Sonnet Pasithea DAC Review and Measurements

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by purr1n, Jun 17, 2022.

  1. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Don't get me wrong. I love hearing others' impressions especially when they are totally different from mine. It's just that these misconceptions about NOS have been driving me nuts for over a decade now. It only took an expenditure of $30k for the APx555 to finally clarify these points! Don't defend NOS via science. Just love it for the sound!

    I mean, look at me. Vinyl totally sucks balls when it comes to science and measurements. Although I would point out that despite the errors, they are analog errors and not quantization errors at the ADC, but that's another story.

    As an aside please donate. See bottom right of the Home Page. I don't ask you guys for this every post like Amir. We need money for the sake of audio science.

    upload_2022-6-20_11-13-18.png

    By the way, NOS DACs without analog reconstruction filters don't need a cap in the output stage - this could provide some benefits with respect to "immediacy". Also, impulse response with NOS DAC will likely not have the amount of ringing found from the use of linear or minimum phase filters used in oversampling DACs. Although this ringing is out-of-band, and in theory not audible, I have some doubts. People have been able to hear differences. (Heck, people here have been able to discern absolute polarity), something that I didn't think possible! There are lots of parameters that can be tuned with such filters: roll-off, transition-band width, passband cutoff, stopband demarcation, stopband attenuation. Changing these will have different effects, so it's hard to generalize.

    Foobar users can download the Resampler-V plug-in to play with some of these parameters for linear and minimum phase filters. No need for HQplayer, although HQPlayer does offer way more complex filter types.
     
    Last edited: Jun 20, 2022
  2. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    This behavior is seen on several DACs, some of these DACs sound very different from each other. I can't say I've noticed any kind of specific sound phenomena related to this. It's not new. The UltraAnalog modules back in the 90s did this, but I'm unsure if they glued their two chips the same way.

    The reason it's done is to get better THD+N and linearity toward the softer sounds. Look at the slope of the THD line from 0dbFS to -55bdbFS. If we didn't attach another ladder, we wouldn't get more than 16-bits performance. Personally I wouldn't have cared, but the world's gone ASR/DXO/Rtings.
     
    Last edited: Jun 21, 2022
  3. Armaegis

    Armaegis Friend

    Pyrate BWC
    Joined:
    Sep 27, 2015
    Likes Received:
    7,464
    Trophy Points:
    113
    Location:
    Winnipeg
    I did not follow any of the technical discussion here, but I would like to state that I really appreciate seeing the reasonable adult discussions/disagreements/etc and resolution that did not dissolve into poo flinging and fingers-in-ears arguing.
     
    • Agreed, ditto, +1 Agreed, ditto, +1 x 8
    • Like Like x 2
    • List
  4. crenca

    crenca Friend

    Pyrate
    Joined:
    May 26, 2017
    Likes Received:
    3,825
    Trophy Points:
    113
    Location:
    Southern New Mexico
    Thanks @skem & @purr1n for the discussion. It helped remind me of a few things, not the least of which is the difference between OS (digital) filter and output analog (reconstruction) filter. Your revised post makes more sense now skem - at first it seemed to me you were saying "phase" & "timing" (of upper vs. lower FR) was two different things.

    The two quotes above are Interesting, in that on the one hand you acknowledge the usefulness of linear (i.e. "in phase") OS filter yet in the case of Hotel California you prefer an out-of-phase OS solution. Perhaps the recording of Hotel California just does not have that much content above 12, 13k (besides maybe tape hiss) so the filter does not skew anything of consequence? Perhaps with this particular recording your ears just prefer a bit of out-of-phase fuckery? Perhaps the phase fuckery of a "minimum" phase filter is moved up out of the audible band when OS to 192k (Anyone know if this is so)?

    In any case the high tap linear filters typically used these days (last time I looked HQPlayer had a few typically chosen) have such ridiculously low level (and, as purr1n points out of band) "ringing" distortion I find it hard to believe it affects anything (timbre, attack, etc) in an audible way - heck the analog output stages of (chip or otherwise) DACS and amps would seem to not be able to resolve such incredibly low level differences/distortion. But...like marv points out perhaps its so...maybe it has an effect analogous to high levels of feedback, something we can hear but have no way to measure.
     
