Depends on your technique of exposing and the situation, but traditionally yes you would want to be at 850, because you would get the most highlight room above middle grey with the least about of noise.
C100: When to use ISOs below 850
- Thread starter Eric Coughlin
- Start date
Depends on your technique of exposing and the situation, but traditionally yes you would want to be at 850, because you would get the most highlight room above middle grey with the least about of noise.
So if I understand it correctly, setting ISO below 850 doesn't change sensor sensitivity or linear gain, instead of it lowers gamma value toward black press, thus reduces noise level in dark areas and rises chance of overexposure. Setting ISO above 850 on the other side applies linear gain. ISO 640 maybe means standard gamma value 2.2
No, ISO 640 is 0dB Gain. ISO850 is +2.5dB gain.
- set exposure "to be right without overexposure" (on the WF monitor) with canon-log and ISO850.
(* with exactly the same exposure and gamma canon-log set ISO to 640 and ISO320 and look to WF monitor .-) )
- with exactly the same exposure set ISO to 320 a gamma to cine1 (you will get exactly the same "white" without overexposure).
- with exactly the same exposure try ISO320 and ISO400 with WDR gamma (you will get exactly the same "white" without overexposure).
But colors and black differ.
Cine1 has the best colors (cyan, yellow, blue, magenta and red are ok).
WDR has more different cyan (blue), canon-log much more different yellow, cyan.
If you try different gamma and different level of exposure (60pct vs 80pct vs 100pct on WF monitor) you get some different colors (even with the same gamma correction .-) ).
This is my experience.
- set exposure "to be right without overexposure" (on the WF monitor) with canon-log and ISO850.
(* with exactly the same exposure and gamma canon-log set ISO to 640 and ISO320 and look to WF monitor .-) )
- with exactly the same exposure set ISO to 320 a gamma to cine1 (you will get exactly the same "white" without overexposure).
- with exactly the same exposure try ISO320 and ISO400 with WDR gamma (you will get exactly the same "white" without overexposure).
But colors and black differ.
Cine1 has the best colors (cyan, yellow, blue, magenta and red are ok).
WDR has more different cyan (blue), canon-log much more different yellow, cyan.
If you try different gamma and different level of exposure (60pct vs 80pct vs 100pct on WF monitor) you get some different colors (even with the same gamma correction .-) ).
This is my experience.
Last edited:
There is confusion in what the negative gain and gain between 0dB (ISO 640) and 2.5dB (ISO 850) really does.
That means that the ISO setting from 320 to 850 doesn't change signal level, but alters the selected gamma. For Canon Log gamma, the ISO 850 makes probably the most logarithmic gamma.
alex1
Well-known member
Maybe it is just me but at 00:53 the photographer looks really awkward holding the camera, especially for having a vertical grip.
Gain (iso) is applied in analogue signal (as Alan Robert said in Addenum 70 on page 15) before A/D converter:
""This is inevitable where the gain is applied in the analogue signal, since 6dB gain doubles the signal level, but only doubles the noise power rather than noise voltage, thus 0.5dB/dB is the expected slope of the curves".
http://thebrownings.name/WHP034/pdf/WHP034-ADD70_Canon_EOS_C300.pdf
The AD converter change linear analogue signal to nelinear digital signal (with gamma applied), so this A-D conversions is non-linear.
The exposure of the chip is one point of view, it depends on the chip construction. On the output is an analogue amplifier where is gain (ISO) setting, this signal goes to the A/D converter, where (on the output) is gamma correction applied in the digital domein. You have the same signal from chip (the same exposure), then when you set ISO850+canon-log gamma or ISO320+cine1 gamma you will get the same level of white - "just without overexposure" in both cases (canon-log+ISO850 is about 103pct of the modulation on the waveform monitor, cine1 + ISO320 is about 107pct of the modulation on the waveform monitor, but both white level are the same - "just" whithout overexposure for the same signal and correct exposure of the chip. If you try change exposure of the chip, you will get in both case the same amount of overexposure - one in 103pct and the second in 107pct level of modulation on the waveform monitor). The differences in colors and black levels are the second point of view .-) .
If you want to understand to this "canon-log ISO setting", try set exposure of the chip (just without overexposure) and with canon-log change ISO from 850 to 320. With ISO320 change exposure of the chip (a little more amount of light) - you will get overexposure on canon-log gamma at 80pct modulation of the WF monitor, exactly the same amount of overexposure as with ISO850, but in different modulation level (ISO 850 has the same overexposure on 103pct of modulation).
