The A2DP support & sound quality & power consumption of the Microsoft and Widcomm BT stack compared; a lot of new A2DP i

Now that I’ve made the stereo Bluetooth headphones (A2DP) support work with my Dell Axim x51v with the original Microsoft BT stack (see the tutorial here), I’ve thoroughly tested it and compared the sound quality and the power consumption of the two stacks.

First, the power consumption test, along with the CPU usage (all measured on the same x51v two times - with a hard reset in between. I've long been promising something similar to the AximSite folks in this thread):

BT stack:MicrosoftWidcomm
BT disabled112 mA112 mA
BT enabled, A2DP off138 mA143-149 mA
A2DP on140-142 mA, some temp 150 peaks now and thenThe same
WMP; Värttinä test playback213-226 mA; in general, about 220 in avg220-246 mA; about 233 in avg with the FlowControl hack; without the hack, about 5-6 mA less with smaller peaks
device.exe9.3-11.3% CPU7.3-9.7% CPU

As can clearly be seen, while device.exe, which is in charge of encoding the A2DP stream, consumes slightly less CPU time with the Widcomm BT stack, the latter still consumes about 13 mA’s more than the Microsoft stack. This, however, isn’t that big a problem if you also take into account the really superior sound quality.

What should you use to check out the sound quality yourself?

I really recommend for example the demo MP3 of Värttinä’s Oi Dai (linked from here); in there, you’ll clearly hear the quality difference between the A2DP mode of the MS and the Widcomm BT stack. Just make sure you try to listen to the generic distortion in the middle band when the girls start to sing.

BTW, use the first about five seconds to fine-tune the Subbands value with the Widcomm BT stack (see my earlier, related tweaking article here) – it’s mostly in the increased noise of strong lows (like at the beginning of this song) that the drastically decreased Subbands value has a really audible effect. With other kinds of music, the difference is very hard to hear.

Tweaking

I’ve played a bit with the MS BT stack parameters under HKEY_LOCAL_MACHINE\ SOFTWARE\ Microsoft\Bluetooth\ A2DP\Settings I’ve listed in my previous A2DP fine-tuning article. I’ve paid special attention to the following DWORD’s:

MinSupportedBitPool: it does have an impact on the sound transmission: while you can’t fine-tune the quality, a sufficiently high value (for example, 64; with 48, it still works) will make sure no connection can be made between the headphones and the PDA. That is, it’s useless to try to “tweak” this value.

UseJointStereo: the default 0 is “false”. When set to 1, it results in a mono transmission. With 2 (I’ve also tested with 3, with the same result), it seems it indeed uses joint stereo mode. It, however, hasn’t resulted in any kind of sound quality improvement. Conclusion: forget it too.

I’ve, in addition to trying to tweak BitPool, also tried to raise MaxSupportedBitPool , MinBitPool and, of course, BitPool – without success. I couldn’t achieve any sound quality improvement.

All in all, it seems the Microsoft BT stack is definitely worse, A2DP quality-wise, than that of Widcomm / Broadcomm and nothing (no tweaking) can be done to fix this.

Note that you can very easily spot the sound quality problem with the MS BT stack by

1. it clearly has weaker highs than the Widcomm BT stack. This isn't so big a problem - after all, the (slight) lack of highs isn't so big a problem as constant, content-dependent aliasing at around 2-3 kHz
2. the sound sounds exactly like the aliasing in analog -> digital digitalization lacking a properly designed anti-aliasing filter, which would be needed to comply with the Nyquist–Shannon sampling theorem.

Summary

There is not really competition between the A2DP quality of the two Bluetooth stacks. If you happen to have a WM5-upgraded Dell Axim x50(v) or x51(v), don’t hesitate: install the Widcomm BT stack on it if you want the best available sound quality. You’ll be presented definitely better sound quality, on the expense of slightly (but not much!) decreased battery life. And, of course, you get a lot of additional Widcomm goodies like Bluetooth PAN and the like.

Similarly, if you need to choose between, say, a Widcomm BT stack-based device and a MS-based one (and there's no known Widcomm hack for the latter), go for the earlier, if, otherwise, the two devices are the same.

This also means my sound quality problems with the HTC Wizard weren’t (also see my mourning here) because of the HTC Wizard hardware or the bad A2DP hack (see this for a more thorough elaboration on the hack) but because of the A2DP algorithm used in the MS BT stack. The official A2DP implementation of the really powerful Dell Axim x51v AKU2 isn’t at all better, quality-wise, than the A2DP hack for the Wizard.

I’ll try, of course, to push the Microsoft folks very hard to find the cause for the audibly lower sound quality.

UPDATE (11/05/2006): discussion: HowardForums

UPDATE (11/07/2006): PPCT frontpage - please do read this thread, I've posted a LOT of interesting stuff in there!!!