    Last edited: Jun 22, 2022
  5. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    There's no free lunch. However, one is fundamentally more wrong than the other. Minimum phase is computationally much less expensive, and thus used for real-time applications. Linear phase EQ on a digital mixing board is not a good idea, will be too much lag.

    Linear Phase
    upload_2022-6-22_12-44-19.png

    Minimum Phase
    upload_2022-6-22_12-49-57.png

    Linear phase reconstruction
    [​IMG]

    Minimum phase reconstruction (not according to original samples because of phase shift in high frequencies)
    [​IMG]

    Moral of the story:

    • Just like what you like instead of appealing to authority based on audiophilistic marketing bullshit based on misunderstanding of how math and science really work.
    • We can't cherry pick one method, praise it on its scientific merits, and then pick other method and praise that on an alternate set of scientific merits.
    • There's no free lunch. Pay on way or another way.
    Looks like Felix is back at it again, this This was six, seven, years ago. Some people scare me. This dude actually contacted Craig (Eddie Current) and asked him to join his FB tube amp group so his tube amp friends could argue and debate points of tube amps with Craig. Craig told me about this and asked me "WTF, do you know this guy, who the heck is this guy?"
     
    Last edited: Jun 22, 2022
  6. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Minimum phase, all parameters equal, tends to sound more incisive than linear phase. This can work better or not work better depending upon the DAC and personal taste. I fed the old Gungnir DS minimum phase OS with SoX, and liked what it did to the soundstage since the old DS has a compact stage. Flat out preferred it this way. No appeals to science of lack of pre-ringing, etc. Most other DACs, I prefer linear phase.

    foorbar Resampler-V
    upload_2022-6-22_13-24-58.png
     
  7. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Here's one linear phase.
    upload_2022-6-22_13-32-1.png

    Here's another linear phase were I moved up the stopband attenuation and stretched out the passband to get a nice "NOS" type impulse response. There's a penalty with HF. There's no free lunch.
    upload_2022-6-22_13-33-32.png

    Yes, all of this sounds different.
     
    Last edited: Jul 16, 2022
  8. crenca

    crenca Friend

    Pyrate
    Joined:
    May 26, 2017
    Likes Received:
    3,825
    Trophy Points:
    113
    Location:
    Southern New Mexico
    Which seems to support the rough comparison to feedback.

    Now correct me if I'm wrong, this particular distortion (i.e. "pre ringing") of a linear filter only occurs in an impulse response, 0 to 100 situation. In other words computationally it only happens from a rapid (basically infinite frequency) LSB to MSB change, whereas with in band signal (i.e. normal music) it does not occur at all (computationally), not that it is just "buried " in the signal. At least, this is what I remember ultrabike saying in his write up (could be misremembering).

    Someday I will have to play with all this again, but I don't have a NOS dac, find HQPlayer frustrating, etc.
     
  9. SoupRKnowva

    SoupRKnowva Official SBAF South Korean Ambassador

    Pyrate Contributor
    Joined:
    Sep 26, 2015
    Likes Received:
    4,249
    Trophy Points:
    93
    Location:
    Austin, TX
    I know the math states that the "preringing" you see in impulse graphs of filters is a requirement of the reconstruction, that you can't accurately reconstruct the waveform without it, and that the filter will only "ring" in the resultant output if the input isn't properly bandlimited.

    But what I dont fully understand is how that works out with the non-ideal filters we are actually using in digital devices, aka non-sinc(x) filters. In these non-ideal filters is the pre-ringing something that can actually effect the signal without aliasing being present? though I still mostly consider "ringing" in digital filters to be an audiophile boogeyman
     
  10. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    It can be argued that some bands use square wave like signals in their music.

    With respect to ringing (pre or post), it can be argued that they will be buried in real musical signals. The better augment is that this ringing is above the range of human hearing.

