respond to these MQA Render flags. It would be interesting to compare the consistency of the filters across DACs especially for those not based on the ESS chips when we see more Renderers released.

One very consistent impulse response is that which is used for MQA 96kHz rendering:


If there is one "typical" impulse response for MQA, it would be this one. Even though the Meridian Explorer 2 does not respond to the Render parameters in the stream, when sending 24/96 data to the DAC, it already uses that digital filter. As such, we can say that MQA Core expects the DAC to play back the unfolded data using that kind of digital filter setting for 96kHz "original" recordings.

A few months ago, I took a stab at MQA filter emulation using iZotope RX. Well, now that I've got these impulses for confirmation on the filter design, the actual filter is even less steep than I thought! Here are the parameters for "pseudo-MQA" for 96kHz playback:


So, with that, what one could do is rip the 24/96 digital output from an MQA Core decoder (TIDAL or Audirvana) for a song that is originally at 24/96. Then use iZotope RX with those settings to upsample to 24/192kHz. This final file would include the effect of the MQA-like digital upsampling filter (eg. the poor antialiasing performance and whatever time-domain benefit they might want to claim).

You can then take that 24/192 file and play it back with any reasonably accurate / high quality DAC and get a sense of what MQA would sound like with an actual MQA Rendering DAC like the Dragonfly. If you give this a try, let me know what you think! A good example on TIDAL of 24/96 "original" MQA-encoded music would include stuff like the Led Zeppelin "Masters" which I know is the same mastering as what HDtracks sells.

For those a little technically inclined, you'll know that doing what I'm describing above would not be difficult at all. In fact, I wish MQA could have just taken a few moments and provided a few post-processed files in 24/192 so consumers could hear the difference between "original" Studio Master compared with "de-blurred" MQA output with all the processing including the effect of the upsampling filters. There would actually be no need for A/B comparisons at audio shows since everyone would then be able to figure out if MQA provides any benefit for themselves at home. Let's not be na├»ve though, of course MQA would not want to do something so simple; this would only burst the bubble of the implied mystique they're apparently after with all the secrecy and hyped-up claims of sonic improvement...

------------------------

To close, I was wondering if anyone can show me a single example of where a simple, direct generic USB2.0 cable between a computer / Pi / streamer to a USB DAC actually results in compromised analogue output from the DAC.

The reason I ask is because over the last few years, there has been a proliferation of devices that supposedly clean up the USB interface by doing all kinds of things like "filtering" noise, "regenerating" the data signal with an improved clock to improve jitter (as if this is beneficial with modern asynchronous USB DACs), "isolation" apparently also removing noise in the data lines, "injecting power" with cleaner power supplies, and maybe other stuff not mentioned.

Remarkably, some of these devices even claim that with the device in place, it often "sounds like you are listening to a different DAC"! Really? Like I said, it would be nice to demonstrate just one example where there is objective evidence of such an effect with a description of the computer / streamer and DAC used. It should not be hard to show the difference especially since it could sound like a "different DAC", and differences of this magnitude would be easy to show. (As usual, subjective opinions are easy to find, especially for the faithful early adopters willing to put $$$ down.) Remember that a year ago, I discussed that even with different USB devices like the i5 NUC, ODROID-C2, laptops, or full HTPC computers, they made no difference to the DAC output. One could later add the Raspberry Pi 3 to the list.

Even more remarkable is how some people appear to be so convinced by these claims that they would even stack multiple of these devices together! For example, this reviewer admits that there are 9 devices in his playback chain to deliver the audio files between server to DAC with ethernet and optical cables in between. Some of it might be reasonable but surely it seems unreasonable to suggest all nine of these devices would be able to additively or synergistically improve the sound, right?

From my perspective, it's much more likely that one is simply wasting money without cause based on a placebo effect or simply faith in manufacturer claims. But even worse, there is the likelihood that all those devices may worsen sound quality by creating potential data error with all that extraneous reclocking / filtering / isolating going on! Would you ever consider it a good idea to hook up your USB hard drive with important data on it to the computer in this same fashion without some hesitation around data integrity or check if transfer speed may have slowed down? Does anyone really think that reputable DAC engineers these days are so incompetent as to create DACs with such horrible noise, jitter, and susceptibility to power irregularities as to require a number of third-party add-ons like this?

It has been at least 2 years now since such devices like the AudioQuest Jitterbug and Uptone USB Regen have been released. More recently we see stuff like the iFi iPurifier. Some of these like the SOtM tX-USBUltra cost more than very good DACs. In all this time, I have not seen any measurements to confirm significant benefit to audio playback from any DAC. At best, they've used USB eye patterns to somehow imply audibility (and act as if very expensive USB signal analyzers somehow are necessary for justifying the sonic voodoo being peddled). This is just as meaningless as HiFi News and their USB cable test years ago. I've even seen some folks stick these USB devices in front of S/PDIF / AES/EBU / HDMI-I2S converters like the Singxer SU-1 as if this is able to achieve anything good by the time the signal reaches the DAC!

No thanks. Personally, I just don't find this stuff "cool" nor a wise use of dollars that could be directed towards purchasing a better DAC if needed and good music among many other beneficial acquisitions and experiences. They seem to be nothing more than a psychological salve for those with unfounded insecurities (typical of those who also have a strong faith in things like expensive cabling).

----------------------

Okay, back to the summer schedule :-). Just had a great time in Arizona, Nevada, and Utah enjoying the Grand Canyon and Zion National Park with the family over the 4th of July.

Hope you're all enjoying this time of the years... And of course enjoying the music while doing so!

25 comments:

  1. Here you go
    https://www.computeraudiophile.com/forums/topic/30593-group-test-usb-gadgets-aq-jitterbug-uptone-regen-ifi-iusb30-ipurifier-ipurifier2-intona-w4s-recovery-any-more-%E2%80%A6/#comment-620472

    ReplyDelete
    Replies
    1. No, this doesn't show anything except the noise in the USB line. Given that USB is digital, it should reject noise below about -20dB perfectly. The untreated USB port in the article is below -100dB at 1MHZ (it doesn't show higher frequencies). 16/44k1 data is about 1.4MB/S (and HD even higher). The article seems to imply that audio frequency noise is a problem (since that's what it measures). It is irrelevent!!

      Ian

      Delete
    2. So the graphs are showing that there is something going on using those small USB "regen" devices, but you think the changes are irrelevant for audio signal = we can´t hear them?

      Delete
    3. No, only that we can draw no conclusions from those graphs about audible effects.

      Delete
    4. Thanks for the link Bi A.

      Interesting tests. It is good that the Regen and iFi devices can decrease the noise from the USB power lines from the poster "DM"s computer (presumably the HP TouchSmart?). But no... This doesn't mean the sound has changed from the DAC.

      A few things to consider.

      1. Just because the USB noise level is different does not correlate to sonic output. That is, one cannot conclude that the noise somehow makes its way to the DAC's analogue output which is what we're really interested in.

      2. As "Unknown" noted above, those levels of noise are not of concern in a digital system in terms of data integrity.

      3. I am a little concerned about the 'Sound' impressions being posted in the thread. The posts do not say how this was done... When testing this stuff one has to be CLEAR what DAC is being used, what computer is the source, and which headphones / speakers. For example, if these were tested using his Abbingdon AMR DP-777 DAC, then we have to realize that this is an old 2012 DAC with measurable jitter on the digital side and poor noise level due to tube stages on the analogue. It may sound good to some but clearly this is *not* a high-resolution capable player. (I've heard this DAC in a dealer's show room many years ago... No thanks.)

      4. Since DM has his 0404USB ADC connected, why did he not bother running a few tests off the DAC with the different devices? Then we'd see whether there were noise, jitter, and distortion differences. Maybe then he could explain why he likes the Intona despite poor noise rejection. Maybe then he can show us why he felt there was more "immediacy and details, the dynamic transients were larger and cleaner" with the iFi. Or "improved bass, dynamic, blacker background and improved space" with the Regen - surely there will be *some* notable change in the DAC output if even a few of these descriptions were objectively true and represent a real-world change!?

      Here's a thought...

      Maybe DM can record 1 minute of a song he thought changed in quality direct from the USB out and the same song with something like the iFi iUSB3.0 or even that Intona+iFi micro iUSB3.0 with the highest subjective score. Let's have a listen... The 0404USB at 24/96 should have no trouble capturing the sonic change from a DAC like the AMR DP-777!


      Delete
  2. Not sure about graphs though - I do use my ears. With a ifi iusb3 - sound comes alive. Just cant explain why. Even my dun give a damn wife detects the difference. Archimago - you have hear audible difference (not measurement) in your setup?

    ReplyDelete
    Replies
    1. Hi SYM.

      I have not tried the iUSB3 but over the years have listened to the Jitterbug, and have used the Corning optical USB3 cable and able to show less interference through my preamp (May 3, 2015). But I would not say either changed the "sound" of the actual output from my DAC.

      I'll see about borrowing one of these devices to test but to be honest, I cannot see myself spending say $400 on something fancy like the iFi Micro when it simply makes no sense and as far as I am concerned, the burden of proof should come from the manufacturer.

      As a side note, remember a year ago, I posted a critical article on iFi's "USB Audio Gremlins Exposed" (~March 2016) which I took down because someone complained about copyright violation (presumably the company did not like me quoting from the article). If there is no tolerance for criticism of what IMO are gross deficiencies in their claims, then how can I send any money to these guys?

      Should a person pay the guy who sells snake oil to try first before declaring that it's most likely BS? IMO, of course not. We figure out how things work. The onus of proof is on the snake oil guys. If they cannot do this, then they don't deserve the consumers' dollars.

      Delete
    2. Ah yes, the old 'veils were lifted' and 'even my wife could hear it'. This is why audio DBT results can be such fun.

      Delete
    3. @Archimago, this business of gagging honest writers for exposing snake-oil products is very concerning.

      I remember Mark Waldrep gagged by a 'cease and desist' letter from a legal firm, for writing that a listening test, by a snake-oil cable company at an audio show, was rigged.

      And now I see that you have been gagged by the mere threat of copyright violation for quoting an infomercial in an analysis of its veracity.

      I sure hope nobody gets discouraged by this. A couple of simple techniques should ensure freedom from censure, for example, even when we are certain of a fact, just write "it is my opinion that" before every single opinion, and when quoting marketing material or infomercial 'articles', quote only the segments that you wish to analyse, in short bites, and link to the whole article.

      Delete
    4. @tnargs: Indeed. "Opinions" should be fine. Yup, "infomercial" is exactly the type of article that was. Speaks sadly to the kinds of articles being posted by some of the websites out there.

      @Steven: Blind testing you say? I'm working on something that might be fun :-)...

      Delete
  3. Earlier USB DACs could potentially have performance issues as a result of the computer's USB power. Lots of thought (and money) was put into "clean" power but it was my experience that you get voltage fluctuations on USB ports that were not directly connected to the motherboard. Connect to one of the ports on the back and the problem was magically solved.

    Newer USB DACs, however, have been designed for use with portable devices and as a result do not require near the power. I doubt that issue is there any longer.


    ReplyDelete
  4. "... it often "sounds like you are listening to a different DAC"! Really?"

    All reasonable DACs sound identical because they are DACs - that's what they do. If two DACs don't sound identical then at least one of them is a long way from meeting the definition of a DAC - it's some sort of effects box. So for me if a USB doohickey makes a DAC sound like "a different DAC" it means it hasn't changed the sound at all.

    ReplyDelete
    Replies
    1. That's one way to look at a it TRA :-).

      However, we can of course see in measurements and potentially hear different noise levels, distortion, aliasing suppression, jitter rejection between DACs.

      At the end of the day, I see that there really are only 2 things a USB doohickey can do:
      1. Truly lower noise floor.
      2. Truly change time domain performance with lowering jitter.

      Hopefully the doohickey doesn't introduce any data errors of course. If #1 changes, then that means the source computer is emitting LOTS of noise in the USB data/power line and the DAC has difficulty rejecting that and allowing it to pollute the output. In my experience, it would truly be a bad DAC that allows this to happen which is why I'd like to see such a situation and avoid the computer/streamer + DAC!

      If #2 is the case, then it means there's something wrong with the DAC's USB interface and it's not truly asynchronous. I saw modifiable jitter anomalies back in the days of isochronous adaptive USB DACs and S/PDIF which perhaps a doohickey could improve on. But that was years ago and not likely these days!

      Delete
  5. I don't see a lot reviews about USB cables and gadgets etc. Like I used to. But I don't even read most of that stuff anymore. Some of the reviews are laughable. I did notice differences in sound from different cables but I attribute that to jitter. Some DACs are probably a little fussy about cables. I now use a $2 USB 3.0 cable with my DAC and it's never sounded better.

    ReplyDelete
    Replies
    1. Hi Larry,
      Yup. I'm glad to see the reduction in cable reviews and other non-sense of that sort. A few "news" outlets like Hi-Fi+ still act like the mouthpiece of the cable industry. I do hope that as audiophiles become more aware of what is / is not possible, that they will realize that expensive cables make no sense at all... And that the onus of proof is on the *manufacturer* before anyone drops a few hundred bucks or even thousands on what is likely snake oil.

      Jitter is not a characteristic modifyable by a typical 6 or 12' length of cable and certainly not an issue with modern asynchronous USB DACs or those using asynchronous ethernet inputs.

      Delete
  6. If those USB gadgets can really vastly reduce the measured DAC output noise, then it simply means that specific DAC has severe design flaw and I will sell that DAC immediately. How much was your old benchmark device, EMU 0404 USB? It sold at <$200 when it was still in the market, and it has something like 110dB+ I/O performance isn't it?

    ReplyDelete
  7. Also, I liked how you showed the MQA filters, but I don't like you use the expensive iZotope RX just for doing such things, because iZotope RX can do a lot of amazing things!

    For that kind of filter alone, it can be freely reproduced by using an Audacity plugin here:
    http://forum.audacityteam.org/viewtopic.php?f=44&t=65226

    Generated audio file:
    https://drive.google.com/open?id=0B_2hpSLtpOP8R1VHdHVJOEUwbXM

    ReplyDelete
  8. Isn't it the base of digital transmission: as soon as the S\N is good enough so the receiver can decode without error the 0s and the 1s, then the transmitted data is perfectly decoded and noise under that level no longer matters. It is what we can easily observe with "over the air" digital TV broadcasting : perfect image as soon as the signal level is high enough to be correctly decoded by the receiver, once "sync" is achieved, reducing noise or increasing the incoming RF signal level doesn't improve the image quality that is already perfect. Why would it be different for audio ?

    ReplyDelete
    Replies
    1. Hello Denis, yes and it is the same for the transmission of digital audio. Many years ago I worked on this Dataset patent as an electronics tech. My job was to tune the circuit design and watch the “eye pattern” for best SNR and minimum jitter to be able to transmit/receive the digital signal as far as possible before the Bit Error Rate became too high and the eye pattern collapses.

      For folks that want further understanding, this application note on Understanding Eye Pattern Measurements is as good as any to understand eye patterns and their relationship to the transmission of digital signals. While measuring eye patterns during the R&D phase of a project is important, once maximum performance of the digital transmission has been achieved, then it is simply a pass fail test called a Mask Compliance Test for production units. See page 16 and Figure 30 in the article I linked. As one can see, if the eye pattern loses no samples, i.e. no samples in the keep out area, it passes the test. It is that simple.

      As far as digital audio is concerned, it is exactly the same. If there no dropped samples, then one is getting bit perfect transmission. If there are dropped samples, then one hears that as literally a drop out or stutter or click in the audio. Other than that, it has no other audible effect on the digital audio signal.

      Delete
  9. Hi Archimago. I just wanted to say keep up the good work, I look forward to every update on the fine blog. One post I've been waiting for is the iPhone lightning to headphone adaptor measurements. , I'm sure you mentioned that your wife had received one with a new phone. Anyway, I've been using mine with several iems and subjectively it sounds great. I'm interested to see if it is punching above it's weight and if such a small device can be a transparent dac.
    Thanks once again for your fantastic blog, James.

    ReplyDelete
  10. Speaking of filters, I have read various blogspots of yours that show different type of filters but I don't understand the different types.
    The User's Manual for the Mytek Manhattan II has various options for PCM and DSD filters...

    8.2.11 PCM FILTER shapes
    • FRMP - fast roll-off, minimum phase filter
    • SRMP - slow roll-off, minimum phase filter
    • FRLP - fast roll-off, linear phase filter
    • SRLP - slow roll-off, linear phase filter
    • APDZ - apodizing, fast roll-off, linear phase filter
    • HBRD - hybrid, fast roll-off, minimum phase filter
    • BRCK - brickwall filter

    8.2.12 DSD FILTER (DSD Filter Bandwidth)
    • AUTO – the filter is selected automatically depending on DSD rate: for DSDx64 - LO, DSDx128 - MED, x256 – HI. It is highly recommended to leave this option enabled.
    • LO - 47,44 kHZ IIR Filter
    • MED - 60kHz IIR Filter
    • HI - 70 kHz IIR Filter

    ReplyDelete
    Replies
    1. Addendum to my recent post...

      Maybe this would be a worthwhile topic for your readers?!

      Delete
    2. Yes, a sort of beginner's blogpost regarding filters would be greatly appreciated :) I've read descriptions of various filters here and there, but I'm puzzled about how they affect sound. That way I can understand the various filters in HQPlayer :p

      Delete
    3. It might also include a Pros/Cons if that is not too subjective?

      Delete
  11. Hi Archimago, when considering audibility of artefacts in a sound reproduction system, the dB is the standard unit of measurement, so would it make more sense to view impulse response graphs with a dB vertical scale? (Need to take absolute value, of course.) Then one could infer audibility of transient ringing in terms such as “at +1 ms, it’s 43 dB down”, etc.

    ReplyDelete