Astell&Kern is, of course, best known for its fine digital audio players. But the reality is that many people use their phones to listen to music. So the company has brought its expertise to bear on providing higher quality audio than that available, normally, from phones. And since an increasing number of phones have no analogue audio output at all, an external DAC is pretty much a necessity for a music lover. That expertise is expressed in the Astell&Kern AK USB-C Dual DAC Amplifier Cable PEE51.
- USB Type-C to 3.5mm analogue stereo digital to analogue converter
- Dimensions: 137mm by 17mm by 10.2mm
- Weight: 27.3 grams
- Compatible with Android, Windows, MacOS
- Not compatible with iOS (iPhone/iPad)
- Supports PCM up to 32 bits and 384kHz sampling
- Supports DSD to at least DSD128, rated to support DSD256
- Measurements show ultra-low noise, extended high frequency performance and power to support most headphones.
- Summary: Excellent sound quality, solid high resolution audio support, DSD iffy from some sources
- Price: $169
- Available at fine high fidelity retail outlets, and direct from distributor's retail division here
All about the Astell&Kern AK USB-C Dual DAC Amplifier Cable PEE51
What’s with that product name, eh? Henceforth I shall mostly refer to it is the PEE51, or perhaps the A&K PEE51.
This is a small device. It indeed has a cable, as the name suggests, on one end of which is a USB-C plug … as the name suggests. At the other end is a small boxy thing, with interesting angles (as is Astell&Kern’s way). See the photos for details. It contains the bulk of the electronics and provides a signal to your headphones via a 3.5mm stereo output socket.
It employs two Cirrus Logic CS43198 combo-DAC chips and headphone amplifiers. Cirrus Logic isn’t quite as sexy a brand name for DAC chips as AKM and ESS, but it knows its stuff and has been used in many a high quality product, including the thousand-dollar Astell&Kern A&norma SR25 digital audio player.
Unlike some similar-sized portable DACs, this one is a full USB Audio Class 2.0 device, which means that it can go well beyond 24 bits and 96kHz, as you can see above. It will work with Android phones that support “OTG” on their USB-C connections – that would, these days, be most of them with USB-C – and Window 10 (post the 2017 “Creator’s Update”) and Mac computers. Astell&Kern make explicit that it does not work with iOS devices.
So of course I plugged it into my iPhone, just in case, via an Apple USB-to-Lighting adaptor and a USB-C to USB-A adaptor. No go. The iPhone didn’t even seem to be aware of its presence. Normally if you plug in something it doesn’t like, an iPhone pops up a message telling you the thing is incompatible.
Using the PEE51
As I’m writing this bit, I’m listening to the second movement of Beethoven’s 7th Symphony – a Viennese critic at the time wrote that Ludwig Van ought to be sent to the lunatic asylum for this work, so who says that we hifi reviewers are harsh? – through Focal Elear headphones, and it’s sounding pretty magnificent.
Not all was perfect, though. Because I’m preparing this review before the actual release of the product, I don’t yet have access to the drivers that Astell&Kern promise. I’ll update once I get my hands on them.
So, here’s the go. Both Windows and Mac reported that the PEE51 supports up to 32 bits and 384kHz sampling. And with PCM the PEE51 worked perfectly with all my test signals up to 384kHz, providing clear, reliable music.
With DSD, things were kind of iffy. With my Mac, using JRiver Media Center (on a Mac JRiver can only use DoP – DSD over PCM – not native DSD), even DSD64 wouldn’t work. Yes, some music came through, barely, over a sea of digital hash. On a Windows computer – again with JRiver – and using the built-in Windows 10 drivers, DSD64 and DSD128 sounded lovely … except for all-too-frequent dropouts. I’ve experienced this before with the Windows drivers. I expect that when A&K issues its drivers with ASIO support, that won’t be a problem. As for DSD256, the Windows driver would not accept it.
