Topping E50 DAC Review and Measurements: I Can't Go for That

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by purr1n, Nov 21, 2023.

  1. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    92,279
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    It's been a about a year since I last reviewed a Topping product. The last was the D10s (reviewed here) which was forgettable - that is I actually don't remember anything about it. I know Topping gear has next to zero interest here, but I think it's important to take the pulse of things Topping every now-and-then to see how far they have progressed. We never know, maybe one day their audio engineers will design by ears first and measurements second, rather than design purely by APx555*. I know expectations are low from everyone here; but hey, I love to be proven wrong. And in the past, I have been proven wrong and admitted it.

    IMG_1503.jpg

    The E50 DAC uses the latest and greatest ESS9068AS. I know, it's hard to keep track since not that long ago, it was the 9038. ESS's line is confusing. I'm not even sure the "next-gen" ESS9038 is really even that much better than the ESS9028. The only thing I can gather is that maybe the ESS9038 is easier to deal with? I say this because before the ESS9038, almost any ESS DAC made by a Chinese manufacturer exhibited Sabre-itis. By this I mean harsh highs, unnatural sibilance, lower treble glare and hardness, excessively sharp transients, grainy, raspy top octave, flat stage, lows with poor pitch differentiation or texture. Sabre-itis would manifest as one or more of these elements. The advent of the ESS9038 changed all that. This is when Chinese audio companies started making ESS DACs that didn't sound half bad. The Sabrey sound didn't totally go away, but they were much toned down. The story I had heard was that ESS engineers knew about the sonic f'ckery of their chips and worked hard to fix that for their next-gen chips.

    Still, the interesting thing is that Western manufacturers such as MOTU, Schiit, Weiss, Apogee, have seemed to figure out how to work with the ESS chips, even their old ones such as the ESS9018 and ESS9028. Phrases such as "inner warmth", "vinyl-like" have even been used to describe these DACs! If you had asked me around 2012 if there was a strong sound to ESS DACs very dependent upon the chip (as opposed to implementation), I would have answered yes. Today, I would urge readers to not necessarily jump to conclusions on an "ESS Sabre" sound.

    *I mean, how many variations of ESS DAC chip + OPA1612 can one make. This what happens when one designs purely by APx555.
     
    Last edited: Nov 21, 2023
  2. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    92,279
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Some teasers:
    upload_2023-11-22_9-28-6.png
    upload_2023-11-22_9-30-16.png
    The 1kHz tone on the analyzer looks great, but why is the SINAD so low?
    $25 to the first person who can explain why (before I explain)
    HINT: It has to do with the definition of "SINAD"


    Spectrum for 10kHz tone 0dbFS
    upload_2023-11-22_9-29-46.png

    To be continued...
     
    • Like Like x 1
    • Epic Epic x 1
    • List
    Last edited: Nov 22, 2023
  3. M3NTAL

    M3NTAL Friend

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    1,783
    Trophy Points:
    113
    Location:
    Arizona
    Maybe the new game is to see how low SINAD can go? Western People like gobbly-goop - here ya go.
     
  4. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    92,279
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I don't want to address the "SINAD" thing yet because I want one of you guys to win the prize. So let's proceed with the weird behavior with the 10kHz 0dbFS stimulus. BTW, the E50 running at 44.1kHz. Why are we seeing spikes at 2kHz and 4kHz? And two at 16kHz and 18kHz? Normally may see some "grass" in random fashion, but the ordered grass is a bit disturbing.

    Topping E50
    10kHz 0dbFS
    upload_2023-11-22_10-52-46.png

    Let's do two things: put the frequency axis into linear scale and extend the bandwidth to 90kHz
    upload_2023-11-22_10-58-36.png

    Whoa, grass galore!

    The thing I don't get is why the multiples of 2k, 4k, 16k, 18k, 22k, 24k, 40k, 42k, 50k, 54k, and so on. The 20kHz, 40kHz, 60kHz are understandable as they harmonics of 10kHz. The rest doesn't make any sense though. I thought about IMD, but 44.1kHz is the rate I am running. Ee don't see a 44.1kHz spike that would a beat frequency for IMD. (I don't think it matters with ESS because there's some internal resample to delta-sigma hybrid to one trillion megahertz).

    Sure the ordered grass is low in level, but just super curious why it's there. One can also argue that most of it is in the ultrasonic region where we theoretically can't hear, but how do we explain the likes of Texas Instruments, a manufacturer of state-of-the-art DAC chips, caring enough to address ultrasonic junk: https://www.superbestaudiofriends.o...ging-back-true-science-and-exploration.13584/

    And again, why the disparity? Audio Science Review "SINAD" for E50:
    upload_2023-11-22_11-35-48.png

    What I got for E50 for "SINAD":
    upload_2023-11-22_11-36-24.png
     
    Last edited: Nov 22, 2023
  5. Biodegraded

    Biodegraded Friend

    Pyrate Contributor
    Joined:
    May 28, 2017
    Likes Received:
    8,696
    Trophy Points:
    113
    Location:
    Vancouver BC
    SINAD low because Noise low! Edit: high. Dummy, I keep thinking of the old-school definition where low is good.
     
    Last edited: Nov 22, 2023
  6. cameng318

    cameng318 Friend

    Pyrate
    Joined:
    Jun 14, 2020
    Likes Received:
    333
    Trophy Points:
    63
    Location:
    Wyoming
    That pattern is probably the cause by bad upsampling. When you insert 0 between the signals to blow up the 10kHz signal sample rate from 44.1kHz to say 352.8kHz, 34.1, 54.1, 78.21, 98.21...kHzs shows up:

    Tall lolipops are the oringinal signal, then you insert 7x zeros between them...
    1.jpg
    and you get tall grasses on the frequency plot:
    untitled.jpg
    The other kHzs are probably just IMD with these grasses and 10kHz harmonics.

    Then it's up to the FIR filter to mow them down. The in chip FIR filters typically only have hundredish taps, so they don't want to clean it below 120kHz, or else it would sacrifics other performances. Also, the more OS it does, the more tap it needs. That's where HQPlayer and FPGA/DSP comes in.
     
    • Like Like x 4
    • heart heart x 1
    • Agreed, ditto, +1 Agreed, ditto, +1 x 1
    • List
  7. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    92,279
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Will be back with subjective impressions. Need to do pre-turkey day shopping.
     
    • Agreed, ditto, +1 Agreed, ditto, +1 x 1
    • List
  8. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    92,279
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Anyway, let's proceed with subjective impressions which I know what most of you are interested. I stacked up some similarly priced DACs to serve as baselines for comparison. I will be using the SE outputs of the E50 because I feel this is the most common use case at this price point. (Heck, most of my own headphones have not been re-terminated to balanced.)

    The E50 actually sounds fairly decent. It will do the job for most newcomers. However, the competition is better. We are talking about small differences which may be more or less discernable to people depending upon downstream gear and individual ability. Also, I can see some people actually preferring the sound of the E50.

    The presentation of the E50, is "loud". What I mean by this is as if the "DR" or dynamic range score (used for recordings) for the E50 is on the low side. I don't feel an urge to turn up the volume when using the E50 as a source. The iFi and Schiit DACs in comparison sound quiet. I feel I need to crank up the volume more. (It should be noted that the output voltage level for the Schiit is the same at 2Vrms for 0dbFS). What I'm talking about is similar to when I listen to higher DR recordings. The thing people bitch about when they hear high DR recordings is that they are too quiet. The answer of course is turn it up! (to the point where most passages are moderate volume, only for the loud passages to be very loud!)

    Another way to express the E50's presentation would be contrasty. In the sense of a TV screen having it's contrast turned up so that the greys are crushed into blacks and the slightly bright parts of even brighter. This is what I mean by some people maybe preferring the sound of the E50. There sort of a fake clarity and "blackground" going on as discussed here: https://www.superbestaudiofriends.o...blacker-background-and-less-resolution.11802/

    The other two DACs, the iFi and the Schiit, there's so much going on "in between", more granularity in volume levels. Not only that, more microdetail. While the iFi uses an ancient BB chip that is outclassed by today's best chips, it still manages to pull off more microdetail than the E50. As an aside, the iFi is the most mellow and softest among the DACs here. It's also a bit cloudier than the other two. These aspects are additional considerations for an ideal component choice in a system.

    Finally there are two additional aspects of the E50 which I find subpar: bass textures and rendering of the highs. The bass seems to lack pitch differentiation. It sounds like the low notes are blobbed or smooshed, but in a temporal sense, but in a frequency sense. "One-note" maybe? With so many overtones in bass instruments (and all other instruments), it's naturally more difficult to make out pitch and textures. The E50 just seems to make this harder than with the other two DACs. (Generally, bass reproduction is where I feel R2R or multibit designs excel). In the highs, I find metal percussion, cymbals, high-hats a bit unnatural. No, it doesn't hurt my ears as with bad ESS implementations of the past, but there's excess splash that just doesn't sound right. It's as if good quality silky cymbals got turned into cheap starter cymbals. (I think the drummers here can get what I mean, and no, YT percussion reviews don't quite demonstrate how things sound in RL).

    All in all, I can see potential audiophiles just starting their journey preferring the E50. However, those who are accustomed to better gear, who want to get into the act or make a change, may prefer the other two.

    I'll slap the E50 and the Modi+ on my ABX tester machine to see I can tell the difference. I love blind testing.

    Moar measurements...
     
    • Like Like x 6
    • Epic Epic x 2
    • List
    Last edited: Nov 22, 2023
  9. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    92,279
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Some kind of aliasing?

    I do find the bad upsampling rationale hard to believe because that would be saying ESS engineers are incompetent. Yes I get it, sometimes there are bugs, but something like this with their latest chip the ESS9068 is hard to believe. Let's take a look at other ESS implementations to see if the behavior is consistent.

    Topping E50
    10kHz at 0dbFS
    [​IMG]

    Schiit Modi+
    10kHz at 0dbFS
    upload_2023-11-22_16-27-27.png

    Let's take a look at ultrasonic garbage. tl;dr: the E50 has ton of crap in the ultrasonic region.

    Topping E50 ultrasonics (to 96kHz)
    10kHz at 0dFs
    upload_2023-11-22_16-37-3.png

    Modi+ ultrasonics (to 96kHz)
    10kHz at 0dbFS
    upload_2023-11-22_16-41-26.png

    So I don't know what's up. Is there something messed up with the ESS9068? Or did Topping engineers not follow the liner notes in the ESS datasheet, or do some weird stuff to make "AmirNAD" or SINAD at 1kHz look better at the expense of other maybe just as important measurements?
     
    • heart heart x 3
    • Like Like x 1
    • List
    Last edited: Nov 22, 2023
  10. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    92,279
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Since no one has provided the answer...

    SINAD, SINAD, SINAD, SINAD, SINAD. What exactly does this mean. There is no such thing as SINAD. Well, not unless it's qualified under the parameters of how it is measured. The difference in this case comes down to bandwidth.

    upload_2023-11-22_17-10-54.png

    A legit argument would center around whether ultrasonics should be considered. I've already pointed out that Texas Instruments seems to care. Heck, TI fixed the ultrasonic junk in their SOTA TOTL D-to-A chip and even asked Schiit to give them another chance! This is an argument for another thread. Or maybe more research is needed. Oh wait, research? What's that? Maybe I should demand that you bow to my "truth" with a Pink Panther figure!

    Anyway, cool Atomic Bob Distortion Surface or ABDS™ coming up.

    P.S. You guys know I could give two hoots about measurements at the scale of Ant-Man; but I figure, if you want to live by measurements, then you should certainly be willing to die by them.
     
    • Like Like x 4
    • heart heart x 1
    • Epic Epic x 1
    • List
    Last edited: Nov 22, 2023
  11. cameng318

    cameng318 Friend

    Pyrate
    Joined:
    Jun 14, 2020
    Likes Received:
    333
    Trophy Points:
    63
    Location:
    Wyoming
    Ah yes, aliasing was it. I forgot the word. The image I showed in the last post was aliasing without filtering. 34.1, 54.1, 78.21 was really prouncing in the measurement too. 2kHz and 4kHz could be the IMD of these frequencies and 10kHz harmonics.
    upload_2023-11-22_10-58-36.png

    From ESS9069 datasheet page 33 I got this:
    Screenshot 2023-11-22 6.56.48 PM.png
    So this filter only knocks down the aliasing by 120-140ish dB, other filter options were only 100dB ish. It's not going to show up anomalies in many situations. TBH this filter is probably one of the best among built in DAC filters. It cuts off pretty close to 22.05kHz, while the older DAC filters were bit worse. I bet ESS wants to build in more taps in their filter, but it just takes too much computing power in an all-in-one chip. The computing bits might matter too, it would be a luxury to have 64+ bits.

    Schiit products uses their megacomboburrito filter in their DACs, not sure about the Modi+. But if I reverse engineer my Gungnir Multibit by my ear, the filter is probably around the magnitude of 10k taps and linear phase. That's plenty of taps to do any sharp and deep cuttings. The analog filter stage probably cuts deeper than the typical ESS9018+OP1612 designs too.
     
