Saturday, 23 November 2019

MEASUREMENTS: Yeeco TI TPA3116 Class D Amplifier (aka Nobsound NS-01G, Douk Audio F900S, Mogu F900S)

As I posted back in June of this year, we can get inexpensive little amplifiers these days based on Class D chips such as the TI TPA3116D2 (first released in 2012) which is what is inside the device here:

Yeeco TPA3116 amp being probed... Notice that input is through the AUX phono jack in front from the RME DAC.
As you can see, I have a "Yeeco" branded device but these are produced in China and will come in a number of brands like the Nobsound NS-01G from Douk Audio, and other names like Fac"Mogu" discussed in this review and here's Z Review with an older version.

I mentioned in the previous post that once I have the amplifier measurement rig going I was going to have a look at the output from this little amplifier. I figure this would be a good device to start with using the MOAR amplifier procedure described last time.

So what does US$30 buy us in terms of sound quality in 2019?!

I. Basic Amplifier Characteristics

For these tests, I'll use the 12V/5A "wallwart" SMPS that comes with the amplifier package as shown in the previous post. It's certainly not the most beefy or heavy-duty power supply I've come across, it's just what most purchases will come with.

Using my oscilloscope to look at Vrms output from this device, I measure an Amplifier Gain of 25.4dB with the volume knob turned to 100% when fed a low voltage 500Hz signal. The TI datasheet shows that there are a number of settings for the amplifier chip from 20 to 36dB. This result suggests to me that it's internally set to 26dB gain.

Here's what the Damping Factor looks like with my 7 standard test frequencies:


Interesting. At low frequencies, it maintains itself around 16-17x but the output impedance rises quite significantly after 1kHz, hitting a very low 1.6x damping factor at 20kHz!

Let's now have a look at the Frequency Response of the amplifier across a 4Ω resistor and the Sony SS-H1600 speaker:


We see that the frequency response dips down to around -2dB at both extremes of 20Hz and 20kHz using the resistive load. Reasonably flat from 100Hz to 4kHz. As a result of the diminishing damping factor in the higher frequencies, coupled with the Sony speaker load impedance increasing from 1kHz+, we see a concomitant "lift" in frequency response. This could be an example of why some have complained of "harshness" over the years with lower end Class D amplifiers with poor damping factors. Frequency response increases from just before 20kHz and up with those Sony speakers!

While we're looking at the frequency response, we can have a quick peek at the phase response:


Clearly the flat phase response of the DAC output as shown previously has been affected passing through this inexpensive Class D device. I've already measured Class AB devices where phase remains relatively flat compared to this. It will be interesting to compare this with better Class D amps in the days ahead.

II. Single-Tone Harmonic Distortion and Noise

While for most devices, my standard amplifier measurement protocol will use an input voltage of 1Vrms from DAC to amp, this amp's AUX input performed best at around 0.5Vrms from the signal generator. So as not to disadvantage this amp and report on the best score I can get, I therefore kept input voltage to that lower level.

Okay... Let's look at a few 1kHz THD(+N) FFT's:


Click on the image to zoom in for the gory details. As you can see, this isn't pretty for a simple 1kHz signal! I had a look at the FFT either as left channel only or stereo and there was no difference to be found for the above output levels. The distortions are not primarily power supply related limitations.

I tried to run the REW "Stepped Sine" test to check out Harmonic Distortion vs. Frequency. Alas, I tried a few times but kept getting bad data. Nonetheless, we can still have a look at the harmonic distortion components across the audible spectrum using the sine sweep at 2V into 4Ω (1W).


Basically we see relatively high amounts of distortion. Similar to the 2V (1W) THD+N FFT above, we see a predominance of the 3rd harmonic being stronger than the 2nd harmonic for much of the audio band.

Here's what the THD+N vs. Voltage output looked like to correlate distortion and power - both channels driven with 0.5Vrms input from the DAC at 1kHz (no benefit if I fed signal into single channel). The device went into protection mode shortly after 8Vrms into 4Ω (16W). Not too surprising I suppose given that the amp is fed by a 12V rail. Though not tested, the amp should be able to provide a bit more power with a higher voltage power supply.

For ease of comparison / conversion, I've included the equivalent % figure in red beside the dB value.
The graphs for THD (not shown) and THD+N were practically identical as most of the distortion is harmonic. It performed at best only around -60dB (0.1% THD+N) at 2Vrms (1W) into 4Ω. The distortion hits 1% by 12W (~7Vrms) into 4Ω.

I tried out my simple Crosstalk procedure using a 300Hz and 4kHz tone into either channel to look at how much "seepage" came across. This was ugly. Only about -13dB on either side and the FFT looks like that of a terrible two-tone intermodulation distortion test!


This result suggests that the amp has a great deal of trouble reproducing signals that are very different in each of the two channels despite much lower harmonic distortion when just playing a single tone together as per the THD+N graphs above.

Let's move along to multitone tests... :-)

III. Multitone Testing - Intermodulation Distortion and Triple-Tone TD+N

Okay, check out these IMD test results - all signals at 2V output into 4Ω:


Again, not pretty. Calculated intermodulation distortion of -46dB SMPTE, -52dB CCIF/ITU-R, and -52dB for the Linkwitz tones.

Here's the TIM test tone (1kHz square, 12kHz sine, 96kHz bandwidth), 4.47Vrms output level into 4Ω load:


We can see the 2kHz sidebands forming around the 12kHz sine wave. Considering the distortions elsewhere, these sidebands are not terrible in that they're down at around -74dB or so in relation to the 1kHz square wave peak.

Finally, here's my Triple-Tone Total Distortion and Noise, at the standard 2V output into 4Ω load:


We can see various harmonic and intermodulation distortions in that FFT. TD+N result of -51.5dB. It could have been worse :-).

IV. Square Wave and Wideband Noise


Remember that the 1kHz square signal above is being formed digitally from playback of a "non-aliasing" 24/384 square wave. While not as precise looking as the signal direct from the DAC, the general morphology looks OK, decent bandwidth. Notice that the right channel (CH2) appears to be slightly louder than the left channel.

Using the Rigol oscilloscope, here's a rough look at whether there's much ultrasonic content up to 1.5MHz while amplifying a dual-tone signal with 5kHz and 93kHz components (24/192) at ~2V into 4Ω:


Although the resolution is not great, we can see that there are some higher level ultrasonic peaks particularly around 400kHz, 700kHz, and 800kHz. Presumably a significant amount of this represents ultrasonic amplifier switching noise. Amplitude isn't particularly high.

Finally, before leaving the oscilloscope, let's try something "fun" given that this is a Class D amp and we see evidence of the ultrasonic noise above. We can have a peek at a lowish amplitude 100mV 1kHz square wave comparing the DAC's output directly, and what comes out from the Class D amp into the 4Ω load as visualized through the oscilloscope:


The Rigol is running at 1Gigasamples/s, 8-bits capture only. Notice the amplifier's output "fuzz" at the plateaus of the square waves. A combination of the generally high noise floor and Class D high frequency switching noise. We can compare this with other amps in future measurements.

V. Other Stuff: USB, Bluetooth, Battery Power

Up until now, we've been looking at the AUX analogue input to the Yeeco amplifier. In this last measurement Section, let's have a quick look at the Yeeco's other inputs - notably Bluetooth wireless and the simple USB input.

Starting with Bluetooth (4.0+EDR), here are the 1kHz sine and CCIF/ITU-R IMD tones at 2Vrms out into the 4Ω load sent from my smartphone, a Samsung Note 5:


Remember that Bluetooth is lossy and I assume it's simply using the default SBC codec. Looking at the 1kHz THD(+N) graph, we see that the DAC is running a bit slow... Instead of precisely 1kHz, the output frequency is closer to 999Hz. The THD+N of -56.1dB through Bluetooth is a bit worse than using the RME DAC (-60.7dB THD+N as above in Section III).

Notice that the IMD graph and measurements are messed up! Due to the imprecise DAC inside the Yeeco, REW is not locking onto the frequencies accurately. Since I am using the same FLAC 24/96 test file as with the RME DAC, Bluetooth must be also resampling the data presumably to 48kHz which in itself may (will) contribute to some loss in fidelity.

Again, using my Samsung Note 5, this time with a USB adapter and USB Audio Player PRO, I played the test tones into the Yeeco over USB:


The USB input only handles 16/48 so the test tones would have been downsampled from 96kHz and the 24-bit data also been reduced to 16-bit (don't know if properly dithered or just truncated). The THD+N is no better than Bluetooth, and again, the clock is inaccurate, this time running a bit too fast such that 1kHz became 1.001kHz and throwing off the expected IMD reference tones and where the intermodulation products should be.

I think it comes as no surprise that we're obviously not looking at a particularly high quality DAC inside this inexpensive device!

In my post in June, I talked about listening to this little amp battery-operated with the Talentcell 12V/8300mAh lithium power pack also. So, compared to the inexpensive 12V SMPS that came with this device, did the battery pack perform any better?


