@Yuriv Great post! You answered some key questions about the 4G products from Apple and they're mostly encouraging. So that's good news. The output impedance issue is my biggest complaint about the Touch 3G (well ignoring iTunes at least).
As for the square wave response I agree with you and tried to make it clear in the review it's about balancing compromises and there are different opinions as to which way is "best". I've done ABX comparisons of different filter options and can confirm at least some of them do indeed sound different from each other. In practice, the differences between the Clip+ and the Touch 3G probably are not audible, or even if they are, likely don't make enough difference to get very excited about. I'll try and revise my reviews to be a bit more "neutral" on the topic.
But the output impedance is way more likely to cause audible differences. My UE SuperFi's sound different on the iPod vs the Clip+. As I tried to explain on my blog, some might prefer the frequency response variations induced by the iPod's high output impedance (see below) but I don't.
Benchmark, like many headphone/DAC manufactures, took the low impedance approach to their design (they call it "zero ohm") and I agree with that approach as it minimizes the interaction with various different headphones like you see in the graph below. However, as I point out in my blog, some headphones are designed with a higher output impedance in mind. And, with those headphones, one can argue a low impedance source will change their sound from what the manufacture was targeting. I agree there's no well accepted standard these days.
Fortunately, most of the headphones with the really wild impedance swings are higher-end models and are likely designed for a low impedance source. So, in practice, I'm not sure there's much of a problem with low impedance sources with most any headphone except for perhaps a few very esoteric ones. But, as shown below, there are problems with higher impedance sources.
As for the small differences in our output impedance measurements, I can do some more research there. The should be done at a frequency where there are not frequency response issues driving the load--I use 1 Khz. Ideally, they should be done with two different loads--say 15 ohms and 150 ohms--rather than a single load and no load. The dScope has selectable input impedance, and one of them is 150 ohms. So I'll try using that instead of no load (100 K ohms). It just makes the math slightly more complicated
The Class D opinion is hard to call from what I've seen so far. You bring up a good point about the output inductor and I'd have to make more measurements there. But there actually
is a fair amount of ultrasonic hash on the Clip+. It has way more than the Touch 3G. The other ways to get response down close to DC with a single ended low voltage power supply involve likely more costly and power hungry designs. Given the Clip's obviously tiny battery and low cost, Class D (or similar) seems to make more sense.
As for clipping levels, I did the best I could with a 0 dBFS signal and the discrete volume steps available on the devices. As I point out in the review, you often have the choice between "well under clipping" and "well over clipping" when you really want a volume setting in the middle. So I get as close as I can to 1% THD without going over. I may start publishing a gain figure for the maximum volume setting to compare the max volume setting possible between players. Classical music lovers with inefficient headphones might, for example, want more gain available. The downside is this confuses some people into thinking max gain = max loudness when it doesn't. It's like the guitar amplifiers in Spinal Tap going up to "11" instead of "10".
Finally, I haven't tried any formal showdown between the DAC1 Pre and the little Clip+. At 40 times the price ($1600 vs $40) it seems a bit unrealistic in some ways. But I agree it could be rather interesting! The DAC1 is my main source for listening with both speakers and headphones when it's not being recruited for "bench test duty".
Here's the plot showing what the iPod's output impedance does to the frequency response: