Friday 12 September 2014

MEASUREMENTS: Microsoft Surface Pro 3 Audio.

Over the last couple years I have been using my ASUS Taichi ultrabook as my daily work/business traveller laptop. I've certainly appreciated the ultrabook form factor and speed! Although these little devices aren't cheap, there's no way for me to go back to anything larger these days in terms of a full function portable work machine...

"Unfortunately", I felt the need to upgrade the daily use machine as I was starting to run into headroom issues with 4GB RAM and needed something with substantially more battery life for the longer trips. Enter the newest member of the Microsoft Surface family with the Intel Haswell core - Surface Pro 3. Sure, I could wait another 6 months and get the Broadwell maybe in early 2015 but like anything in technology, you can't wait it out forever when there are true advances being made.

I got the i5, 8GB RAM, 256GB SSD model. I suspect this will be the most popular model of the i3/i5/i7 versions as a reasonable compromise between features and price. I was actually looking at trying the i7 model but I couldn't find any locally! I believe the i5 was released about a month and a half ago and the i7 just came out so demand was higher. In any event, the i5 should be more than adequate and I hope there will be some benefits to a less speedy model in terms of heat production, noise, and maybe improved battery life.

With blue Type Cover 3.
I'm not going to bother with a full review. You can find plenty on-line like AnandTech, or IBT. Overall I do agree with the reviews. It's certainly not the perfect tablet - too heavy and large still for casual browsing and media consumption. And it's not the perfect laptop/ultrabook for the price. It is however made very well with a nice magnesium case which is solid with a sturdy kickstand. It's certainly very easy to travel with, and the newest Type Cover 3 functions decently as a keyboard/trackpad. I actually was interested in the first Surface Pro when it came out in early 2013 but there were too many issues at the time with heat, battery life, and keyboard/mouse issues that I went with the ASUS Taichi then. With the improvements since, I thought I'd give this iteration a try.

Just a few quick comments:
1. The 3x2 screen is different. First time I've bought a non-widescreen (~16:9) computer/monitor in the last decade. For a small device like this, it works well. It's a little smaller than standard letter paper size. I love the screen contrast and 2160x1440 resolution - almost as high as my 27" main monitor! Note that some old software might not scale appropriately leading to ridiculously small fonts which might make it tough to use the touch screen accurately.

2. The digitizer pen feels good and works well. I started using OneNote more and can appreciate the capabilities. Seems to work well so far taking notes in meetings. Pressure sensitivity was good and allows some casual sketching (not that I'm much of an artist!). I find myself sometimes accidentally pressing the buttons on the pen which I need to get used to.

3. Mixed feelings about the kickstand and "floppy" keyboard. I'm actually typing fine on the Type Cover 3 keyboard - speed about the same as my previous ASUS ultrabook. The fact that there's no spacing between the keys increases the chance of inadvertent typos. What's different is the center of gravity and feel of the base. Normally the keyboard is the base of the laptop rather than the lower edge of the slate + the kickstand being the supports with the center of gravity pushed distally. It's not a big deal on a table since the kickstand is very stable; but feels initially strange and one has to be careful when placed on the lap - it can be easy to tip over with leg movement. This "lapability" issue is known and at least the kickstand can be adjusted further to improve balance but this also extends the front-to-back length.

4. Please Microsoft, consider two USB ports in future versions; I'd hate to carry around a hub. Sometimes it's nice to plug in my Logitech Unify receiver for an external mouse/keyboard and still have another USB for something like the Dragonfly or AudioEngine D3.

5. I'm glad they're using the DisplayPort for external monitor output. Easier and cheaper for analogue VGA adapters. Support for 4K, etc...

6. One beneficial result of the kickstand form factor is that it's cooler on the lap since there is no contact with the area around the CPU and heatsink. The fan is still audible with moderate to heavy computing but at least the heat is dissipated away from where there's physical contact.

On to the headphone sound...

For such a small device, the built-in speakers are fine and I do believe sounds better than the older ASUS Taichi. The first thing I noticed when looking at the sound playback settings is that the hardware is identified as "Realtek High Definition Audio" and the maximum samplerate possible is 24/48! I tried the latest Realtek HD Audio drivers here (R2.75) as well and still 24/48. It looks like previous versions could do 24/96. Though my preference would be at least 24/96 these days, I don't believe this is a problem; surprising nonetheless since almost all devices should be able to handle 24/96 by now. Perhaps the hardware is capable but there's something with the OS settings I haven't adjusted yet (using the current stock Windows 8.1 installation that came with the machine).

A look at the oscilloscope graphs:
1kHz Sine 0dBFS at 16/44.

