Disco Robo
Well-known member
After reading casual observations and anecdotal evidence I chose to run my own latitude test for the T2i. This test is mostly scientific and flawed only by the quality of tools used. The conclusions I derive from the data may be different than your own however the data is unbiased and as accurate as possible given the tools used.
I started by building a light box with 4 individual cells. The 4 cells were then gelled with nothing, ND.9, ND1.8, and ND2.7 or if you prefer nd8, nd64, and nd512 and for the old school video guys 1/8, 1/64, and 1/512. The cells were then backlit with 4 t5 28w 5500k 92cri fluorescent lights. The 4 cells only represent 2 feet of the overall 4 foot length of the tube as the tubes tend to be hotter at one end than the other. Lightmeter readings were used to confirm that each cell was indeed 3 stops removed from the one on either side.
The camera was set at 1080 24P, ISO 100, f3.5 with an initial shutter speed of 1/2000, and the Daylight preset. I used a Tamron 17-50 2.8 lens set at 48mm to limit the effects of barrel distortion. The angle of the camera was set so that the variance from cell to cell (pre-gel) was 90 IRE +-1 IRE. To accomplish this I simply rolled the focus all the way out until each box was a blob of light so that fine detail was not a factor in setting level. BTW this is how you're supposed to set white balance IYDK. I used the widely agreed upon best practices profile of Neutral -4 -4 -2 0.
The focal subject is the lighting itself, this objectifies the test and does not subject it to reflective readings, because a non-backlit , 11 step chart is limited to about 6 stops.
The tools I used were Lee ND.9 gel, a Sekonic lightmeter, FCP, and a Matrox MXO2. The footage was recorded to card and FCP simultaneously via the composite output recorded at 10 bit uncompressed NTSC. Both videos were converted to Pro-Res 4444 through mpeg streamclip, the composite recording was also upscaled to 1080p and CC'd to bring the black to 0. All IRE numbers are from the card, the composite follow was there so I didn't have to use flash cards, the paper kind, to get shutter data. The top portion of the composite feed was cropped and the card footage is moved up but that was the only manipulation of the footage. The final output for vimeo is h.264. ALL waveform readings relate to the 4444 card footage within FCP using it's WF only.
The data will be presented in the following format 2000-70/26/3/0 meaning shutter speed 1/2000 cell 1 70IRE, cell 2 26IRE, cell 3 3IRE, cell 4 0IRE.
Bored yet? On to the data.
SHTR-0C1/C2/C3/C4
2000-070/26/03/0
1600-075/30/05/0
1250-080/35/07/0
1000-086/40/08/02
0800-093/46/10/03 Yikes thats 10 stops or is it 9
0640-100/49/12/3.5 Just peaks
0500-100/55/16/04
0400-100/61/18/4.5
0320-100/66/21/5.5
0250-100/70/26/06
0200-100/76/30/6.5
0160-100/80/35/07
0125-100/86/40/08
0100-100/92/46/09
0080-100/100/50/11.5 Almost the same as SS 640 3 stops removed....Ahhh science
0060-100/100/55/15
0050-100/100/60/19
0040-100/100/65/22
0030-100/100/69/25 Whew
A real WFM with gain would have been nice to more accurately judge the IRE numbers but my methodology remained consistent as mostly proved by the data.
The results....Well it looks like the T2i is capable of physically resolving about 10 stops. How usable the information above 93 IRE or below 3 IRE is up to you. I figure it's good for about 8 usable stops where there is enough data to create useful picture information.
Note: I've been a video engineer and director for 13 years and I'm on vacation, smart asses with dumb questions will be summarily lit up. Beyond that I'm glad to give back to the community.
Take Care.
I started by building a light box with 4 individual cells. The 4 cells were then gelled with nothing, ND.9, ND1.8, and ND2.7 or if you prefer nd8, nd64, and nd512 and for the old school video guys 1/8, 1/64, and 1/512. The cells were then backlit with 4 t5 28w 5500k 92cri fluorescent lights. The 4 cells only represent 2 feet of the overall 4 foot length of the tube as the tubes tend to be hotter at one end than the other. Lightmeter readings were used to confirm that each cell was indeed 3 stops removed from the one on either side.
The camera was set at 1080 24P, ISO 100, f3.5 with an initial shutter speed of 1/2000, and the Daylight preset. I used a Tamron 17-50 2.8 lens set at 48mm to limit the effects of barrel distortion. The angle of the camera was set so that the variance from cell to cell (pre-gel) was 90 IRE +-1 IRE. To accomplish this I simply rolled the focus all the way out until each box was a blob of light so that fine detail was not a factor in setting level. BTW this is how you're supposed to set white balance IYDK. I used the widely agreed upon best practices profile of Neutral -4 -4 -2 0.
The focal subject is the lighting itself, this objectifies the test and does not subject it to reflective readings, because a non-backlit , 11 step chart is limited to about 6 stops.
The tools I used were Lee ND.9 gel, a Sekonic lightmeter, FCP, and a Matrox MXO2. The footage was recorded to card and FCP simultaneously via the composite output recorded at 10 bit uncompressed NTSC. Both videos were converted to Pro-Res 4444 through mpeg streamclip, the composite recording was also upscaled to 1080p and CC'd to bring the black to 0. All IRE numbers are from the card, the composite follow was there so I didn't have to use flash cards, the paper kind, to get shutter data. The top portion of the composite feed was cropped and the card footage is moved up but that was the only manipulation of the footage. The final output for vimeo is h.264. ALL waveform readings relate to the 4444 card footage within FCP using it's WF only.
The data will be presented in the following format 2000-70/26/3/0 meaning shutter speed 1/2000 cell 1 70IRE, cell 2 26IRE, cell 3 3IRE, cell 4 0IRE.
Bored yet? On to the data.
SHTR-0C1/C2/C3/C4
2000-070/26/03/0
1600-075/30/05/0
1250-080/35/07/0
1000-086/40/08/02
0800-093/46/10/03 Yikes thats 10 stops or is it 9
0640-100/49/12/3.5 Just peaks
0500-100/55/16/04
0400-100/61/18/4.5
0320-100/66/21/5.5
0250-100/70/26/06
0200-100/76/30/6.5
0160-100/80/35/07
0125-100/86/40/08
0100-100/92/46/09
0080-100/100/50/11.5 Almost the same as SS 640 3 stops removed....Ahhh science
0060-100/100/55/15
0050-100/100/60/19
0040-100/100/65/22
0030-100/100/69/25 Whew
A real WFM with gain would have been nice to more accurately judge the IRE numbers but my methodology remained consistent as mostly proved by the data.
The results....Well it looks like the T2i is capable of physically resolving about 10 stops. How usable the information above 93 IRE or below 3 IRE is up to you. I figure it's good for about 8 usable stops where there is enough data to create useful picture information.
Note: I've been a video engineer and director for 13 years and I'm on vacation, smart asses with dumb questions will be summarily lit up. Beyond that I'm glad to give back to the community.
Take Care.
