mb126 wrote:
Based on magnitude of difference between 80D vs 90D and 80D vs A6300, I don't think we are likely to see anywhere near a 1 stop improvement as OP suggests.
I would agree that the impact on subjective noise would likely not be close to 1 stop better, but that is about how it translates to total sensor performance and how we evaluate the differences.
It must be emphasized that there is no discrepancy between read noise data from cgarcia and bill claff. At ISO 1600 the read noise is measured at 14.2535 DN vs 15.032 DN. That difference is 0.077 Ev and inconsequential.
alundeb wrote:
I would agree that the impact on subjective noise would likely not be close to 1 stop better, but that is about how it translates to total sensor performance and how we evaluate the differences.
It must be emphasized that there is no discrepancy between read noise data from cgarcia and bill claff. At ISO 1600 the read noise is measured at 14.2535 DN vs 15.032 DN. That difference is 0.077 Ev and inconsequential.
Maybe I'm misreading something in OP or Bill Claff's data.
In OPs last post on first page he is claiming 1 stop improvement (e.g. 80D ISO 1600 DR should be equivalent to 90D 3200 DR). I'm saying that considering the difference between the A6300 vs 80D is about 1/3 of a stop throughout the range, it does not stand to reason that this will hold up given that Read noise difference between 80D vs 90D <<< Read noise difference between 80D vs A6300.
I'd love to be proven wrong when Bill updates the PDR with the 90D. But I'm going to remain skeptical until then.
alundeb wrote:
I would agree that the impact on subjective noise would likely not be close to 1 stop better, but that is about how it translates to total sensor performance and how we evaluate the differences.
It must be emphasized that there is no discrepancy between read noise data from cgarcia and bill claff. At ISO 1600 the read noise is measured at 14.2535 DN vs 15.032 DN. That difference is 0.077 Ev and inconsequential.
I think you have the decimal point in the wrong place.
mb126 wrote:
Maybe I'm misreading something in OP or Bill Claff's data.
I was concerned about differences I saw in the graphs too, but
bclaff wroteon the DPP forum:
I performed a Read Noise in DN analysis and the results agree with those by cgarcia. My estimates of DxOMark Landscape Dynamic Range (8MP normalized Engineering Dynamic Range (EDR)) also agree.
mb126 wrote:
Maybe I'm misreading something in OP or Bill Claff's data.
In OPs last post on first page he is claiming 1 stop improvement (e.g. 80D ISO 1600 DR should be equivalent to 90D 3200 DR). I'm saying that considering the difference between the A6300 vs 80D is about 1/3 of a stop throughout the range, it does not stand to reason that this will hold up given that Read noise difference between 80D vs 90D <<< Read noise difference between 80D vs A6300.
I'd love to be proven wrong when Bill updates the PDR with the 90D. But I'm going to remain skeptical until then.
Fully agree. Based on very preliminary data the 90D has increased pixel count without impacting noise/DR performance at pixel level, and has slightly better performance on a normalized basis. Which is nice but not the same as "much better". And for High ISO it is now about as good as the 7D2, which is an improvement over the 80D (but only 1/3 to 1/2 stop, nothing spectacular). Nothing dramatic like a real 1 stop DR advantage for 90D over 80D (or even several stops improvement as was suggested earlier in the thread).
mb126 wrote:
FWIW Bill Claff's data does not appear to support this
The full PDR graph is not yet available for the 90D but presumably it will reflect what we see in the read noise:
1. Base ISO will be similar to 80D
2. High ISO performance will be closer to 7D2 or slightly better
PDR incorporates both shot and read noise into its metric, whereas EDR uses only read noise (and saturation point). This means PDR is a proxy for noise across both shadows and midtones whereas EDR represents only noise in the deep shadows. If EDR has improved by 1EV at High ISO then the 1EV improvement will be observed in the shadows. This means a significant improvement for low-light and/or low-key High ISO images, for example nighttime photos, and less improvement for mid and high key shots, for example daytime photos that raise ISO to achieve fast shutter speeds for action shots.
Note that as ISO increases the impact of read noise on total noise increases because the shadows shift into the lower midtones.
snapsy wrote:
PDR incorporates both shot and read noise into its metric, whereas EDR uses only read noise (and saturation point). This means PDR is a proxy for noise across both shadows and midtones whereas EDR represents only noise in the deep shadows. If EDR has improved by 1EV at High ISO then the 1EV improvement will be observed in the shadows. This means a significant improvement for low-light and/or low-key High ISO images, for example nighttime photos, and less improvement for mid and high key shots, for example daytime photos that raise ISO to achieve fast shutter speeds for action shots.