Differ black levels - much more than white level, which is probably practically the same. But after the color correction on canon-log gamma footages you will get only a minor differences. Look at here - https://vimeo.com/67939867 - at part with color correction (one is for log with ISO850 and the second is for log gamma with ISO640 - versus cine1 and normal4 gamma with knee 90pct). Black levels, quality of colors (a blue box under window) and amount of noise in black part after the color correction are points of view.
Last edited:
1/ ...set exposure to "just without overexposure" - it is "the fixed point" for signal level on the chip in this test.
- it means that you set exposure of the CMOS chip to level when each pixel of the chip is just before saturation (change aperture of the lens - get a little more amount of light and you will see a clipping on the WF monitor).
This signal goes to the analogue amplifier where you set ISO (GAIN) and then it goes to AD converter where is the gamma correction applied on the output - it is non-liner A/D conversion.
2/ The modulation is level of signal on the waveform monitor, it is the result after non-linear AD conversion. Black has 0pct level of modulation and white has 100pct of modulation for example. On WF monitor is 107pct superwhite headroom.
You get different level of white modulation with different gammas (it is still exactly the same signal from the chip "just befor clipping", but with different ISO and with different gamma applied on the ADc output, you see this level on the WF monitor).
You get different levels for white peak on WF monitor, but if you try change exposure of the chip, you will get exactly the same clipping of the white peak. THis clipping is exactly the same, but on the different levels on WF monitor
.
---
Look to this picture - it is clipping (bad exposure of the chip) for canon-log gamma with ISO640 (a modulation level is about 95pct):
http://www.videoproduce.cz/images/02b_640ISO_0dB_gain_gamma_log_prepal_odclonenim.jpg
and on this picture is exactly the same signal with ISO set to 320, with exactly the same clipping (ISO change only modulation level, not clipping):
http://www.videoproduce.cz/images/04b_320ISO_minus6dB_gain_gamma_log_prepal_odclonenim.jpg
So for canon-log gamma changing ISO from 320 to 850 only change the level of modulation (an amplifier gain), exposure of the peak white on the chip is the same. But black levels (modulation of dark places in the picture) and noise level are different, they depend on the non-linear gamma curve and on the gain applied. If you set chip exposure to max. white peak level (without clipping) and set GAIN to +2.5dB/ISO to 850 then with canon-log you get max. modulation level on WF monitor without clipping (it is the main rule for canon-log, his non-linear A/D conversion). Change ISO to 320 and set cine1 gamma and you will get 107pct of the modulation level - without white peak clipping .-) . That is all about white peak level without clipping.
--
There are pictures with the same level of signal on the chip (the white peak just befor clipping) with different gammas and ISO after color correction:
http://www.videoproduce.cz/images/TEST00_normal4.tif
http://www.videoproduce.cz/images/TEST01_cine1.tif
http://www.videoproduce.cz/images/TEST02_log640corr.tif
http://www.videoproduce.cz/images/TEST03_log850corr.tif
There you can see this differences in black parts levels and in colors quality and in noise levels (in black parts and after color correction), it is the result of non-linear applied gamma curve. But white peak (the chip exposure) is the exactly the same (a white peak level is just befor clipping on the chip). It is "the fixed point" for signal level (on the chip) in this test. The same scene, the same amount of the light, the same exposure (white peak on the chip), but different ISOs and different gammas.
The main surprise for me was that white peak was without clipping (exactly the same) for both canon-log and cine1 gamma (if you used a different gain level). I hope that there will be significant differences in the dynamic range. The result of test is that cine1 has accurate colors (can be output without correction) but WDR or canon-log has sometimes pleasant image, with better (but not such accurate) colors (I must always used a secondary color correction). Big question is level of noise in dark parts (after color correction with the 8bit signal and without external recording), ISO640 with canon-log may be sometime the best.
- it means that you set exposure of the CMOS chip to level when each pixel of the chip is just before saturation (change aperture of the lens - get a little more amount of light and you will see a clipping on the WF monitor).
This signal goes to the analogue amplifier where you set ISO (GAIN) and then it goes to AD converter where is the gamma correction applied on the output - it is non-liner A/D conversion.
2/ The modulation is level of signal on the waveform monitor, it is the result after non-linear AD conversion. Black has 0pct level of modulation and white has 100pct of modulation for example. On WF monitor is 107pct superwhite headroom.
You get different level of white modulation with different gammas (it is still exactly the same signal from the chip "just befor clipping", but with different ISO and with different gamma applied on the ADc output, you see this level on the WF monitor).