    This is a 1kHz output from an ESS DAC using the sharp linear filter option. Count the little upses and downses. There are 24 of them in a one square wave cycle (up and down). That's 24 kHz.
    upload_2022-6-22_14-22-50.png

    I'm not a filter math wiz. However, we can always play connect-the-dots like the Denafrips "NOS" filter.
    (from @GoldenOne)
    [​IMG]

    The part we have to pay with the connect-the-dots filter is this which is supposed to be a 15kHz sine, which looks to me like a triangle wave with some phase shift turning it into a sawtooth wave.
    [​IMG]

    "Filters" were never a thing until Meridian Digital honcho Bob Stuart started pushing it with audiophile marketing speak on how how minimum phase were better because they had no pre-ringing, and how pre-ringing was unnatural, not found in nature, made the timbre of instruments sound off, while audiophiles made up something about castanets. Meridian's house sound was kind of warm and slow, so I figure the minimum phase filters made them sound better, less sleepy. I like simple explanations.

    Audiophiles care more about filters than the DAC designers themselves. The DAC designers just pick a filter with appropriate parameters and just go with it, never looking back. Chi-Fi DAC manufacturers picked up on the neurosis of audiophiles and their insatiable appetite to constantly tweak (because most are lost in a perma-unhappy state and want to listen to filters instead of music), started providing six to eight different filters with their DACs.

    As far as Bob Stuart. Does that sound familiar? Hint: MQA.

    Remember when DSD was all the rage and how one-bit had almost infinite bits in the audio band? And then we got betterer DSD with DSDx2, x4, x256, x1024, etc. Well DSD sure seemed to go away fast after MQA came out. What about HDCD? Somebody here should ask Amir over at ASR what happened to HDCD since Microsoft bought them when he was in charge of the consumer stuff at that time. I thought HDCD was a pretty cool idea and hoped Microsoft would have done something with it.

    Please shoot me.
     
    • Like Like x 5
    • Epic Epic x 1
    • List
    Last edited: Jun 22, 2022
  11. crenca

    crenca Friend

    Pyrate
    Joined:
    May 26, 2017
    Likes Received:
    3,825
    Trophy Points:
    113
    Location:
    Southern New Mexico
    Have to be the ultimate playback chain stress test. We need a thread devoted to good castanet recordings, preferably in DSD10,862 and/or DXD cubed

    On a serious note, after going through a short "minimum phase" (via HQPlayer - in front of OS DACs which is silly because "just different") preference but after a while realizing I did not like what it did to the upper frequencies, I learned a bit more about OS filtering. For playback without DSP, some form of linear with enough taps to make sure they're ringing is ultrasonic (if you're convinced "ringing" is even relevant) seems the way to go. But all that's on paper before you get ears on something admittedly...
     
    Last edited: Jun 22, 2022
  12. skem

    skem Friend

    Pyrate
    Joined:
    Nov 1, 2017
    Likes Received:
    1,911
    Trophy Points:
    93
    Location:
    Charles River
    Soup and I have argued over the pre-ringing forever. I can’t say I hear pre-ringing *per se* but I definitely do hear a wooshing sound on piano and other percussive instruments when using most linear filters. But the tonality is much more pleasant in linear. That’s what I hear, and it seems to correlate to these technical details, but again that’s what I hear.
     
  13. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I hear this whoosing sound a lot of the time on piano. It's the sound of the dampers over all of the strings being lifted when the pianist steps on the sustain pedal. For close mic'd recordings on large grand pianos, it's more evident. Maybe it's easier to hear with linear filters because minimum phase smears the sound? Whooshing sound on drums could be brushes or the sticks whooshing while they rub against the inside of the hand if using a light grip.
     
    Last edited: Jun 22, 2022
  14. Armaegis

    Armaegis Friend

    Pyrate BWC
    Joined:
    Sep 27, 2015
    Likes Received:
    7,464
    Trophy Points:
    113
    Location:
    Winnipeg
    When I was playing around with the filter settings on the iFi iDSD Pro, I only liked the BItPerfect and BitPerfect+ filters and hated the rest. I had no idea what each filter was though.

    These were my point form notes
    Bitperfect - feels slow
    Bitperfect+ - feels better
    transient aligned - sorta harsh, like a twang drags out but also feels like a bad harmonic or out of alignment
    gibbs - wtf harsh
    apodising - wtf harsh

    vs what the manual says about each filter:
    upload_2022-6-23_0-9-28.png

    I'm not sure if that means I prefer linear phase, though it kinda sorta reads that way?
     