Note that as ISO increases the impact of read noise on total noise increases because the shadows shift into the lower midtones. ...Show more →
Yes, so it depends very much on the type of photography if one will see an advantage. Personally I don't see much value in better High ISO performance in low light conditions, but it's probably nice if you are doing people/party shots, stage photography etc. with the lens wide open and borderline acceptable shutter speeds because of motion blur. For me the real problem with the 80D sensor is noise in High ISO action images and there doesn't seem much improvement there (probably a little, depending on conditions like lighting etc. etc.).
But at least the good news is that after so many years in last position Canon is finally getting closer to the best-in-class sensors.
technic wrote:
For me the real problem with the 80D sensor is noise in High ISO action images and there doesn't seem much improvement there (probably a little, depending on conditions like lighting etc. etc.).
All of the current APS-C cameras have nearly identical shot noise metrics, so there should be no difference for the High ISO action imagery outside of the shadow areas. That changes for Ultra High ISOs (12,800+), where read noise in the shadows become the lower mid tones, and with the 90D's improvement there it should fully match the performance of current APS-C bodies. Here's DxOMark's 18% SNR values for the D500, A6300, and 80D: https://photos.smugmug.com/photos/i-5hhPwF3/0/77e8d25f/X2/i-5hhPwF3-X2.jpg
Note that the A6300's knee between ISO 6400 and 12800 is the result of raw NR.
snapsy wrote:
All of the current APS-C cameras have nearly identical shot noise metrics, so there should be no difference for the High ISO action imagery outside of the shadow areas. That changes for Ultra High ISOs (12,800+), where read noise in the shadows become the lower mid tones, and with the 90D's improvement there it should fully match the performance of current APS-C bodies. Here's DxOMark's 18% SNR values for the D500, A6300, and 80D: https://photos.smugmug.com/photos/i-5hhPwF3/0/77e8d25f/X2/i-5hhPwF3-X2.jpg
Note that the A6300's knee between ISO 6400 and 12800 is the result of raw NR.
It's unclear to me what you mean by "shot noise metrics".
It's true that 18% SNR is dominated by shot noise; in this case photon noise.
But normalized (Print) values are more a reflection of Full Well Capacity (FWC) than anything else.
Per unit area the cameras you cite are quite close together.
My figures are (per square micron):
Nikon D500 2607e-
Sony ILCE-6300 2548e-
Canon EOS 80D 2538e-
bclaff_too wrote:
It's unclear to me what you mean by "shot noise metrics".
It's true that 18% SNR is dominated by shot noise; in this case photon noise.
But normalized (Print) values are more a reflection of Full Well Capacity (FWC) than anything else.
Per unit area the cameras you cite are quite close together.
My figures are (per square micron):
Nikon D500 2607e-
Sony ILCE-6300 2548e-
Canon EOS 80D 2538e-
Hi Bill, by shot noise metrics I meant photon shot noise, which is represented by the DxO graph I presented - 18% SNR, which is determined by the combination of FWC and QE.
Quick comment — from someone who has not followed the entire thread*...
Whether the new 32.5 sensor is the same, slightly better, or a full stop better... one important take-away is that the oft-repeated concern that the new higher MP sensors will perform worse! is again contradicted by the existence of a new sensor that seems to at least equal the performance of the older sensor with fewer photo sites. That is an important big picture point.
Dan
* I've been in the Sierra back country for more than a week and just returned very late last night.
snapsy wrote:
All of the current APS-C cameras have nearly identical shot noise metrics, so there should be no difference for the High ISO action imagery outside of the shadow areas. That changes for Ultra High ISOs (12,800+), where read noise in the shadows become the lower mid tones, and with the 90D's improvement there it should fully match the performance of current APS-C bodies. Here's DxOMark's 18% SNR values for the D500, A6300, and 80D: https://photos.smugmug.com/photos/i-5hhPwF3/0/77e8d25f/X2/i-5hhPwF3-X2.jpg
Note that the A6300's knee between ISO 6400 and 12800 is the result of raw NR.
That's the theory but I notice that High ISO images from the D500 in the range of about ISO 800-6400 seem to hold up MUCH better than those from my theoretically identical 80D
technic wrote:
That's the theory but I notice that High ISO images from the D500 in the range of about ISO 800-6400 seem to hold up MUCH better than those from my theoretically identical 80D
That's because 18% SNR is not a good indicator of low light performance since it's photon noise dominated.
Low light performance is mostly about where photon noise just starts to overcome read noise which is why read noise measurements are more indicative.
technic wrote:
That's the theory but I notice that High ISO images from the D500 in the range of about ISO 800-6400 seem to hold up MUCH better than those from my theoretically identical 80D
That's because as you increase ISO read noise becomes an increasing influence on total noise and the D500 has a 1EV advantage over the 80D in High ISO read noise:
With the 90D's High ISO read noise improved by 1EV the D500's advantage will disappear. Btw, to get a sense for where the cross-over point of shot and read noise occurs at each ISO and for what tonal aspects of the image play with this link I created for you into Dpreview's IQ widget:
gdanmitchell wrote:
Quick comment — from someone who has not followed the entire thread*...
Whether the new 32.5 sensor is the same, slightly better, or a full stop better... one important take-away is that the oft-repeated concern that the new higher MP sensors will perform worse! is again contradicted by the existence of a new sensor that seems to at least equal the performance of the older sensor with fewer photo sites. That is an important big picture point.
Dan
* I've been in the Sierra back country for more than a week and just returned very late last night.
At equivalent sensor technology a higher-resolution sensor will perform worse in terms of High ISO read noise and DR, even when normalized to the same resolution, as a function of its higher aggregate read noise. But this has generally masked by the fact that each step up in resolution has been accompanied by sensor tech improvements in per-pixel read noise.
snapsy wrote:
At equivalent sensor technology a higher-resolution sensor will perform worse in terms of High ISO read noise and DR, even when normalized to the same resolution, as a function of its higher aggregate read noise. But this has generally masked by the fact that each step up in resolution has been accompanied by sensor tech improvements in per-pixel read noise.
That is (almost) precisely my point — though it requires some qualifications.
First, part of the problem is that folks presume that only one factor is in play — the fact that sensors using smaller photo sites, all else being equal, produce more noise than sensors using larger photosites.
Second, they forget that various technical advances have improved the ability of sensors and their individual photo sites to capture a wider range of data values to lower the effect of noise.
Third, the overlook the very significant improvements in how the captured data are processed.
When you look at the big picture, what matters is the quality of the file you get out of the camera in the end. And as photo site density has increased radically over a period of decades the quality of the image files has gotten better and better. The early reports about what comes from cameras using the 32.5MP sensor is yet one more bit of evidence.
snapsy wrote:
That's because as you increase ISO read noise becomes an increasing influence on total noise ...
I hope you won't find my comment out of line because you might think it's simply semantics.
As the ISO setting is increased read noise does not become a "increasing" factor, if anything the effect is slightly reduced.
The key factor is that less signal is collected, Full Well Capacity (FWC) is not fully utilized.
NormQ wrote:
Egrets represent little challenge for combinations much further down the totem pole that the 5D4 and 500LII. The only reason for the performance you claim is user or equipment error. I suggest you have your gear looked at by Canon. I use the same combo for much faster and smaller species and maintain a vey high keeper rate. Handheld. Out of curiosity can I ask if your concern with the 90D having spot AF is because you use this AF selection for BIF?
Yes, they should represent little challenge as compared to a small passerine, though if you ask those who have been in the field with Canon for years, it is a fact of life. The AF is THE reason Art gave up Canon in favor of Nikon and Sony. There is nothing "wrong" with my gear (am a CPS member), have many years experience with Canon equipment, have calibrated all of the gear, etc. FWIW I was tracking using center AF point with surround on a Egret flying across our field of view with a sun angle to our back, at 9am against open water/trees. I've experienced this with GBH's, Sandhill Cranes, Eagles and other large birds for no real explicable reason other than focus simply doesn't lock on and track properly. It doesn't happen all of the time, but frequently enough to be frustrating. Five minutes later, I captured a beautiful series of a barn swallow under similar lighting - smaller subject, more erratic flight, etc. This kind of failure never happened with my 1d series but was more common with the 7d's, 5d's, etc. in spite of proper technique, settings etc. I've wondered for years whether it had something to do with separate processors for AF, voltage differences driving the lens, or some other factor?
My hope was the 90d would have an improved AF system which would surpass the 7dmkii/5dmkiv. I need more than 20 mp full frame otherwise, I'd be shooting with a 1dx2. With respect to the spot AF, I use it a fair amount for perched birds for improved accuracy.