You get different levels for white peak on WF monitor, but if you try change exposure of the chip, you will get exactly the same clipping of the white peak. THis clipping is exactly the same, but on the different levels on WF monitor
.---
Look to this picture - it is clipping (bad exposure of the chip) for canon-log gamma with ISO640 (a modulation level is about 95pct):
http://www.videoproduce.cz/images/02b_640ISO_0dB_gain_gamma_log_prepal_odclonenim.jpg
and on this picture is exactly the same signal with ISO set to 320, with exactly the same clipping (ISO change only modulation level, not clipping):
http://www.videoproduce.cz/images/04b_320ISO_minus6dB_gain_gamma_log_prepal_odclonenim.jpg
So for canon-log gamma changing ISO from 320 to 850 only change the level of modulation (an amplifier gain), exposure of the peak white on the chip is the same. But black levels (modulation of dark places in the picture) and noise level are different, they depend on the non-linear gamma curve and on the gain applied. If you set chip exposure to max. white peak level (without clipping) and set GAIN to +2.5dB/ISO to 850 then with canon-log you get max. modulation level on WF monitor without clipping (it is the main rule for canon-log, his non-linear A/D conversion). Change ISO to 320 and set cine1 gamma and you will get 107pct of the modulation level - without white peak clipping .-) . That is all about white peak level without clipping.
--
There are pictures with the same level of signal on the chip (the white peak just befor clipping) with different gammas and ISO after color correction:
http://www.videoproduce.cz/images/TEST00_normal4.tif
http://www.videoproduce.cz/images/TEST01_cine1.tif
http://www.videoproduce.cz/images/TEST02_log640corr.tif
http://www.videoproduce.cz/images/TEST03_log850corr.tif
There you can see this differences in black parts levels and in colors quality and in noise levels (in black parts and after color correction), it is the result of non-linear applied gamma curve. But white peak (the chip exposure) is the exactly the same (a white peak level is just befor clipping on the chip). It is "the fixed point" for signal level (on the chip) in this test. The same scene, the same amount of the light, the same exposure (white peak on the chip), but different ISOs and different gammas.
The main surprise for me was that white peak was without clipping (exactly the same) for both canon-log and cine1 gamma (if you used a different gain level). I hope that there will be significant differences in the dynamic range. The result of test is that cine1 has accurate colors (can be output without correction) but WDR or canon-log has sometimes pleasant image, with better (but not such accurate) colors (I must always used a secondary color correction). Big question is level of noise in dark parts (after color correction with the 8bit signal and without external recording), ISO640 with canon-log may be sometime the best.
Last edited:
anthonyfurl
New member
Thank you for these. I see what looks like h264 artifacts in these tif images and I just wanted to ask if that was the format you captured in.
So the test shows that in C-Log mode with decreasing ISO below 850, the exposure lattitude (maximum exposure / sensor noise) remains the same, but dynamic range (maximum modulation / sensor noise + quantization noise) nonlinearly decreases with decreasing modulation and constant quantization noise. So for both the highest lattitude and DR, ISO between 640 and 850 is best. On the other side in cine1 gamma mode, the widest lattitude seems to be at ISO 320 and decreases with increasing ISO, DR remains the same. What are the maximum modulation levels in Wide DR with ISO 320 and 640?
Last edited:
WDR gamma has the white peak without clipping with ISO320-400 (the same signal "just without clipping" on the chip).
anthonyfurl - there can be some differences, if you don't record internally to H264, also the result depend on where and how you did a color correction. But canon-log with 8bit never be an ideal thing.
- The exposure lattitude is dynamic range ON THE CHIP (!!). It is the linear contrast ratio (4000:1 or 72dB).
- Modulation is some level of signal on the WF monitor after A/D non-linear conversion, depends on the ISO/GAIN applied or on the selected gamma correction.
White peak clipping can start with bad exposure of the chip or with bad ISO/gain setting (too high level). In both case you will see a white peak clipping on the WF monitor. It is a very important thing for this test to have a one fixed point - it is the chip exposure "just without white peak clipping".
anthonyfurl - there can be some differences, if you don't record internally to H264, also the result depend on where and how you did a color correction. But canon-log with 8bit never be an ideal thing.
- The exposure lattitude is dynamic range ON THE CHIP (!!). It is the linear contrast ratio (4000:1 or 72dB).
- Modulation is some level of signal on the WF monitor after A/D non-linear conversion, depends on the ISO/GAIN applied or on the selected gamma correction.