  15. crenca

    crenca Friend

    Pyrate
    Joined:
    May 26, 2017
    Likes Received:
    3,825
    Trophy Points:
    113
    Location:
    Southern New Mexico
    I think (someone can confirm/deny):

    The first two are iFi's version (same-but-different, probably to avoid royalty/contract/NDA) of MQA/Bob's beliefs about digital OS filtering {which is basically don't - let the images/aliasing through! ;) } . The second two are out-of-phase "minimum phase" filters, meaning low FR and high FR are no longer congruent in the "time domain". The last one is a standard (realtively) high tap linear phase with the interesting marketing name of 'Transient Aligned' seemingly saying "Hey audiophools, if you want your time domain to be in sync with your frequency domain, and your "transients" (and everything else - timbre, etc.) to sound Memorex Live, this is your filter...".
     
  16. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    It's been a long time since I posted. Hi all :)

    I agree with @purr1n, and perhaps I can contribute a little.

    On a DAC you usually have what we call a digital upsampling operation followed by an analog reconstruction operation.

    Analog reconstruction is done with a non-linear phase IIR filter filled with poles, because such is life with analog components. Severe phase and amplitude issues can happen with these filters, but mostly around their transition band.

    Digital upsampling increases the bandwidth of the signal so that the necessary analog reconstruction filter transition band can be as far away as possible from the audio band. Furthermore, it helps keeping the passband flat since the DAC usually does a sample/hold operation that results in distortion of the passband as well.

    Unlike analog reconstruction filters, digital interpolation filters in the digital upsampling operation are usually FIR filters, with no poles, that can achieve linear phase from passband to stop band.

    So you have:

    Upsampling -> DAC -> Analog Reconstruction

    When talking about ringing, the focus is on the interpolation filter inside the digital upsampling operation.

    The classic picture is that of the impulse response. The impulse response ringing corresponds to the transition band imposed to a digital impulse signal by the digital interpolation filter. As discussed, the ideal situation is for the start of the transition band of the filter to happen outside of the audio passband. The result of this is that the ringing would be ultrasonic and lower than anything in the passband of the audio signal.

    Implementable linear phase digital filters that can achieve linear phase in the passband, and fully ultrasonic ringing are well understood and readily available. They are not impossible sync filters.

    It is possible to use minimum phase filters as well. But these filters are not linear phase from passband to stop band. They will also exhibit ringing, though because they are minimum phase, the main tap is now closest to the first tap. This means that you have almost no pre-ringing and about double the post-ringing for the same filter order. The real trade of here is: very little phase distortion close to the transition band (as opposed to non if using a linear phase filter), for minimum filter output latency.

    Again... Hi all :)
     
    • Like Like x 13
    • Epic Epic x 3
    • List
    Last edited: Jul 1, 2022
  17. mitochondrium

    mitochondrium Friend

    Pyrate
    Joined:
    Sep 1, 2017
    Likes Received:
    1,115
    Trophy Points:
    93
    Location:
    A Cell
    Happy to hear the voice of reason again.
     
  18. Herman Visser

    Herman Visser New

    Joined:
    May 22, 2022
    Likes Received:
    3
    Trophy Points:
    3
    Location:
    Australia
    I am the happy owner of a Sonnet Pasithea DAC and I also still own a Metrum Pavane L1 DAC. I've had the Pastithea DAC for about two months now and had to wait close to two months for delivery. I thought I'd share some notes for prospective buyers. In my buy I considered only one other DAC, the Holo Audio May L3. I chose the Pasitihea as I really like the Metrum sound, and it had a high quality volume control built in. The cost was also about $1k Australian less compared to the Holo Audio L3.

    Metrum DAC's are famous for a very weak USB input. I learned that the hard way when I hooked up my Pavane to a Cocktail Audio X50D player, and couldn't believe that it was the same DAC. Since then I have gone straight to high quality digital sources to drive these DAC's via their I2S inputs. Currently I drive the Pastihea from the Cocktail Audio player, serving as my all in one hifi. I use a Singxer SU-6 to drive the Pavane from my gaming computer. In both cases I use Neumann KH310 studio monitors for both setups. In the case of the Pasithea, I added a subwoofer to cover the very low octaves, and that sounds lovely. I use pure silver low impedance cables (XLR's) and that is about it. I've also done a fair bit of acoustical treatment to the room my Pasithea plays in.