1kHz Square 0dBFS at 16/44.
Oscilloscope indicates that at 100% volume with 0dBFS signal, there's no clipping. Nice. Also, channel balance is excellent! Maximum output voltage is ~1.6V. Not capable of playing music as loud as the Dragonfly v1.2 or AudioEngine D3 so make sure to go for higher sensitivity, less power-hungry headphones.

The 16/44 impulse response looks like this:

As expected, standard symmetrical linear phase digital filter. Less than usual pre and post-ringing suggesting earlier roll-off in the frequency response.

Time for the RightMark 6 results. Here are the 16/44 results with a few other devices for comparison. I used the usual measurement chain:
Microsoft Surface 3 headphone out --> shielded phono-to-RCA --> E-MU 0404USB --> shielded USB --> Windows 7 laptop


The Surface Pro 3 is in the first column, followed by both the Dragonfly and AudioEngine D3. I then threw in my two compact portable devices the Nexus 5 phone and Nexus 7 tablet. Followed by the Transporter and Oppo BDP-105 as higher end devices. Apart from the stereo crosstalk being a bit "weak", we're seeing that 16/44 is not a problem for any of the devices these days. The defining factor likely would be the quality of the headphone outputs (eg. power, impedance characteristics) rather than accuracy of the DAC. As usual, a few graphs (click images to enlarge):
Frequency Response - Surface 3 slightly earlier high-frequency rolling off than others.
Noise Level
Stereo Crosstalk - unusually high for Surface Pro 3.
Since 48kHz is the highest samplerate supported, it's worth measuring since this would be the natural downsampling level for all the 96 and 192 kHz music. I don't do many measurements at 24/48 but I'll throw in the results from the Squeezebox Touch and Receiver from awhile back for comparison as well as 24/96 results from some laptops I had measured last year to show their 24-bit performance:

Noise level and dynamic range suggests that the Surface Pro 3's headphone output has about 17.25-bit dynamic range. With 24-bit data, the best devices I have like the Transporter will give me a measured result around -113dB (almost 19-bits) which is also likely a limitation of the measurements using the E-MU 0404USB. Distortion characteristics of the Surface 3 are comparable to the Touch, but again, we're seeing quite high stereo crosstalk of -66dB which is poor for digital devices.

Despite the 48 kHz limitation, it's nice to see that the Surface 3 is capable of better dynamic range than my old Acer Aspire 5552 laptop and ASUS Taichi ultrabook. From those tests last year, I was quite impressed by the Apple MacBook Pros and I see that the Surface 3 measures about the same as the old MacBook Pro from 2009 except for the poorer stereo crosstalk.

Again, a few graphs for completeness (only the 24/48 results):
Frequency Response
Noise Level
Stereo Crosstalk
Let's finish off the objective stuff with the jitter spectra - as usual, the Dunn J-Test; 16-bit and 24-bit versions:

While the 16-bit J-Test results aren't the cleanest nor does it look ideal, it'd be hard to make a case for jitter audibility. The 24-bit results looks very nice. Overall minimal "skirting" around the primary frequency.

I'm not going to spend much time on the subjective descriptions here... The fact is, when I'm traveling, I just don't see why I'd need the full high-fidelity experience. I'm only likely ever going to be listening with this machine for any length of time on plane flights where the background noise is high and I'm likely distracted anyhow doing work. Suffice it to say, in an atypical and wholly unrealistic situation for me sitting in the quiet of my home with my Sennheiser HD800 headphones plugged into the Surface Pro 3, the sound is excellent. It doesn't provide as much power as the recently measured Dragonfly v1.2 or AudioEngine D3 so don't expect it to play very loud. It does however convey the resolution I'm used to like Donald Fagen's immaculately produced "I.G.Y." from The Nightfly. I had a listen with the AKG Q701 as well which is one of my more inefficient headphones. Again, the detail was excellent but perhaps a bit bass shy and at maximum volume, it's just capable of slightly louder than my usual listening level. This volume level would be fine with my high sensitivity IEM headphones I usually bring on my travels though.

Although stereo crosstalk is high compared to other digital devices, subjectively I do not believe this is a problem. Remember that vinyl has stereo separation typically between 20-40dB only and it can still sound excellent. It's one of those things where technical specification strongly exceeds real-world necessity (I would still like to see "perfect" execution of course from an engineering perspective).

I'm impressed by how much technology can be packed into a small package these days! As I write this last bit, I have been using the machine for close to a month now and it's performing very well. The small form factor is convenient and shaves a few ounces off my work briefcase compared to the previous ultrabook. As need for increasing CPU speed appears to be plateauing and monitor resolution really doesn't need to get any better for smaller devices, it is going to be the form factor, weight, and usability that matters. While not ideal at this point, the Surface Pro 3 is clearly taking a step closer.