White peak clipping can start with bad exposure of the chip or with bad ISO/gain setting (too high level). In both case you will see a white peak clipping on the WF monitor. It is a very important thing for this test to have a one fixed point - it is the chip exposure "just without white peak clipping".
Last edited:
I have meant dynamic range of codec input signal. On the sensor output, lattitude = DR of course. So C-Log with ISO 850, Wide DR with ISO 400 and Cine 1 with ISO 320 have the same lattitude, they differ mainly in amount of highlight compression.
ad. quantization: 8bit quantization is not that big problem, bigger problem is AVCHD internal quantization, so external recording to higher bitrate codec would help.
ad. quantization: 8bit quantization is not that big problem, bigger problem is AVCHD internal quantization, so external recording to higher bitrate codec would help.
Yes, this is exactly what I mean (my english is really bad, sorry if I wrote something what is not perfect...). And normal4 gamma with knee on 90pct is also without clipping.
I really don't like an external recorder, so I still try to work with ISO to get the best result (noise is the main problem). AVCHD is sequence of 11 frames, every third image is different and the first image of sequence is also different from every other - it does noise much more visible (is it mosquito noise, yes?). I saw footages from C300, they looks a little better but no miracle again. It's needs some 200Mbps (I-frame only 422) record.
Canon-log is a very pleasant-looking image, but on the faces it sometime makes not a very pleasant result. I like F3 s-log (even on the internall record looks a very good), but this camera ergonomy is not for one man show for fun .-) . It looks that I will need ninja2, but I like a freedom...
I really don't like an external recorder, so I still try to work with ISO to get the best result (noise is the main problem). AVCHD is sequence of 11 frames, every third image is different and the first image of sequence is also different from every other - it does noise much more visible (is it mosquito noise, yes?). I saw footages from C300, they looks a little better but no miracle again. It's needs some 200Mbps (I-frame only 422) record.
Canon-log is a very pleasant-looking image, but on the faces it sometime makes not a very pleasant result. I like F3 s-log (even on the internall record looks a very good), but this camera ergonomy is not for one man show for fun .-) . It looks that I will need ninja2, but I like a freedom...
Last edited:
My english is not perfect too, I would write in czech, but I want also others to understand.
So when using other than C-Log gamma, lowering ISO is better than using ND filters, with C-Log gamma ND filters are better.
Problem with noise in compression codecs is, that the noise is low level signal component consisting of high number of frequencies, so it is the first part of picture that is decimated by codec quantization. AVCHD uses reference frames, so the noise could be partialy reconstructed using data from that reference frames, but only in spatial domain, not in time domain, resulting in the so called "mosquito noise" which is not so apparent on single frames but much more in playback. In MPEG2 or MJPEG codecs blocking appears instead of mosquito noise. But the camera noise could be efficiently masked by adding the right amount of film grain.
So when using other than C-Log gamma, lowering ISO is better than using ND filters, with C-Log gamma ND filters are better.
Problem with noise in compression codecs is, that the noise is low level signal component consisting of high number of frequencies, so it is the first part of picture that is decimated by codec quantization. AVCHD uses reference frames, so the noise could be partialy reconstructed using data from that reference frames, but only in spatial domain, not in time domain, resulting in the so called "mosquito noise" which is not so apparent on single frames but much more in playback. In MPEG2 or MJPEG codecs blocking appears instead of mosquito noise. But the camera noise could be efficiently masked by adding the right amount of film grain.
Thanks, I understand what you mean now. The czech language would be ok .-) .
Yes, you are all right. On other gammas is lowering ISO ok, on the canon-log gamma you must be careful because canon-log has a very special levels. Black parts of image needs some ISO level in this case, I tried ISO850, 640 and 500 and the best choise is between ISO640-800 (depends on noise level). I saw the result after color correction.
Much more interesting is ISO320 with cine1 gamma for some common work. Colors are much more accurate and dynamic range is ok, noise level is better than with canon-log or WDR.
But cine-fans likes canon-log with nice desaturate highlights .-) .
Yes, you are all right. On other gammas is lowering ISO ok, on the canon-log gamma you must be careful because canon-log has a very special levels. Black parts of image needs some ISO level in this case, I tried ISO850, 640 and 500 and the best choise is between ISO640-800 (depends on noise level). I saw the result after color correction.
Much more interesting is ISO320 with cine1 gamma for some common work. Colors are much more accurate and dynamic range is ok, noise level is better than with canon-log or WDR.
But cine-fans likes canon-log with nice desaturate highlights .-) .
Last edited: