p.15 #2 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
Steve Spencer wrote:
For some applications (e.g., video) that is absolutely true, but for other applications personally I find all the other things you mentioned much more important. I can turn the ISO up a third of a stop almost all the time in my shooting and the slight decrease in transmission matters little.
I purchase fast lenses firstly to be able to shoot in dimmer light...so for me the t-stop matters a lot if it is much higher than the theoretical f-stop. I wish manufactures would publish this value.
I shoot quite often at ISO 12,800 so increasing the ISO is not the desired answer.
Very cool samples! To me, it renders similarly to the 24 GM.
Mar 05, 2020 at 10:09 AM
Steve Spencer Online • • • • • • • Upload & Sell: On
p.15 #4 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
chez wrote:
I purchase fast lenses firstly to be able to shoot in dimmer light...so for me the t-stop matters a lot if it is much higher than the theoretical f-stop. I wish manufactures would publish this value.
I shoot quite often at ISO 12,800 so increasing the ISO is not the desired answer.
I don't think anyone has shown, however, that the difference between T-stop and f-stop is much different between fast lenses and slower lenses. Even the one example in which T-stop was measured by the Chinese site the difference between the difference in T-stop and f-stop between the faster lens (f/1.4) and the slower lens (f/1.8) was considerably less than a third of a stop (about a 1/6th of a stop as measured). So by any reasonable standard that is a very small difference and in fact my argument has been it is so small that I have not seen any one show that we can measure T-stop precisely enough to even measure this difference reliably.
Said another way, if the the difference between T-stop and f-stop is pretty much the same for all lenses, then f-stop, which is a physical measure and can be measured precisely, tells us basically the same thing and T-stop. Only if the difference between T-stop and f-stop varies quite a bit (and T-stop can be measure accurately and fairly precisely) does T-stop provide any additional information.
p.15 #5 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
Steve Spencer wrote:
I don't think anyone has shown, however, that the difference between T-stop and f-stop is much different between fast lenses and slower lenses. Even the one example in which T-stop was measured by the Chinese site the difference between the difference in T-stop and f-stop between the faster lens (f/1.4) and the slower lens (f/1.8) was considerably less than a third of a stop (about a 1/6th of a stop as measured). So by any reasonable standard that is a very small difference and in fact my argument has been it is so small that I have not seen any one show that we can measure T-stop precisely enough to even measure this difference reliably.
Said another way, if the the difference between T-stop and f-stop is pretty much the same for all lenses, then f-stop, which is a physical measure and can be measured precisely, tells us basically the same thing and T-stop. Only if the difference between T-stop and f-stop varies quite a bit (and T-stop can be measure accurately and fairly precisely) does T-stop provide any additional information....Show more →
DXO publishes their t-stop values and they do vary from lens to lens, some more drastic than others, some are bang on their f-stop rating. If one lens is rated at t-stop 1.8 and another much larger heavier lens is rated at t-stop 1.6...then that matters to me and I need to take that into my decision as to whether the more expensive and heavier 1.4 lens is warranted.
On the DXO site there are f2.8 lenses that have a t-stop of 3.2...quite a difference IMHO.
For example the Tamron 24-75 f2.8 lens has a t-stop of 3.3...so you really are not getting a 2.8 lens as far as low light shooting goes.
Mar 05, 2020 at 11:37 AM
Steve Spencer Online • • • • • • • Upload & Sell: On
p.15 #6 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
chez wrote:
DXO publishes their t-stop values and they do vary from lens to lens, some more drastic than others, some are bang on their f-stop rating. If one lens is rated at t-stop 1.8 and another much larger heavier lens is rated at t-stop 1.6...then that matters to me and I need to take that into my decision as to whether the more expensive and heavier 1.4 lens is warranted.
On the DXO site there are f2.8 lenses that have a t-stop of 3.2...quite a difference IMHO.
For example the Tamron 24-75 f2.8 lens has a t-stop of 3.3...so you really are not getting a 2.8 lens as far as low light shooting goes....Show more →
But what we don't know is how precise and how reliable those DXO numbers are. Do they have the precision in their measurements to tell the difference between 3.3 and 3.1? If they tested 30 lenses what would be the confidence interval around their T-stop measurement? Would it be .05 or would it be .4? We don't know any of that which makes it quite ambiguous what these numbers mean. Of course, a T of 2.8 vs. 3.3 is just over a third of a stop, but if their measurement is precise to even a sixth of a stop (which I highly doubt it is even that precise) then the lens might only be giving up a sixth of a stop versus the nominal f-stop which would not be bad at all and a lens measuring T of 2.8 could also be giving up a sixth of stop it is just the margin of error happened in different directions which is not only possible but will happen with one test of one lens. So unless they can test with more than a sixth of a stop precision and they start testing multiple lenses it is quite possible that a lens they rate with a T-stop of 2.8 is no different from a lens they rate as having a T-stop of 3.2.