As for the sonic measurements. Stereo crosstalk is high for a digital device and I'm surprised by the 48kHz sampling rate limit. It does handle 24-bit data with decent dynamic range rivaling the old Apple MacBook Pro I have. Compared to previous Windows laptops, despite the size, it does quite well keeping the noise level low.

To end off, a few thoughts over the last couple weeks:

1. Sometimes the darn 'Connected Standby' sleep feature on the Surface Pro 3 makes the machine not wake up! Look here for a thorough article on this issue. It happened to me twice in the 1st week of ownership. I turned off the Hyper-V feature and thereby disabling this "feature". I've since turned Hyper-V back on and have not noticed a problem in the last 10 days... Maybe there was some software update that fixed this.

2. Here's a serious question/suggestion for high-end DAC manufacturers. I know HDMI costs money to license yearly. How about using the DisplayPort as a standard for audio data transmission - a royalty-free VESA specification? I've always wanted a high-end multichannel DAC/streamer (attention paid to resolution, lower jitter, etc.) - something like the Transporter with robust server system that can handle *all* my music files - both stereo and multichannel (maybe even DSD stereo/multichannel but that's lower on my priority list). What I'm currently doing with sending multichannel via the computer's HDMI to the Onkyo NR-TX1009 receiver works and will get me multichannel sound. But the Onkyo DAC isn't the best, occasionally I will run into disconnections and I would love to see unification on the software side where something with the power of Logitech Media Server can be used to serve both stereo and multichannel! The DisplayPort has the capability for 24/192 multichannel no problem and presumably cheaper for manufacturers without worrying about the licensing costs.

3. There's a guy who claims to be able to ABX high-resolution audio files as well as ABX using the Bozza track from the 24-bit test. He claims to be successfully doing it with an HP Zbook 14 laptop and Etymotic ER-4P. Given what I have seen with measuring the laptop headphone outputs, I'd be very curious what that HP Zbook 14's output quality looks like to confirm that it is capable of >16-bit resolution. Furthermore, I have the Etymotic ER-4B (as discussed in the Dragonfly measurements) sporting higher 100-ohm impedance - it's good but I can't say I've ever considered it that good in conveying sonic detail. Not saying that the guy cannot tell the difference but I would be a bit cautious given the equipment used and posting a "successful" ABX log is easy; would have been better if he responded to the survey and got 3/3 in a blind test... Needless to say, in my 24-bit test, I specifically asked folks not to just use the computer motherboard/laptop headphone jack simply because it more than likely outputs compromised sound compared to something superior like even the inexpensive Dragonflies of this world (if not a fully external powered DAC) if one is serious about trying to listen to high resolution audio.

That's all for now.

Have a great weekend and week ahead everyone!


  1. I really appreciate your reviews.

    Would you consider using 30 and 300 ohm dummy loads for measurements of devices that would be directly used with headphones in the future?

  2. Hi xnoreq,
    Good point.

    I will consider it in the next while... Just put in an order for the iPhone 6 for my wife :-)
    [I'm an Android guy!]

  3. Some headphone outputs may perform much worse when loaded with a low impedance (33 Ohm) load.
    The load the EMU (in High Z input mode) provides may skew the results in a very positive way as it doesn't load (1 MOhm) the headphone output at all.
    Roll-off in the lows, amount of distortion(s) and channel separation are most likely affected when measuring with a dummy load. Also squarewaves may may look different.
    The results are still valid when you connect a (high quality) headphone amplifier to the headphone out of the mentioned devices though.

  4. The "positive ABX guy" killed it for him by showing the positive 16/24 bit log. Until then there was maybe a slight possibility that he wasn't just gathering visits to his website - but after that - nope, he is very probably just full of it. "Trained" or not - claiming to hear differences >90 dB below music at normal levels on a system with at best similar dynamic range is just too much. Nice publicity stunt but nothing more.

    1. Thanks for the note cer. Indeed, even if he had used amazingly accurate equipment, one just has to take that 100% (or was that 0%?) log with huge suspicion.

      I find it amazing that that post has some subjectivists on places like Audio Asylum agreeing that it represents some kind of smoking gun with regards to audibility of 24-bit audio... Amazing to see comments to that regard even from folks I (at least used to) consider as somewhat insightful.

      October promises to be an interesting month for 24-bit audio; assuming Pono delivers on time. Grab your popcorn and prepare for the media blitz if this happens. :-)

  5. Enter the newest member of the Microsoft Surface family with the Intel Haswell core - Surface Pro 3. Sure, I could wait another 6 months and get ...

  6. Thanks for the detailed analysis. Have you also measured the output impedance of SP3 headphone out?

  7. A good amount of simply because so many critics can regard as this really is no true, After i fully grasp develop which often getting this know-how associated with almost any theme submitting place of work in fact hoards an individual rendezvous. Kinds notion offers myself enigma a good option there are several that are looking for this genus associated with drive. royal mail phone recycling