UPDATE (22 March 2021): As the struck-out bit above indicates, DSD didn't work well with the built-in drivers on a Windows 10 computer. That's by no means unusual. Astell&Kern has now provided its own drivers. As usual, these provide drivers for Windows' own direct sound, and also both ASIO and WASAPI drivers which can be accessed by quality player software. A quick install of those had DSD working perfectly. Using JRiver, if you use the WASAPI driver all DSD is routed to the DAC disguised as PCM -- this is called DoP for DSD over PCM. With this driver, regular DSD64 and double-speed DSD128 worked perfectly and sounded beautiful. No dropouts. DSD256 would not work with WASAPI. But it worked perfectly, again sounding delightful, using the ASIO driver and with the DSD output from JRiver set to Direct rather than DoP.
Regardless of mode, the volume control in JRiver was disabled, which was initially a slight concern. But it turns out that you use the Windows volume control on the task bar. No, that does not mean that somehow the ASIO data is being routed through Windows audio processes. All it means is that Windows is issuing level control commands to the hardware volume control in the PEE51.
Using the USB Audio Player Pro app on my Android phone, the PEE51 worked perfectly with DSD64 and DSD128 using the DoP system. It decided to convert DSD256 to PCM. With Native DSD selected, DSD64/128 still worked perfectly, but DSD256 just produced a nasty buzz.
There was one problem with the phone, though. The USB-C plug on the PEE51 was a touch too large to properly engage with the USB C on my phone while it was wearing its thin, included-with-the-phone case. I had to take the case off to use it.
I suppose I should talk about how it sounded. So here goes: it sounded fine. All my headphones – I tried half a dozen different ones of radically different designs – sounded just as they did with a fine analogue audio headphone amplifier. And they all went easily as loud as I could tolerate.
I ran my usual suite of tests on the AK PEE51. So let’s start with the noise performance.
The noise measured at -118.2dB A-weighted when I fed the PEE51 24-bit test signals and my computer was running from battery, not connected to the noisy Ethernet in my office. But when the computer – a Microsoft Surface Pro 2017 – was connected to both power and Ethernet, the noise level increased to-97.6dBA. This continues the trend of portable DACs not being particularly effective at stopping analogue noise delivered over the USB connection from making it into the final analogue output. Here’s the graph:
Is that connected-noise audible? Actually, no. That’s at about the same level as the noise levels inevitably associated with 16-bit audio. Speaking of which, here was that performance (overall, it was -97.7dBA):
The noise in this graph is at the theoretical minimum possible level. Sixteen bits can’t resolve anything lower in level. That’s the kind of thing to look for. Check out the green trace in the 24-bit graph above, which is the same for 24 bits.
Again with 24-bit signals, THD was at just 0.0004% -- which matches A&K’s specification exactly. Measured intermodulation distortion was also ridiculously low at 0.00062%. Crosstalk was at -111.8dB. (If you’re lucky, you’ll get a crosstalk measure of perhaps 25dB from an LP!)
Now, the CS43198 DAC chips employed by the PEE51 have five selectable filters for PCM signals. Astell&Kern has chosen one of these. Obviously. You’d need special controls to make it user selectable. But the question is, which one?
The five available filters are Fast Roll-Off Linear Phase, Fast Roll-Off Minimum Phase, Slow Roll-Off Linear Phase, Slow Roll-off Minimum Phase, and NOS. That last stands for “non-oversampling”, and in most implementations it not only doesn’t oversample, it doesn’t even employ a low pass filter. I find NOS all manner of offensive in a technical sense.
So, which one? A few measurements quickly established that. First, the frequency response at all three test frequencies went out pretty much dead flat to around 45% of the sampling frequency, and then dropped fast and hard. Checking the Product Data Sheet for the CS43198 (137-page PDF), it was clear that this matched the “Fast Roll-Off” filters. Let’s check them out. Here is the frequency response with 44.1kHz sampling – which is the one which should be most important to you, since the great majority of all your music will be in this format:
Notice: flat to pretty much spot on 20kHz. And here’s with 96kHz sampling:
Same, except extending to 41kHz. And 192kHz sampling has a slight wobble at the extreme treble, leading to an insignificant 0.5dB droop at 81kHz:
As for Linear Phase vs Minimum Phase, I captured the output of the PEE51 when sent a super short impulse (one sample at -3dB amidst total silence in a 44.1kHz file), thus:
And a 1kHz square wave, thus:
The notional ringing all occurring after the impulse and after the rise or fall of the square wave says Minimum Phase rather than Linear Phase (the latter’s ringing is symmetrical around the impulse).
I’m fairly certain that you cannot hear the difference between Minimum Phase and Linear Phase, so I’m indifferent as to which is used. The “ringing” is a necessary artefact of any bandwidth-limited signal. The name – ringing – sounds alarming, but it is only describing what you see when a signal is viewed on an oscilloscope, not anything audible.
I’m pretty confident that you can’t hear the difference between Fast Roll-Off and Slow Roll-Off either, but I lean towards the former anyway because it actually does deliver an extended, flat frequency response out to very close to the theoretical maximum available for the relevant sampling frequency.
Most importantly, NOS has not been used. Good.
You can read why I so dislike NOS, plus all about various filter settings, here.
Now, onto more important matters: will the PEE51 drive your headphones or earphones effectively? Well, given my listening impressions above, the answer is obviously yes. But let’s get a little more objective about this.
First, I measured the internal impedance of the headphone output at a touch over two ohms. Astell&Kern specifies it as two ohms, which enhances my confidence in my measurements. Now, the purist in me likes to see it as low as absolutely possible. Some devices manage around 0.5 ohms. That said, the practical effect of a two ohm impedance is effectively nil, even with low impedance headphones or earphones with an impedance of their own that varies widely by frequency.
The PEE51 was able to deliver a clean 13-ish milliwatts into a 300 ohms load, representing high impedance headphones. That means around 11 decibels higher than the sensitivity rating of the headphones in use. I’m guessing that older, very low sensitivity, high impedance headphones may have trouble reaching entirely satisfying levels. That said, I’ve just donned my twenty-something-year-old Sennheiser HD 535 open-backed headphones, and I find that The Offspring’s Americana is being delivered at thunderous levels with the volume on the Windows computer set at just fifty percent.
As for low-impedance eargear, the PEE51 produces half a volt, or somewhat more than 15mW into 16-ohm loads just before clipping. That works out to around 12dB above headphone/earphone sensitivity.
In practice, the output was ample for whatever level I wanted – and sometimes it was a rather high level! – with whichever of the six sets of headphones and earphones I used.
Now, just for fun, I do want to qualify that 15mW figure for output into 16-ohm loads. Life is short, so I use only three signals to measure this aspect of performance. They are pure sine waves I digitally generated, peaking at 0dBFS in the 16-bit digital space. Their frequencies are 100 hertz, 1002 hertz and 10,000 hertz. In this test with the AK PEE51, all was fine except for a weirdly wonky 10kHz signal:
I was somewhat tempted to undertake some further tests to analyse the particular type of distortion this waveform represents. But, again, life is short. And whatever it is, it is clearly harmonic, and since this is 10kHz, the lowest possible element in the distortion is at 20kHz, and all the rest is at 30kHz, 40kHz and so on.
I like measurements. I like finding oddities. But, seriously, this particular one means nothing. I found that I could regularise the sine wave by reducing the output to 0.2 volts. That equates to 2.8 milliwatts, which is 4.5dB above headphone sensitivity rating. If you try driving 10kHz into your ears at this level, expect a life-long hole in your auditory perception centred on 10kHz.
So, for all practical purposes the 15mW figure is real-world stuff.
As is, the Astell&Kern PEEE51 portable DAC ought to suit the tastes and needs of just about all mobile high fidelity enthusiasts – assuming that they aren’t adherents of iOS. Given how well other A&K products work with DSD, I’d be surprised if the issues I’ve raised aren’t dealt with in an early firmware upgrade and with proper Window drivers. It’s small and light enough to take just about anywhere, whether for use with your computer or your (Android) phone.
I am puzzled, though, why it doesn’t work with iOS. Perhaps a future firmware upgrade can change that.