While no worse, we are seeing <1dB improvements with the 1kHz sine THD+N and essentially no difference with the ITU-R intermodulation signal. Compare with Section II and III above for the respective measurements using the stock 12V SMPS.

Here's the TIM test signal comparison with 4.47V amp output into 4Ω:


No difference; not that we necessarily expected anything.

Likewise, here's a comparison of the Triple-Tone TD+N using the 12V SMPS and the 12V lithium pack:


Nothing impressive here. Only -0.3dB "improvement" with the lithium power pack. For an inexpensive amp of this output quality, there's no reason to think that we would need a particularly good power supply. I explored higher output levels and I did not see much difference in THD+N between the 12V SMPS and lithium power pack up to around 10W.

VI. Impressions and Conclusions

I think it's quite clear that the objective results demonstrate that this device is not "high fidelity" by any means. (Again, I trust this comes as no surprise! :-)

I basically said as much back in June when I spoke of my subjective listening impressions. While the little Yeeco amp is able to produce a reasonable rendering of the music as to be enjoyable, the sound quality is compromised. Indeed the noise floor is not impressive hence the audible hissing I reported. I mentioned that the bass was a bit light which could be a result of that low frequency dip in the frequency response. What is likely more important is the low damping factor in the high frequencies which can result in an accentuated treble lift in some speakers (including the little Sony bookshelf measured). This would create the impression of both a sense of relative bass weakness and the accentuated treble could sound harsh with certain speakers. Perhaps this was why I thought the amp sounded "airy" and thin at times (which as I mentioned could still be enjoyable with some recordings).

Obviously as you can see, the level of harmonic distortion is high. Intermodulation distortion levels likewise are poor. Other than at 1W into 4Ω, I was unable to achieve better than -60dB (0.1%) THD+N through any significant portion of the power range. With a 12V power supply, the amp reached 1% THD+N at just over 12W into the 4Ω load. Clearly, it's unrealistic to be advertising this as a "50W x 2" or "100W" amplifier especially when paired with a stock 12V power supply. No surprise that advertising departments tend to take liberties with specs on inexpensive devices like this all the time. Sure, one could try a higher voltage 19V laptop A/C adaptor (I have a generic one to try out but could not find the appropriate plug at this time) and even up to 25V (I'd need to open up the unit to confirm that the capacitors can handle this) if one needs a few more watts - no guarantee that this would reduce distortion or if the heat sink would be able to manage.

I was surprised by how atrocious crosstalk is with only -13dB between the two channels. Not good for creating a precise sound stage especially with that relatively high noise floor. Again, maybe this can be pleasant with more "airy" new age music!

For those who believe that clean power supplies can make a difference to devices like this, the evidence really isn't all that supportive of any difference comparing the inexpensive 12V stock power supply and the use of a lithum power pack (a comparable Talentcell battery pack was demonstrably better in my previous post looking at DAC/ADC noise performance). We can expect a higher voltage power source will increase usable power output. The limitations of this Class D chip amp simply cannot be substantially overcome with the power supply used.


As I noted at the start, it's good to have a look at an inexpensive device like this in 2019 to examine fidelity at the extreme low end of the price range. IMO, correlating the objective results and experiencing the sound quality can provide some context to help judge value once we start exploring devices that cost many times the price. Seeing the objective results is also good when we consider the subjective reviews out there and when people speak about their listening experience in long-running forum threads like this. As you can see, there are folks who seem to like the sound of this kind of amplifier. Obviously there will be variation in sound quality depending on different manufacturers, but I believe core sonic qualities will be similar to what I've measured here.

You too can fork up the $30 or so for something like this and experience for yourself whether you subjectively like the sound! Remember, distortions up to a certain amount could sound "euphonic" to some even if objectively poor on the measurement test bench. Nothing wrong about admitting that one likes the sound of an amp like this with insight that obviously this is not "accurate", "transparent" or "high fidelity".

Another reason I wanted to measure this device is because this is a Class D amplifier. We are currently well into the ascent of Class D designs and I would not be surprised if by the end of the 2020's, almost all new devices are Class D by then for many good reasons! What we see in this device is a reminder of various limitations we should be aware of - potential low damping factor, phase variation, and ultrasonic noise.

Despite all the imperfections, in terms of "bang for buck" considering the features including Bluetooth input and built-in USB DAC, these little Yeeco/Nobsound/Facmogu TI TPA3116-based devices are still quite capable packages so long as high quality sound is not the primary intent.

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

Here's a funny article courtesy of Rafe Arnott on AudioStream (a fellow Vancouverite from the looks of it).

Really? You now think that wireless digital audio transmission can sound good when it's convenient because you have a few more "audiophile" products like B&W's Formation, Devialets or KEF LSX Wireless you want (need) to sell to audiophiles?! (It's rather obvious this is the reason for the silly article.)

Sure, while it's "unthinkable" that the moon gets blasted out of Earth orbit (imagine what would happen to the people on Earth if this happened!), what's so mind blowing about the fact that wireless digital audio transmission can sound good? Haven't we been routinely watching 1080P YouTube videos and streaming hi-def Netflix over WiFi without quality loss compared to wired ethernet for pretty well the last decade?

Yeeesh... IMO, this kind of audiophile reporting is unhealthy, implying the author did not understand the technology in the first place, and the more reason to see articles (and websites) like this as nothing more than thinly veiled advertising. It's like these people are living based on the theory of a "flat earth" (because presumably that's what their untested perceptions and Industry handlers tell them) and suddenly being aware of what has been obvious reality all along when the time is right to push the products and technologies. For years now, we've known that so long as the wireless strength is good, there's absolutely nothing to worry about with data integrity, and WiFi does not worsen noise or cause distortions to streaming DAC devices. The only reason anyone would even think that there could be some kind of sonic issue is because people like Rafe (and other less knowledgeable and unperceptive audiophile writers) "reported" that they either heard a difference or believed there was one to be found without the ability to provide evidence.

Maybe some day they'll also admit that with today's technology, jitter is a toothless Boogeyman, the sonic impact of which is primarily born of their imaginations and the eventual realization that "Bits Are Bits" might result in truly blown minds. Perhaps then he'll write an article beyond Space: 1999, and move on to adult sci-fi like 2001: A Space Odyssey! :-)

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

I've been listening to Van Morrison's recent album Three Chords & The Truth (2019, DR8 - don't bother with 24-bits) this week. Man, this guy is one prolific septuagenarian with all 14 tracks of original material. Overall, a good collection of material for those interested in Van's style of gospel-tinged blues/R&B with pretty decent sound quality. Check out track 2 - "Fame Will Eat The Soul", a duet with Bill Medley for a taste.

The other new album I'm checking out is Coldplay's just-released Everyday Life (2019, DR7). Yeah, I agree with the critics, a bit uneven and a mishmash of genres ranging from symphonic instrumental to electropop to gospel to acoustic stripdown depending on which track you're listening to. On the other hand, this variety makes for an interesting eclectic mix-tape kind of experience.

As always, hope you're enjoying the music!

17 comments:

  1. Below the first graph of damping factor as a function of frequency, you mention:
    "At low frequencies, it maintains itself around 16-17x but the output impedance drops quite significantly after 1kHz, hitting a very low 1.6x damping factor at 20kHz!"

    Did you mean to say that the output impedance rises quite significantly? (Presumably to 2.5Ω at 20kHz.)

    As for the Arnott article, I just threw up a little in my mouth. Thanks Arch. ;-)

    ReplyDelete
    Replies
    1. Oops... Yup, thanks Gnu. Updated :-).

      Delete
    2. I learned all I needed to know about Rafe Arnott from this bit of nonsense.
      https://parttimeaudiophile.com/2018/04/08/review-entreq-olympus-tellus-grounding-box/amp/

      Delete
    3. "It usually starts when one realizes that power cables actually do sound different from one another". That's brilliant, thanks for the link, JoethePop!

      But it's actually sad when you realize that this is written by the same person who now runs InnerFidelity.

      Delete
    4. I remember that review from a couple of years back but had forgotten it was from Mr. Arnott. Those "Part-Time Audiophiles" are really something else. I'm struck by the size of that box and impressed by the >100lbs of Swedish dirt it must contain for merely US$9650 presumably not including shipping! :-)

      With an article/review like that, Rafe seems like one of those guys willing to say and do anything for a buck no matter how strange. I'd like to be proven wrong.

      Delete
    5. Yes Mikhail,
      While AudioStream was never that useful going when Michael Lavorgna was running the show, I think it's a tragedy what has become of InnerFidelity.

      Delete
  2. Thanks for your work.

    Do you plan to measure your Emotiva amps?

    ReplyDelete
    Replies
    1. Yup. That's coming up in the next while. :-)

      A bit at a time!

      Delete
  3. Re: Streaming Sound Quality.

    I've never bought in to all the audiophile hand-wringing about the "sound" of CD transports, or CD vs wi-fi etc. I happily moved from a CDP to ripping my CDs and streaming them via a server.

    But being human, I'm subject to the same perceptual biases as anyone else.