    Last edited: Nov 22, 2023
  12. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    19,630
    Trophy Points:
    113
    Location:
    On planet
    By now everyone here should know how I feel about SINAD or THD+N reported as a single number at a single operational level and frequency as summarized here
     
    • Like Like x 2
    • heart heart x 2
    • List
  13. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    92,279
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Topping E50 Distortion Surfaces
    • SE output
    • 100k-ohm load
    • 90kHz bandwidth
    • APx555 "cheat" HPSA mode on
    • ASIO -> USB (AP being sucky for making us rely on an ASIO wrapper)
    H2 - second harmonic distortion product
    upload_2023-11-22_18-47-42.png

    H3 - second harmonic distortion product
    upload_2023-11-22_18-52-30.png

    H4-10 harmonic distortion product
    upload_2023-11-22_18-53-5.png

    Analysis of above:
    • Distortion very low!
    • There's tendency for distortion to rise toward the higher frequencies, especially at higher output levels -6dbFS and above. Not sure if this is a good thing.

    Distortion Surface Notes:
    H2, H3, and H4+ above do not include noise
    H2 and H3 tend to be dominant over the higher orders, additional graphs will be added if this is not the case
    H2 and H3 do sound different from each other, as to how, that's another discussion
    H4+, the higher order distortions can be considered to be "crap" factor
    There is thought that higher order distortion may have an outsized negative affect on sound quality
    THD+N does include the noise component as well as THD (sum of H2, H3, and H4 and above)
    Distortion is signal that is not supposed to be there
    H2 or second order at 100Hz means there is a signal at 200Hz that is not supposed to be there
    H3 or third order at 100Hz means there is a signal at 300Hz that is not supposed to be there
    This graphs tells us how much the distortion is at that harmonic
     
    • Epic Epic x 3
    • Like Like x 2
    • List
    Last edited: Nov 22, 2023
  14. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    92,279
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    What? Blasphemy! Hail Amir, I mean Sauron, the Great, Lord of the SINAD, Lord of Mordor!

    More seriously. WTF? What's going on just under 20kHz? See THD+N distortion surface below. That I did not expect. The regular "grass" doesn't explain it. There's something else going on.

    Topping E50
    THD+N Distortion Surface
    upload_2023-11-22_19-3-31.png

    upload_2023-11-22_19-6-30.png

    Let's just double check on the spectrum analyzer. Maybe these stepped distortion frequency sweeps are not settling correctly! (It happens). Ruh Roh! This ain't good.

    Topping E50
    19kHz stimulus at 0dbFS
    upload_2023-11-22_19-10-26.png

    Could this explain the unnatural excessive splashy metal percussion that I was hearing? There should just be a 19kHz signal, but why is there also a highish level signal at 25kHz? Considering how Topping products are supposed to be designed by the APx555, I just can't go for that. No can do man. No can do.

    Topping marketing material for E50 bragging about "SINAD"
    e50.jpg

    Womp Womp!


    Could I have received a defective unit? Or a unit with the wrong capacitor values in the analog reconstruction filter? This wouldn't be the first time I've gotten defective gear from low-end Chinese manufacturers. I swear, I don't know why this happens to me. https://www.superbestaudiofriends.o...e-douk-of-wishful-thinking.13953/#post-419513
     
    Last edited: Nov 22, 2023
  15. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    11,811
    Trophy Points:
    113
    Location:
    Philippines, The
    Having 0dBu in front makes more sense for readability but you reversing the z axis just now made me question whether the "distortion very low" bit was sarcasm or not, haha.

    Distortion surface is new but the ratio is a bit more familiar. Literally cannot recall having ever seen that before.

    I'm ignorant so gonna leave it to the folk that actually know what they're talking about to opine on what might be going on here but at least a few of the folk I have in mind are MoT and would be crass to chime in. Sigh.
     
  16. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    92,279
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Topping is welcome to chime in or suggest what I could be doing wrong. I’ve done this for Grace. I will do it for any manufacturer. I am also willing to send this unit back to Topping (at their expense - they send me a DHL return slip) for them to analyze. If there is something wrong, they can send me a non defective unit and I will remeasure and reassess. It's about being fair, knowing that I am human, and that I can mess up.

    For those who feel I am being biased, let me remind users that I shit on the Schiit MIL DAC which had its own ultrasonic distortion issues in the highs:

    From a subjective point of view, the Yggdrasil MiL is in my opinion, the worst of the Yggdrasils in that it lacks the emotion, lacks the resolution, and imparts an odd spittyness to the highs (which attempts to component match with laid-back amps and headphones could not resolve)...I feel the MiL is a poor value for its subjective technical performance.
     
    Last edited: Nov 22, 2023
  17. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    92,279
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Neither have I. However, after more investigation. Turns out I had left the DAC in Ffilter 2. Filter 2 helped ameliorate the unnatural highs (not by much), but it is also the one with the measurement defect. Filter 3 did so as well but sounded rolled, which is surprising given that it's minimum phase with a sharp rolloff. I've said time and time again, I wish DACs made by entry level DAC chinese manufacturers would just cut it out with the filter choices and select the one that meets their subjective or objective vision.

    Filter 1 "Linear sharp" (per manual)
    upload_2023-11-22_20-38-12.png

    Filter 2 "Linear slow" (per manual)
    upload_2023-11-22_20-38-42.png

    Filter 3 "Minimum sharp" (per manual)
    upload_2023-11-22_20-39-46.png

    Let's take a look at the filters next...
     
  18. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    92,279
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Filter 1
    upload_2023-11-22_20-46-31.png
    This looks like the typical linear phase sharp rolloff filter.

    Filter 2
    upload_2023-11-22_20-47-16.png
    This is the slow rolloff filter. It starts early and ends late. Super wide transition band. Still the attenuation at 25kHz, where that oddball signal existed is about -36db below. Maybe this has something to do with the 25kHz signal around -36db with the 19kHz tone stimulus. However the question looms: why is there even an extraneous 25kHz signal in the first place!

    Filter 3
    upload_2023-11-22_20-53-14.png
    I was wondering why this sounded more rolled than the others, especially that it's "minimum phase fast rolloff" as described in the manual. The transition band is definitely more narrow than Filter 2. With respect to more "rolled than the others", the proof is in the pudding. Look at the attention in the audio range, say 20kHz. Filter 2 despite having a slower roll off is at about -46db or 6db lower at 20kHz. Filter 3 is -53db or 13db lower. Despite having narrower transition band and faster rolloff, more of the transition band of Filter 3 is inside the audio band.

    The filters do make a subtle difference in rolling off the highs, but they don't really affect the timbre much. The E50 rendering of metal percussion still sounds somewhat wrong.
     
  19. Psalmanazar

    Psalmanazar Most improved member; A+

    Pyrate Slaytanic Cliff Clavin
    Joined:
    Sep 27, 2015
    Likes Received:
    5,400
    Trophy Points:
    113
    -140 db is lower than any recording ever made dude.
    most of the modern on chip filters are fine. You can also program your own on the ess chips and screw up or make one that does the job you like the sound of better
    they are inadequate on cheap multichannel chips like these https://media.digikey.com/pdf/Data Sheets/AKM Semiconductor Inc. PDFs/AK4413.pdf which is why rme used the short fir filter for low latency followed by an additional iir filter. phase shift but better attenuation

    or codec chips
    https://www.akm.com/content/dam/doc...udio-codec/ak4621ef/ak4621ef-en-datasheet.pdf

    topping might be doing shitty src outside of the chip. that can vary a lot in quality. see tests here
    https://src.infinitewave.ca/

    it could also be other garbage not the anti imaging and aliasing filters
     
    Last edited: Nov 22, 2023
  20. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    92,279
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    To be fair, let's take a look at ultrasonic junk with a 10kHz signal with the different filter settings. Same or similar junk regardless of settings with Filter 3 being slightly worse than the others.

    Topping E50 Filter 1
    10kHz 0dbFS
    upload_2023-11-22_21-8-23.png

    Topping E50 Filter 2
    10kHz 0dbFS
    upload_2023-11-22_21-8-45.png

    Topping E50 Filter 3
    10kHz 0dbFS
    upload_2023-11-22_21-9-10.png

    For comparison:

    Schiit Modi+
    10kHz 0dbFS
    upload_2023-11-22_21-13-52.png

    iFi Zen V2
    10kHz 0dbFS
    upload_2023-11-22_21-18-23.png
    Not as bad you you think. This is mostly harmonics of 10kHz, not weird ass stuff.
     
    Last edited: Nov 23, 2023

Share This Page