Alan Roberts F3 assesment is now online: http://thebrownings.name/WHP034/pdf/WHP034-ADD68_Sony_PMW-F3.pdf
In the report Alan observes the aliasing that I have seen from the camera, in particular the high frequency moire, so no surprise there. But he also measures the noise at -48.5db. Now I don’t have the ability to measure noise as Alan does and I normally respect his results, but this noise figure does not make sense, nor does his comment that the camera has similar sensitivity to most 3 chip cameras. To my eyes, the F3 is more sensitive than any 3 chip camera I’ve used and it’s a lot less noisy. The implication of the test is that the F3 is noisier than the PMW-350. Well that’s not what my eyes tell me. Take a look at the noise graph Alan has prepared. The hump in the noise figure curves at 0db also appears to be dismissed as insignificant, yet it means a greater than 4db difference between what the curve implies the noise figure should be and the measured noise figure. It really doesn’t seem to fit and is very strange. Video amplifiers and processing are normally pretty linear with gain giving a consistent increase/decrease in noise that follows the gain curve. If you read off the noise figures from the graph, the F3 appears to have less noise at +6db gain (-49.5db) than at 0db (-48.5db). So id we are to believe Alan’s test then we should be using +6db gain or -3db gain (-53.5db) but not 0db, sorry but that just does not add up and to dismiss the 0db noise bump as “not significant” is something I don’t really understand as too me it is significant. Either there is something very strange going on with the F3 at 0db, or there is something up with the test. I suspect the later, perhaps the individual camera had some odd settings, as my F3 is quieter (visually) at 0db than +6db. I would need to check it out on a scope back at home to verify this.
There are also assumptions made about the pixel size and sensor pixel count that are quite wrong. Alan suggests the sensor to be a 12 Mega Pixel sensor, this suggestion is based on Alan’s opinion that the F3 has similar sensitivity to a 2/3″ 3 chip camera, so therefore the pixel size must be similar and the bigger sensor means that it must have 12 MP, yet Sony have published that it is 3.3 mega Pixels (same sensor as FS100). 3.3MP equates to roughly 2422 x 1362 pixels, for a bayer sensor this is a little under the optimum for 1920 x1080 (IMHO) and may explain the aliasing as Sony are probably trying to squeeze every last bit of resolution out of the sensor.
Alans assessment of his zone plate results also concludes that the R, G and B resolutions are the same and that the sensor resolution must be much higher than 2200 x 1240. Well I would not call 2422 x 1362 “much” higher and if this is a bayer sensor (neither admitted or denied by Sony) then the G resolution should be higher than the R and B. So could this be a case of conventional conclusions about an unconventional sensor, or have Sony managed to completely wrong foot Mr Roberts?
An interesting finding was that detail at zero, frequency at +99 and aperture at +20 gives the least aliasing. This is quite different from my own findings and will need further exploration.
The F3 assessment is also missing the usual customary round up from Alan where he suggests whether the camera is suitable for HD broadcast or not. I’m really glad I got my F3 before reading the report as I have seen with my own eyes the beautiful clean images the F3 produces. I strongly recommend anyone considering the F3, but put off buy this report to take a look at the pictures for themselves before making any decisions.