p.15 #7 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
I think the difference between T-stop and f-stop depends on coating and the number of elements? I don't see a reason why smaller aperture lenses lose smaller percentage of light assuming similar coating technology.
p.15 #8 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
Steve Spencer wrote:
But what we don't know is how precise and how reliable those DXO numbers are. Do they have the precision in their measurements to tell the difference between 3.3 and 3.1? If they tested 30 lenses what would be the confidence interval around their T-stop measurement? Would it be .05 or would it be .4? We don't know any of that which makes it quite ambiguous what these numbers mean. Of course, a T of 2.8 vs. 3.3 is just over a third of a stop, but if their measurement is precise to even a sixth of a stop (which I highly doubt it is even that precise) then the lens might only be giving up a sixth of a stop versus the nominal f-stop which would not be bad at all and a lens measuring T of 2.8 could also be giving up a sixth of stop it is just the margin of error happened in different directions which is not only possible but will happen with one test of one lens. So unless they can test with more than a sixth of a stop precision and they start testing multiple lenses it is quite possible that a lens they rate with a T-stop of 2.8 is no different from a lens they rate as having a T-stop of 3.2....Show more →
Why are you doubting the precision of DXO measurements? Seems to me the consumer grade lenses have more variance away from their f-stop which would be what I expect. Also seems to me zooms vary more than primes...again it's what I would expect. What are you basing your negativity around. From what I see in the data, it appears to line up with what I'd expect.
I guess you are a glass half empty guy.
Mar 05, 2020 at 01:30 PM
Steve Spencer Online • • • • • • • Upload & Sell: On
p.15 #9 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
chez wrote:
Why are you doubting the precision of DXO measurements? Seems to me the consumer grade lenses have more variance away from their f-stop which would be what I expect. Also seems to me zooms vary more than primes...again it's what I would expect. What are you basing your negativity around. From what I see in the data, it appears to line up with what I'd expect.
I guess you are a glass half empty guy.
I doubt the precision of DXO measurement for two reason 1) they don't report to more than one decimal place. In science you report to one decimal place beyond where you have precision. So by simply reporting to one decimal place if they are following typical scientific reporting procedure they are telling us they do not have precision to one decimal place. They might have precision to even .2 or even .15, but it could also be precision to .9 or .95 (it is likely more precise than that). Still if they had precision below .1 then they should be reporting to the second decimal place (e.g. 2.95). Suggesting precision is at best .1 and that means in comparing any two lenses at least a .2 swing could happen just by lack of precision (with the error going in opposite ways). 2) They only test one lens rather than multiple lenses. We therefore know nothing about how testing procedure and copy variation would affect T-stop. My guess is copy variation alone may not allow much precision and multiple copies would be needed to get any sort of precision and a report on the range of T-stops one could expect across multiple copies of a lens. When you test only one copy, however, you know nothing about copy variation.
For these two reason we need to take DXO measurement as like a very imprecise measure of T-stop.
Finally let me add I am also appalled by the numerous errors that Bill Claff has found in DXOs reporting of their scores. See the large number of errors for example in his audit of DXOs sports scores:
p.15 #10 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
Steve Spencer wrote:
I doubt the precision of DXO measurement for two reason 1) they don't report to more than one decimal place. In science you report to one decimal place beyond where you have precision. So by simply reporting to one decimal place if they are following typical scientific reporting procedure they are telling us they do not have precision to one decimal place. They might have precision to even .2 or even .15, but it could also be precision to .9 or .95 (it is likely more precise than that). Still if they had precision below .1 then they should be reporting to the second decimal place (e.g. 2.95). Suggesting precision is at best .1 and that means in comparing any two lenses at least a .2 swing could happen just by lack of precision (with the error going in opposite ways). 2) They only test one lens rather than multiple lenses. We therefore no nothing about how testing procedure and copy variation would affect T-stop. My guess is copy variation alone may not allow much precision and multiple copies would be needed to get any sort of precision and a report on the range of T-stops one could expect across multiple copies of a lens. When you test only one copy, however, you know nothing about copy variation.
For these two reason we need to take DXO measurement as like a very imprecise measure of T-stop....Show more →
The only place that I know of that tests multiple copies is lensrentals...and at a very specific situation. So by those standards...every other test / review is not valid.
As far as reporting to multiple decimal points...the audience is not some bithead engineer...the audience is photographers that just might get more confused with more data...just my opinion.
p.15 #12 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
DXOmark is filled with discrepancies when it comes to testing t-stop which makes me not trust their precision either.
Here is one example:
Sigma 85/1.4 Art tested on Canon 5D Mark IV: DXOmark: 1.7 T-stop (link)
Sigma 85/1.4 Art tested on Canon 5DS R: DXOmark: 1.8 T-stop. (link)
Their "lens" t-stop measurement tends to differ depending on the testing body.
p.15 #13 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
moyang_mm wrote:
Can we compare the relative T-stops of two lenses by measuring the exposure of RAW files shooting in a controlled lighting condition, fixing the ISO and the shutter speed?
Yes, it's pretty straightforward. The best way is to shoot the comparison images through a diffuse material (I use a #2447 acrylic sheet) and then compare the raw ADU values. I mostly recently did this in my Sigma ART vs Tamron SP thread. The set setup looks like this:
Steve Spencer Online • • • • • • • Upload & Sell: On
p.15 #14 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
chez wrote:
The only place that I know of that tests multiple copies is lensrentals...and at a very specific situation. So by those standards...every other test / review is not valid.
As far as reporting to multiple decimal points...the audience is not some bithead engineer...the audience is photographers that just might get more confused with more data...just my opinion.
Well, everyone who knows me does not see me as a glass half empty type of guys. Quite the opposite. So, until you get to know me you might want to refrain from making such quick judgements. By the way, I am not an engineer either and certainly not a bithead one whatever that might mean. And I do agree that without multiple copies we can't know about copy variation and lens rentals tests are very useful in giving us this information that nobody else provides.
I am on the record here many times suggesting that I like Fred's tests, those done by the time and Philipreeve.com, and lens rentals and I don't find much use for almost any other tests. I can live with Fred and the team at Philipreeve.com not testing multiple copies because I trust them to get a good copy and they almost always comment on how many tries it took to get a good copy or if they couldn't get one what they saw as the flaws in the copy they tested. They can't give a full account of copy variation in their tests, but it is not as if they just ignore copy variation in their tests.
I do worry about anyone who ignores copy variation in their tests and certainly testing like that at DXO that also shows extensive sloppiness in the reports of their testing and when that sloppiness is pointed out to them they don't even fix their errors on their site. You are free to take those numbers for DXO T-stop tests as gospel if you like, but I am far from convinced they are meaningful and for the reasons I have outlined I strongly suspect they are not precise.
Mar 05, 2020 at 02:02 PM
Steve Spencer Online • • • • • • • Upload & Sell: On
p.15 #15 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
snapsy wrote:
Yes, it's pretty straightforward. The best way is to shoot the comparison images through a diffuse material (I use a #2447 acrylic sheet) and then compare the raw ADU values. I mostly recently did this in my Sigma ART vs Tamron SP thread. The set setup looks like this:
p.15 #16 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
Steve Spencer wrote:
Snapsy what is your test/re-test reliability? Any thoughts about copy variation?
Precision is good. I usually test at least twice. There are some pitfalls to accuracy though, specifically lens model-specific gain adjustments to the raw data, which can include lens vignetting correction (which Sony applies to raw data, even when it isn't expected) and pixel vignetting correction (particularly apparent at very large apertures). I test the dead center of the lens to help avoid some of these issues as much as possible. I've never compared the transmission of two copies of the same lens - since transmission is mostly a function of the base optical design and coatings I wouldn't expect there to be much variation but I really don't have any hard info or knowledge either way on the matter.
p.15 #17 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
As much as I like testing (being a scientist and all) to get back to the original topic of this thread. I just received from my US east coast dealer that they have my FE 20mm f1.8 G ready to pick up. Hopefully I can make it in Friday morning.
p.15 #18 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
There are times when I wonder whether the forums here stray from their objective, which for me are to inform without going overboard. Since I have a perfectly good 35mm all I really want to know is whether or not this lens obviates the need to buy a 16/17-28/35. In my experience as a landscape photographer, mainly recently with Leica, I am usually at one end or other of this range but rarely in between. Whether the extremes are 17-20 or 28-35 makes little difference. That said, I suspect I will be better with a zoom than a prime, particularly as I feel a 20mm fast prime is pointless for most landscape work and, generally, less useful than a fast 35/45.
p.15 #19 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
Fred Miranda wrote:
DXOmark is filled with discrepancies when it comes to testing t-stop which makes me not trust their precision either.
Here is one example:
Sigma 85/1.4 Art tested on Canon 5D Mark IV: DXOmark: 1.7 T-stop (link)
Sigma 85/1.4 Art tested on Canon 5DS R: DXOmark: 1.8 T-stop. (link)
Their "lens" t-stop measurement tends to differ depending on the testing body.
With rounding the numbers which I assume the 1.7 and 1.8 are rounded, the actual results could be only different by 0.01 t-stop.
IE: one reading of 1.750 which would be rounded to 1.8 and the other reading at 1.749 which rounds to 1.7.
I'm not worried about a .1 difference in t-stops, but a difference of .5 gets my attention.
p.15 #20 · Back On Sale: Sony FE 20mm f/1.8 G lens for only $798 ($100 off)
waterden wrote:
There are times when I wonder whether the forums here stray from their objective, which for me are to inform without going overboard. Since I have a perfectly good 35mm all I really want to know is whether or not this lens obviates the need to buy a 16/17-28/35. In my experience as a landscape photographer, mainly recently with Leica, I am usually at one end or other of this range but rarely in between. