While testing the output consistency of various flash units for Robert to add to this thread, I decided to also test the consistency of several units using various automatic flash metering systems. I learned a few other things, too.
The first thing I realized was that none of the battery powered flash units were completely recycled when the ready light came on. In consecutive pops, waiting five seconds after the ready light resulted in as much as 4/10 of a stop more power in manual mode, with the greatest differences being at the higher power settings. This was true of every unit I tested, one each of four different brands. The moral of this story is that if you're picky about consistent flash output, you need to be patient and wait for it to recycle completely, especially if you're using full or half power.
My Alienbees B1600 did not exhibit this behavior. Firing it as soon as the ready light came on gave me just as much light as when I waited five more seconds.
The other conclusions I can draw are as follows:
1) Traditional auto mode is more consistent than E-TTL (or E-TTL II). This is not a general indictment of E-TTL, since there are many differences between these two metering systems and the difference in output consistency is not so great that it should preclude us from using E-TTL when conditions dictate.
2) The Canon 580EX II had more consistent output than the other two E-TTL compatible brands I tested. While the performance of the aftermarket brands was not so poor as to consider them junk, it is perhaps a reminder that we get what we pay for. As I conducted the tests, it was apparent that the pre-flash from the 580EX II was significantly more consistent than the others, and it stands to reason that an inconsistent pre-flash will yield inconsistent exposure with E-TTL.
I would caution everyone not to draw more conclusions from the data than I have already mentioned. While my knowledge of statistics is not sufficient to calculate a reasonable margin of error, the data was taken from relatively small samples (10 pops each unit), and measured with a meter that has an accuracy of +/- 0.1 EV and rounds each reading to 0.1 stops.
Testing protocol info:
The E-TTL readings were taken using a Canon 300D, and the E-TTL II readings were taken using a Canon 20D. My Promaster 7500DX is an older version, not compatible with E-TTL II. The camera was on a tripod, aimed at a grey streaked background about 6 feet away. The readings were taken with a Sekonic L-358 meter in cordless flash mode, taped to a stand just in front of the backdrop.
E-TTL and E-TTL II measurements were taken by using Flash Exposure Lock to separate the pre-flash from the exposure flash.