Friday 7 August 2015

MEASUREMENTS: PonoPlayer - observations & opinions...


A couple months back, I mentioned that I got myself one of these PonoPlayers. I got it lightly used locally and saved myself quite a few dollars (remember, I'm in Canada so would not be able to buy it directly unless specially shipped across the border). As you can see, I got one of the standard black models. Although the yellow one looks iconic and was what Neil Young went on Letterman with way back in 2012, black is more my style... And it looks pretty cool.

Since its availability back in January 2015, I believe most have no doubt already read a number of reviews on this device... I don't think it's so much the device itself, but the promotion of the whole Pono ecosystem by Neil Young that has ruffled so many feathers in the world of consumer electronics based on his bold claims about the audibility and necessity of "High-Resolution Audio"; claims of an audio "revolution". On one side are the technology sites who view the claims as nothing more than "snake oil", while on the other side - among the "audiophile press" - we see repeated praises of Pono "raising the bar" in its promotion of better sound quality.


In this blog, I have commented a number of times about Neil Young and his claims (see here, here). Needless to say, I have problems with his characterization of high-resolution audio as being easily audible. Enough about that... I think by now, most who have tried listening to 24/96+ music will have already realized that 24-bit resolution and high samplerates will not impart "God-like" fidelity to albums like A Letter Home even at 24/192.

Let us turn our attention on the hardware side of Pono then. The subject of today's post; the PonoPlayer.

I. General Impressions

Let me dispense with some observations quickly:

1. The build quality is OK. Reasonable weight. Made of plastic. The body feels rubberized and comfortable to the touch. On top are the two 3.5mm phono stereo outputs - one for headphones, the other for adjustable fixed line out. There's a setting to turn on balanced mode - one 3.5mm jack per stereo channel. The bottom contains the micro-USB port for charging and data transfer. Behind a somewhat flimsy cover is the micro-SD card slot for memory expansion.

2. It's not particularly rugged. Not difficult to make little "dings" in the rubberized coat and I've already scratched up the top and bottom glossy plastic just in daily use putting it in my pocket. Simple 3-button plastic controls (volume up, volume down, and a general purpose on/off and play/pause round button) - large enough and shaped for convenience.

3. Triangular prism / "Toblerone" shape is clumsy in the pocket but is good for sitting sideways on a tabletop with screen angled up. It is what it is... Get ready for even more scuff marks and discomfort in tight jeans :-).

4. A 64GB micro-SD card came with the unit. When connected to the computer, the PonoPlayer shows up as 2 storage devices - the main memory (64GB) and the micro-SD card. You can copy whatever music files you want over and make whatever directories you want. I like this better than the "locked in" structure of Apple music products. I have directories for DSD separate from my FLAC files for example... The PonoPlayer has been good in picking up all my music so far. It takes a few seconds to scan after copying and there was one occasion where it failed; doing a shutdown and restart fixed things.

5. Data transfer is SLOW. Copying music over to the PonoPlayer is about 4MB/s. Be patient with high-resolution content! That's about 4 minutes/gigabyte of music data.

6. The screen (2.5", standard resolution - not the hi-res screens on modern phones/tablets) is functional and touch control is reasonable. Other than the usual adjustment to learning a new GUI, I have not had any major issues. Sometimes it seems a little too sensitive misinterpreting a drag as a tap especially when trying to scroll through the album list. At the main screen, swipe sideways to go through the various tabs like "Albums", "Artists", "Revealer" (as of firmware 1.0.6), "Settings", "Playlists", and "Songs". Up and down swipes to scroll through the listings. All relatively intuitive. Double tap on the screen during playback to show extended metadata, bitrate, and resolution. I'm however not in love with the display quality. Angle of view is limited and as you can see in the picture above, even viewing slightly at an angle results in color distortion and loss of contrast.

7. I must admit that the flexibility to use whatever file format I want is nice. FLAC / WAV / AIFF / AAC / MP3 / ALAC for PCM. DFF and DSF for DSD64 and DSD128. Not unexpectedly, no support for more esoteric formats like APE or WV. Note that it cannot decode DST-compressed DFF files at this point (would be a nice feature if the internal Cortex-A8 CPU has enough computing power!).

8. The PonoMusic World software is a customized stripped-down version of JRiver 20 (Mac and Windows, no Linux so far). No ability for DSP processing with playback for example. You'll also need it to install the firmware upgrades and PonoRevealer conversion to the various file formats and bitrates. I basically did not bother to use it for music transfer as I prefer to just copy the files over myself. I already have the full JRiver 20 on my computer and the two programs coexisted without issue.

9. Battery life as indicated elsewhere isn't great. Claims of ~7-8 hours of continuous play is probably accurate on a full charge with the internal 2900mAh battery (easily replaceable; check out the teardown video, and interesting transistor array shown BTW). If you're not going to be playing music for a few days, make sure to do a shutdown to preserve charge.

10. PonoRevealer is basically a tab on the main screen. What happens is that the PonoMusic World software converts the high-resolution (eg. 24/192) file to various formats - original 24/192, 24/96, 16/44, 256kbps AAC, 320kbps MP3 - uploads these various files to the PonoPlayer, and you can then from the Revealer tab play the song and select whatever quality you want to listen to instantaneously on-the-fly. I'll talk more about this later...

11. Oh yeah, it comes with a nice bamboo box/case and the "unboxing" presentation overall is quite nice.

II. Measurements

Let's just get to it then... For today's post, I'm going to measure from the RCA unbalanced line-level output set at 100% mainly. I'm still awaiting arrival of some balanced mini-TSR to XLR cables. Because the PonoPlayer was measured extensively by Stereophile, this provides an opportunity for me to compare my results with their pro-level Audio Precision system they used to see if the results are similar. Also, I hope to add a few more observations in the process.

A. 1kHz 0dBFS Square Wave

First, let's start with the digital oscilloscope looking at a 1kHz 0dBFS square wave. With the line-level 3.5mm jack:

Here is the same with headphone output at 100% volume, no clipping of the signal at peak volume:

Channel volumes are nicely balanced.

Stereophile measured the output impedance at ~3-ohms. Ayre/Hansen further discusses this in the Manufacturer's Comments. For even more info about PonoPlayer and headphones, check out the InnerFidelity report. Interestingly I'm seeing a 1.4V peak voltage for both the line-level output and 100% volume through the headphone jack. Stereophile measured 961mV. That's about a 3.25dB gain with my unit or -3dB from the typical 2V line-level output. I wonder if there were changes made between whatever model the reviewers received and the retail sample I have here. (Maybe Stereophile was reporting RMS voltage? If so, then I'm seeing 1Vrms.)

B. Impulse Response

I'm just going to use the line-level output for this one. Using Adobe Audition CS6 for rendering.

44.1kHz impulse:

96kHz impulse:

There's the Ayre impulse response (as described in their white paper). Absolute polarity is maintained. Interesting that processing is applied to the 96kHz signal as well even though I can't imagine that this would make an audible difference at the high samplerate.

No such thing as a "free lunch" in this world of course. In order to achieve the impulse response above, frequency response and effectiveness of the anti-aliasing need to be examined. We'll look at the frequency response effect below, but here's what 24/44 wideband white noise looks like going through the Ayre filter compared to a typical linear phase 95% bandwidth upsampler (SoX -v upsampling of 44kHz to 192kHz):

As you can see, the Ayre digital filter is quite "leaky". Tons of ultrasonic artifacts seep through beyond the green 22kHz Nyquist frequency for a 44kHz samplerate compared to a normal steep anti-imaging reconstruction filter like with SoX. I would be very curious if anyone has done audibility testing with this amount of artifact (and of course do people prefer this which presumably is what Ayre must be claiming?)!

The PonoPlayer seems to be emitting a noise spike up around 80kHz also.

C. RightMark 6.4.1 Pro

Measurement hardware is the usual:
PonoPlayer --> shielded phono/RCA cable --> E-MU 0404USB ADC --> shielded USB --> Windows 7 measurement laptop

I'm just going to measure the line-level output since I'm mainly interested in the DAC performance (internal ESS Sabre32 ES9018K2M chipset).

16/44.1 kHz:

As you can see, I've compared the PonoPlayer with a number of other devices I have had the opportunity to measure over the last few years. USB DACs like the AudioQuest Dragonfly 1.2 and AudioEngine D3, along with the Oppo BDP-105 blu-ray player (USB input), Logitech Transporter, TEAC UD-501 DAC and Tascam UH-7000.

Other than the Dragonfly's oddly high crosstalk (beyond variation between reasonable analogue RCA cables) and intermodulation distortion, and the PonoPlayer's frequency response roll-off, we can say that decent modern DACs have no trouble with 16/44.

So, speaking of the PonoPlayer's high frequency roll-off, here's how it looks compared to the others:
As you can see, this is a clear demonstration of the effect of the Ayre filter in the frequency domain compared to the other DACs in this roundup. That early roll-off in frequency is of course the price to pay for the rather strong suppression of ringing in the impulse responses above. The PonoPlayer is about -0.5dB at 10kHz and if I zoom into the graph, I see that the -3dB point is at 17-18kHz based on the results here.

A couple more graphs for completeness. Notice the comparatively high intermodulation distortion with the Dragonfly; I mentioned the higher distortion numbers last year when I first measured it.

Noise Level
Notice the stronger second harmonic (120Hz peak) with the PonoPlayer. This was noted by Stereophile as well.
IMD+N (swept frequencies) - notice Dragonfly is the outlier here.

24/48kHz:

I usually don't measure 48kHz, but did in this case to see if the frequency roll-off is still there... Yes, it is of course based on the same digital filter:

Notice that going from 16-bits to 24-bit audio, we dropped the noise floor and improved the dynamic range in turn by 7.3dB. This means that the PonoPlayer is capable of ~17.5-bits of resolution. This is consistent with Stereophile's measurements as well. The Squeezebox Touch by the way looks pretty good here with very low distortion numbers but slightly more bass roll-off (-1dB at 20Hz).


24/96kHz:

Another big summary table with comparison to many other DACs previously measured. Overall the PonoPlayer fares quite well and comparable to USB DACs like the Dragonfly and AudioEngine D3. Distortion is reasonably low with the PonoPlayer.

24/96 Frequency Response. The other DAC with early roll-off (red) is the Tascam UH-7000. Not audible issue at 96kHz IMO.
IMD+N
Noise level

24/192kHz:

The numeric summary above indicates that the PonoPlayer reproduces the signal reasonably well though a step down from the others. Remember that this is a rather unfair comparison of a low power, portable device against well shielded, significantly more expensive, desktop DACs so I'd say the PonoPlayer fares quite well!

Frequency Response - not an issue for 20-20kHz.
Noise level
IMD+N

D. Jitter

16-bit Dunn J-Test:

24-bit Dunn J-Test:

Just a bit of spurious noise at the noise floor of the measurement device. 16-bit LSB modulation pattern visible through the noise floor. Don't forget that the dB scale is logarithmic and what this implies! Needless to say, I do not imagine there would be any audible issue here...

III. Subjective Evaluation

Over the last 2 months, I have had the opportunity to listen to a number of albums with my better headphones - Sennheiser HD800 and AKG Q701 - off the PonoPlayer. I also spent some time with my closed Audio-Technica ATH-M50 "studio" headphones.

No doubt, despite the objective anomalies, this digital audio player (DAP) does sound excellent... From top to bottom, it sounds tight, bass control is excellent even with relatively bass-shy headphones like the AKG. Rap tunes like Akon / Snoop Dogg's "I Wanna Love You" off Konvicted (2006) sounded thunderous when the low bass kicks in with the ATH-M50 as I would expect. The violin on Anne-Sophie Mutter's Carmen-Fantasie (1993, 2015 HDTracks 24/44 version listened to) sounded full-bodied and beautifully nuanced temporally and dynamically. Lots of details retrieved.

The PonoRevealer feature is interesting. When you upgrade the firmware to v1.0.6, it includes one track to compare - Neil Young's "Heart Of Gold". I converted a few other tracks also - Muddy Waters' "Feel Like Going Home" off Folk Singer (Classic Records 24/192 HDAD from 1999), Cat Steven's "Wild World" off Tea For The Tillerman (2012 HDTracks 24/192) and Marianne Thorsen / TrondheimSolistene's "Violin Concerto no. 4 in D major - Allegro" off Mozart Violin Concertos (2L, 2008 hi-res release). Whereas the first two are not true high resolution as they are sourced from older analogue recordings, the Mozart did originate as a DXD recording. I found that the PonoRevealer upload process is extremely slow - to the point that one is obviously dissuaded from converting too many songs! The conversion process was also the one time PonoMusic World crashed on me when transferring to the player.



So, did I hear the difference with the PonoRevealer? Probably. Basically, it is as I expected... High bitrate 320kbps MP3 and 256kbps AAC sound very good and I don't think it's realistic to say that the difference between these and 16/44 FLAC is anything close to "night and day"; maybe a tiny 5% difference if we want to try being objective about it - at least this is to my 40-something year old ears. There was I thought a more significant change between MP3/AAC/44kHz FLAC and the high-resolution 24/96 and 24/192 versions though. The high-res versions sounded to have more "air", less "dull", more "clear" (of course this is sighted listening and I would not put money down on an ABX!). But let me be clear that I think the reason is more prosaic and explained by the measurements. Basically the Ayre filter I think is the major cause of the audible differences, not 24-bits or high samplerate. With 44kHz material (and 48kHz), the filter basically rolls off the high frequencies within the audible range and I suspect younger folks with good high-frequency hearing can detect the subtle difference even easier when doing instantaneous A/B'ing between tracks. Furthermore, I wonder how much of the difference is because of the poor ultrasonic frequency suppression through the Ayre filter creating distortion. With 88+kHz material, roll-off is less than 0.5dB at 20kHz and we basically can experience a flat frequency response from 20Hz to 20kHz again. Like the difference between high bitrate lossy and FLAC, this is not "night and day" difference to my ears which peak out at hearing 16kHz pure sine wave. It sounds to me like a subtle change in tonality and precision which honestly would not affect my ability to enjoy the music.

Note that the Ayre filter frequency effect can be both good and bad. The "toning down" of the high treble could actually help improve listenability of some early "tinny" and harsh CD masterings. For example, recently I had a discussion with a friend about the superiority of the vinyl version of Bonnie Raitt's album Takin' My Time (1973). I believe a big part of this is that the early CD pressing (late 80's or early 90's) sounds way too bright; almost like someone applied pre-emphasis EQ to boost the high end. The Ayre filter on the PonoPlayer I think helped a bit in this case, making this CD version slightly more tolerable...

One thing I would be careful of is headphone selection; do not listen with "dull" headphones like the Klipsh R6. I find it odd that Neil Young would recommend these in his video interview - I think it's a reflection of his hearing that this combination isn't too dull sounding for him with muddy bass. (Yes, I do have the Klipsh R6 headphones and have listened to the PonoPlayer with it to make this statement.)

The amount of power available from these portable DAPs is understandably limited so don't expect very loud volumes with inefficient headphones. With decent DR10 recordings (not excessive dynamic compression) played at 100% volume, it's not uncomfortably loud at all with my AKG Q701's; fine for general use around the house with low ambient noise.

I did have a listen to the free Beethoven DSD128 samples from A Far Cry to make sure playback worked. Indeed, no problem and it sounds great. As far as I'm concerned, uncompressed DSD (especially something the size of DSD128!) is a non-starter on a portable device with limited storage capacity; made worse with slow USB transfer speed.

IV. Conclusions

The raison d'รชtre of the PonoPlayer is about the ability to play high-resolution material in a portable device at good quality. In this regard, it delivers on the promise. I can carry around with me 24/48+ material in my pocket and listen to all the FLACs on my music server library as is just copied over without any form of conversion. DSD is a bonus.

The physical limitations of the device have been discussed adequately already. Likewise, the feature set is very limited - nothing but a USB port and a microSD slot for any kind of communication/data transfer. It plays music. Nothing else. The PonoMusic World software will need some getting used to if you've never used JRiver. I cannot comment on the Pono music store as I have not purchased any music off there yet (nor is it available for me in Canada currently).

As for the sound, yes, it does subjectively sound good. About the equivalent of a USB DAC like the AudioEngine D3 in terms of noise level and low level of distortion (I'm not a fan of the AudioQuest Dragonfly due to some objective and subjective qualities). It's also got a good subjective "weight" to the sound when playing MP3/AAC/44kHz material (I guess this is what people associate as being more "analogue"). And it will play 24/176.4, 24/192, and DSD64/128 to boot. I do have a few suggestions which may be implementable:

1. See if there is a way to improve USB transfer speed... 4MB/s is irritatingly slow for USB2 in 2015! Probably an unfortunate limitation of the USB chipset and flash memory speed I'm afraid.

2. Let me turn off the Ayre digital filter if I want to! The filter is not transparent. As per my conclusions in the recent Digital Filters Test, choice is good. For well recorded 44kHz material, I might actually prefer the standard ESS Sabre32 linear filter than the Ayre. I suspect a firmware upgrade could do this... It would be fantastic to have the choice of a linear normal or slow-roll off, maybe simple minimum phase filter, or even the choice to turn off the upsampling altogether and go NOS for all those who want to try.  I see plenty of space in the Settings --> Playback menu for "Advanced" features :-). As a side note, I wonder if the Ayre filtering is done through software upsampling using the Cortex-A8. If so, going with the DAC's built-in filters might reduce processing demand and improve battery life.

3. Implement DST decompression of DSD64 if possible. This would be very much dependent on whether the processor is capable of this level of realtime processing. As I have said many times before, DSD data is just begging to be losslessly compressed! With compression, DSD64 files are smaller than 24/96 FLAC, and might be worth putting on a device like this to experience direct SACD-quality rips. Otherwise 24/88 FLAC conversions sound just great and won't eat so much storage.

So, there you have it... The PonoPlayer. Whether this is worth the retail asking price of $399USD is as usual in the eyes/ears of the beholder. Honestly, I'm not sure I'll be using the device all that much in my day-to-day life. In an age when most of us probably have or even depend on smartphones, music can be easily loaded on or perhaps better yet streamed. Furthermore the fact that I'm usually not going to be listening to music in environments conducive to quality audio, nor would I have the nicer headphones on while walking around, the need for a high-fidelity DAP that plays high-res files is rather limited for me (and this impression obviously also applies to Astell&Kern, Fiio, Sony Walkman, HiFiMAN...).

I'll probably do my own measurements when I get the balanced TSR-to-XLR cables for completeness to see the improvement in noise performance for myself.

One day, as we look back at the history of audiophilia, I suspect Neil Young and the Pono story will be remembered more than the PonoPlayer itself. Some will remember Pono with a sense of respect, and for others in infamy. This I believe is the result of the unfortunate polarization based on Young's evangelical claims - are you a believer, or aren't you?

Enjoy the music...

RELATED PonoPlayer Followups:
Balanced Cable Measurements
DSD Playback Measurements

Addendum: January 30, 2016
I measured the output impedance of the stereo headphone jack. 1kHz signal, 20-ohm load. I get a result of 3.3-ohms - essentially same as Stereophile's sample.

Addendum: September 5, 2016
Impedance as measured across the audible spectrum. I'm getting an impedance at 1kHz close to 4-ohms this time...

13 comments:

  1. Thank you Archimago for yet another great job. In your comments you manage to separate the hardware, i.e. the player itself, from the hype, i.e. Neil Young and hi-rez audio. Nicely done.

    ReplyDelete
    Replies
    1. Thanks Jazzfan. I try to be reasonably "hype-free" and reality focused... The only way I feel comfortable to do this and can claim to actually "know" something about the device I'm "reviewing" is to combine the objective material with what I'm hearing. After all most of the time we're dealing with "high-fidelity" which means differences should *not* be that obvious! Hopefully the dual elements of objective knowledge and subjective experience/testimony/opinion informing each other in the final impression.

      I believe reviewers should be able to incorporate both the objective and subjective to achieve understanding for themselves and (hopefully) in order to best express this to the readers.

      Delete
  2. "Interestingly I'm seeing a 1.4V peak voltage for both the line-level output and 100% volume through the headphone jack. Stereophile measured 961mV"

    The 1.4Vp (2.8Vpp) you measured is from the slight overshoot.
    The peak signal I see on the scope picture is actually closer to 1.35V than 1.4V on the scale.
    When you calculate the RMS value (for a sinewave) and use 1.36V as 'peak' voltage you get 0.961V.

    It is likely that you are giving peak values and stereophile gave RMS values (for sinewaves) and thus measured exactly the same values.


    That rolled off Frequency Response (for redbook and 48kHz) is certainly audible in blind tests.
    For higher bitrates this isn't the case.
    This also explains differences with MP3 vs 96kHz yet smaller ones with MP3 vs redbook.

    Won't be buying one as most of my music is redbook based.

    ReplyDelete
    Replies
    1. Yup. The FR difference between standard and high samplerate is my suspicion for why it sounds different as well... Fascinating self-fulfilling situation where they create the difference and then say 'notice how much better hi-res is!' :-)

      I edited the text yesterday regarding Vrms. Thanks...

      Delete
  3. Good stuff as usual Archimago, 4MB/S is a bit on the slow side, did you try other usb ports...I know its a strange question but one never knows.

    ReplyDelete
    Replies
    1. Hi Rafael, yes I tried on both my ASUS motherboard's USB2 and USB3 ports as well as a nice Orico USB3 hub. At best 4MB/s consistently, at worst big fluctuations in speed. I'm copying from my SSD drive so I know it's not the computer's drive speed.

      I notice that I get consistently better speeds with the microSD slot than copying to internal memory.

      Also tried with the Surface Pro 3's USB port - 11MB/s copy to computer. Same ~4MB/s max with writing to device.

      Unfortunate.

      Delete
    2. BTW: obvious work around for the copying speed is to take the miniSD out and copy using a faster computer card reader.

      Delete
    3. Yep, the typical workaround but not very nice on the SD card slot on the long run. I use my android phone as a DAP (I can imagine people shouting "blasphemy" right now XD ) and I do it via Wi-Fi to avoid stress on the USB port of the phone, not sure about the speed but its fast enough for me.

      Delete
    4. Just tested transfer speed on my Google Nexus 5 phone copying an album from the computer HD to the internal storage - 18MB/s. More than 4x faster than PonoPlayer for a device that has been out a year prior.

      This is why I'm not happy with the slow USB transfer speed!

      IMO, nothing wrong with the phone as a DAP... I love being able to stream Spotify when I'm out for a jog which is when I would mostly use a DAP anyways. Life is busy and (unfortunately perhaps) mutitasking is an essential aspect of daily life. As much as I don't like the intrusion, sometimes important phone calls do come through even during a nice jog.

      Delete
  4. Great Review!

    The curves of the impulse response and frequency response looked familiar and I found the white paper from Ayre (the hardware designer) that describes their general filtering philosophy:

    https://www.ayre.com/white_papers/Ayre_MP_White_Paper.pdf

    Looks like that early slight treble rolloff is to reduce the post-ringing of their apodizing filter - and they thing the small reduction in treble, is preferable to having large digital artifacting with an apodizing filter. Goes without saying they feel the pre-rining is entirely unacceptable.

    I own their C-5xeMP, so I tendto agree, and in a lot of ways the Pono is a portable Ayre for this very reason.

    But ... I also find that with even decent headphones (AT50) it is only incrementally better than an iPhone (it is better in clarity and definition, but not in a big way - the Apple product is decent sounding) in sound quality, which I think ultimately limits its appeal.

    I own it because it will play FLAC up to 24/192 easily. And it *does* sound better than the mass market stuff.

    ReplyDelete
    Replies
    1. No Apodizing: PS, the Ayre filter is no apodizing (and I haven't read it in the Ayre paper that it should be (though I have read the paper a longer time ago)). A apodizing filter, would have a suppression at FS/2 of infinite (at least 98 dB for 16 Bit) and the soft Minimum Phase filter of Ayre is not that kind of type.

      Delete
  5. chicken and egg ? is that "large" preringing artefact really more audible than the treble rolloff ? i doubt that. what are we actually hearing with this kind of filter ? If you like it good for you but is it really "better" meaning to actually reproduce the recorded signal as acurately as possible ? just my 0.02$ I do have meridian gear and they to like this apodizing filters or sell it to us as the latest and greatest . I'm just a little bit reluctant to accept this as progress ? But what to sell when nothing really new is happening in audio , our ears are what they are and we have dozens of formats that excedes our abilities ? so how to promote interest , the new DSD fad that really does not do anything for us ?

    ReplyDelete
    Replies
    1. Good point Mnyb about Meridian as well. I recall back in 2014, Meridian was announced to be designing the Pono hardware but then Ayre took over.

      If Meridian had continued, I suspect we'd be looking at images of their minimum phase apodizing filter instead and discussing audibility effects.

      Delete