First of all, that's a really nice report. It confirms what we already know about the Clip+ from others' measurements. I wrote a few notes as I was reading your blog entry. I'll use Pascal's excuse: this post is long because I didn't have time to make it shorter.
1. I'm skeptical about the Clip+ having a class D output stage. Devices that have them typically have a rising source impedance at higher frequencies, due to the inductors in the filter. When I measured the voltage drop of a 20 kHz signal across a 16-ohm load, I pretty much got the same result as I did at 1 kHz. Plus your graphs don't show too much ultrasonic gabage.
2. Older iPods have a flat response to 20 kHz. The iPhone 4 and the iPad have a slightly softer top end, even with no load--about 0.5 dB down at 20 k. Several CD players and DACs have a similar response. It seems that they're using a different reconstruction filter this time around. More on this later.
EDIT 2/25/2011: This result had always bothered me, and I got it a long time ago, when the iPhone 4 came out and the iPad still had iOS 3. Since then, gsmarena and Anandtech have measured the iPhone 4 and the iPod Touch 4 and both of them got ruler-flat responses. So I borrowed the iPhone 4 again and redid the tests. I'm happy to confirm their results. The iPad too, measures now flat. It might have been a human error back then--overwriting the wrong files, etc.
3. It's a little strange that the iPod Touch 3G has Rs = 7 ohms. Stereophile measured 5.5 ohms for a very old model. I'm also getting 5.5 ohms for my 2005 iPod Video. Additionally, I calculate 0.7, 1.4, and 0.9 ohms source resistance for the original Clip, the iPad, and the iPhone 4, respectively. But keep in mind that I'm also getting 0.55-0.6 ohms for the Clip+ compared to your 1-ohm result. It's in the same ballpark.
4. There isn't a definitive standard on what the output impedance should be for headphone amplifiers. It's not like the situation for loudspeaker amplifiers, where manufacturers try to get as close as possible to an ideal voltage source with vanishing source impedance (the damping factor spec wars). Back in 1996 the IEC called for 120 ohms output resistance (IEC 61938, formerly IEC268- 15). It's hard to tell which headphones were designed to operate with such an amp. If the headphone's impedance is uniform with frequency, then they will sound roughly the same with a low-impedance amp. Otherwise the frequency balance will be different.
The IEC has since changed its spec. I don't remember what the new guidelines are. As you can imagine, the current situation is rather chaotic. The Wikipedia entry for headphone amplifier says that most headphone amplifiers have Rs = 20-50 ohms. IEC 61938-type units are practically a current source for very-low-impedance IEMs. Early iPods have Rs around 5-6 ohms. It gets worse: except for the many "Designed for iPod" portable headphones, it's not easy to tell what source impedance a manufacturer had in mind when they designed and tested their headphones.
So, is output impedance still a good figure of merit for headphone amplifiers? Personally, I like my amps to be close to ideal voltage sources. (You can always stick resistors in series, just in case). Some examples: Benchmark DAC1, Grace Designs m902, β22, and now the Clip/Clip+ and new iDevices. Their Rs is still not as low as loudspeaker amplifiers, maybe by design. Speakers, after all, usually stay connected to the amp. Headphones are occassionally inserted and removed when music is still playing, so if there's a temporary short, the extra series resistance can limit the current draw.
5. The maximum volume possible into a 16-ohm load is now higher for the iPhone 4 and the iPad, thanks to the lower source resistance. For example, I'm getting 0.63V for 1% distortion on the iPhone 4:
But if you just reduce the volume by a hair, 0.57V, THD for a 1 kHz tone is now 0.0035%:
I suspect the same for your iPod Touch--you've reached the onset of clipping. But low-impedance earbuds and IEMs don't need to be driven to these levels. Most users won't even get close to clipping before it gets unbearably loud. With higher-impedance 'phones, the iPod can play a clean signal louder, so it all works out.
6. Julian Dunn's J-Test signal is a little different from a pure tone at Fs/4. Its job is to attempt to introduce data-induced jitter. The .wav file is floating around somewhere online. I wonder if using that will make a difference in your tests. Even with these devices the jitter isn't high enough to make a difference when listening to music normally.
7. Apple didn't get lazy with the square wave performance of the new iPods. In fact the old iPods play square waves just like the Clip and the Clip+. Apple really opted for different compromises. Here's what the impulse response looks like for these devices:
Impulse response for Sandisk Sansa Clip+ (above left) and Apple iPad (above right) and Apple iPod Video (below left). iPad 1 kHz square wave (below right)
It's the difference between a linear-phase and a minimum-phase filter. The ringing looks bad for minimum phase, but some mastering engineers find this behavior to be less objectionable than the pre-ringing in the linear-phase filter's response. There's a recent discussion on Gearslutz where they all say that it's more damaging to music. For me, it's very hard to tell which unit is which when listening to music normally. The differences might just be of academic interest.
Compare my pics with some of Stereophile's measurements for these two units:
The Meridian 808.2/808i.2 can actually select between the two reconstruction filters: http://www.stereophile.com/content/meridian-8082808i2-signature-reference-cd-playerpreamplifier-measurements
The Ayre QB9 has minimum-phase. It has a setting that reduces the impule response's decay time, but there's no free lunch: it has to roll off the top end even more:
http://www.stereophile.com/content/ayre-acoustics-qb-9-usb-dac-measurements
I can do that too with the iPad:
The latest news is that Apple will be switching back to Wolfson. The time-domain behavior of future iPods, iPhones, and iPads may change again. I didn't have a problem when they were using Cirrus.
You might not like how your iPod plays square waves and count it as one of its disadvantages. In practice, I really don't notice much of a difference when listening to music. (But maybe that's the catch--like 99% of the folks out there, I'm listening to the music instead of the gear.)
BTW, I had a chance to compare my Clip and Clip+ to a Benchmark DAC1. I thought they sounded very similar, even with multi-armature IEMs. Maybe you can publish the results of a level-matched blind test between your DAC1 and the Clip+. If you don't crank up the level, it might be very hard to spot the differences; they have a few similarities: low-enough source impedance, impulse response, square wave (DAC1 might be flatter), etc.
Keep up the good work!
Edited by yuriv - 2/24/11 at 4:00pm