    So when I switched server platforms a while back, I had one of those "audiophile moments." I had been streaming my ripped CDs using iTunes on my mac to my Benchmark DAC. All was great. For various reasons, I switched to using the Logitech Media Server via a Raspberry Pie. Should be the same sound right? That's what I expected.

    Except when I switched it seemed to me I heard the sound change. Music now seemed to have been tipped up a bit in the highs, for a slightly brighter, thinner sound.
    This surprised and concerned me, as I didn't expect to hear any difference, let alone one for the worse!

    But, knowing our propensity for perceptual biases, I had a friend come over and help me blind test between my itunes/mac server vs the Logitech/raspberry pie.
    Fortunately my sources are in a different room from my listening room, which made it easy to have my pal switching from the other room, with no cues as to if anything had been switched. My pal used coin flips to determine a random switching scheme and we did a few trials. Results: Once I didn't know which server was sending the music I could hear NO difference between them. I simply could not detect which was playing with anything other than chance.

    Whew!

    Once that was done, I just stayed with the raspberry pie system listening happily, and frankly my system seemed to sound the same as it ever did.

    This is where more controlled testing can be your friend :-)

    (It's also saved me before from spending on expensive audiophile AC cables and the like).

    ReplyDelete
    Replies
    1. The #1 remedy for audiophoolery is controlled testing. ;-) I wish ABX hardware test boxes were readily available so that I could compare some devices more easily. That's a project for another day...

      As I mentioned as a comment to another blog entry, I did an extended test with a friend comparing two preamps (and later, two CD players). The ear is exquisitely sensitive to level and tonal balance differences; a well-made piece of equipment won't have the tonal balance problems unless it's a tube amp with poor damping factor; level differences can be corrected.

      I switched from a pair of large analog amps to Crown XLS series and have been delighted by the sound (they are more powerful, so I suspect they have more headroom before clipping). I really wish I could critically ABX them to prove that any audible differences are illusory, but in a tri-amped system the effort would be huge.

      Delete
    2. Right on, greg!

      When I blind tested a DAC and some CDPs I made sure to match volume using a voltmeter at the speaker terminal.

      Not all blind tests come out "negative" of course.

      Delete
    3. Yup, not all blind tests come back negative :-).

      It was IMO a relief to see that there was a difference with people able to rank the inexpensive computer motherboard as sounding "worst" in the test earlier this year for example:
      http://archimago.blogspot.com/2019/05/blind-test-results-part-3-do-digital.html

      I think it's no surprise though that more often than not, we can be quite surprised by how much the ear/brain can tolerate even if objectively the measurements are poor and also how powerful expectation bias can be!

      Delete
    4. Computer audio is limited by the effort (or lack thereof) put into the design for elimination of noise and other defects, since the base hardware is pretty good these days. I had to connect an optical-to-analog converter to my computer to get clean signal for my desktop stereo. The analog outs were full of digital noise from the CPU/memory buses; very disappointing considering it was a high-end motherboard.

      Yes, Bob Carver's early experiments with distortion and frequency response back in the 70s showed that we're very tolerant of measurably poor performance in some areas. Interestingly, Harry Olson did a lot of great work at defining the limits of audible perception prior to 1950 - not enough people are aware of his pioneering efforts.

      Delete
  4. This is good stuff! I know you can't just randomly get your hands on everything you'd like, but I would be interested in seeing tests on one of the new-ish Crown XLS power amps. I've done tests on mine and they look quite good - except I haven't yet added an ultrasonic filter to get rid of the Class-D clock leakage when measuring. They're also nice in that they have built-in crossover filters and a bridged mode which works well.

    Anyway, they seem to be good amps for the money and the sound in my system is excellent; certainly I expect they would pass a blind test versus any of my other amps. As you're surely aware, though, high power amps actually are affected by the quality of the power cord - though not in terms of sound coloration, just in the maximum current they allow the amp to draw. On my 30-amp house circuit, a cheap power cord actually drops the PSU input voltage from 120V to 110V under load and this does affect the maximum power before clipping. I wonder how many people who measure these amps overlook the basics of testing and then complain about the amp's measured performance...

    ReplyDelete
    Replies
    1. Thanks for the note Greg,
      Yeah, I'll keep my eyes open for an opportunity with the Crown amps. Certainly for the price the value looks excellent and I have heard testimonials from folks I trust that they can sound very good!

      Delete
  5. thank you, i've been struggling to find any technical info on this amp.

    ReplyDelete
    Replies
    1. A pleasure Flavio,
      This kind of device is generally not going to be on the radar of typical audiophiles! But I think very important to know how they perform... Much more so IMO than ridiculously expensive stuff that few could afford or truly be interested even if one has cash to burn. :-)

      Delete