    The Pasithea took about a month of on and off playing to run in. This is by far the best sound I have ever owned and better than many systems I have heard in hifi shops.
     
  19. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    UPDATE: @shedaze sent the pi2AES with the RJ45 i2s. This worked without a hitch with the Sonnet Pathesia. Well, actually, I had to pop my own SD card with Volumnio into it because the one provided imploded the OS after I set networking to WiFi.

    PXL_20220716_154830474.jpg

    Using the i2s from the pi2AES, the Sonnet DAC improved most notably in transient focus. There were minor improvements in imaging (lower tendency for a three-blob or multi-blob effect, wider sweet spot), and resolution / ambient decay. The AES from the pi2AES sounded smeared and hazy in comparison to the i2s output. However, I still don't think the macrodynamics match up to DACs which I consider A-tier with this particular aspect. Part of this could be Sonnet's flavor of NOS voicing which puts a premium on the mids, rending human voice with an exquisite lushness with overpowering them with tonal density. (I think there are some side effects to this). Still I have high expectations for a $6k DAC - that is I would prefer better than Bifrost level technicalities (in addition to the beautiful mids) if I were to spend that much.

    However, the jury is still out. I don't have the magical Hydra-Z USB converter and I have no idea whether this Hydra-Z USB to SPDIF doodad is comparable to the pi2AES i2s out from the RJ45. At the end of the day, does it matter and it this even fair? From my point of view, DAC manufacturers, especially ones who sell us premium stuff, should provide all the necessary, tools, converters, whiz-bangs, interfaces, for us to get the most of them without customers having to resort to extra purchases. I've always disliked the Toyota and Porsche options bullshit.

    At the end of the day, I think it's combination of two things: I'm a cheapskate; and while I can appreciate the Sonnet DAC's one strength, I'm probably not a fan of Sonnet's house sound for the music that I tend to play.

    P.S. Bringing this to the show for DAC-offs.
     
    • Like Like x 9
    • Agreed, ditto, +1 Agreed, ditto, +1 x 1
    • List
    Last edited: Jul 16, 2022
  20. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    CLOSING THOUGHTS:

    After coming back home and comparing and thinking about NOS DACs, I'm appreciating the Sonnet Pasadena a bit more. Let's get the can-be-better stuff out of the way. This DACs greatest weakness is with the softness on attacks. However decay isn't so bad. The Sonnet will reproduce 32nd notes in speed metal somewhat competently, not in a total fail fashion like the Abbas 2.3 NOS DAC where all the notes blend together into a pulsing whole note. The Pasithea also does a decent job on microdetail and clarity, especially when the i2s connection is used with a suitable equipped pi2AES. The downside is that the AES and coax inputs seem to suck other than when driven by an OOP Dragonhyperball-Z DDC. (The USB flat out sucks).

    The Sonnet Paramecium's greatest strength is with vocal reproduction. The vocals are rendered without any grain, they are lush, welcoming, and full bodied while at the same time being without excess thickness. For example, the Schiit Bifrost 64 in NOS mode almost pulls it off, but vocals remain too thick. The Holo DACs in NOS mode sound like modern DACs and don't have that NOS density. So in this regard, the Parenthesis DAC is very special, unique. I love what it does here, although I know others like @SoupRKnowva will agree to disagree.

    Where this DAC will do well is with recordings with simple arrangements: girl with guitar or piano, sparse and light jazz, etc. I think if this is primarily the type of music that you listen to, then the Parasithia will suit you well. However, you if like pop, rock, metal, or busier country, or orchestral you may find the sound too soft. As such I still stand by my original statement that for most, the Paseithia will do as an excellent secondary DAC to complement a more well-rounded primary DAC.
     
    • Like Like x 10
    • Agreed, ditto, +1 Agreed, ditto, +1 x 1
    • List

Share This Page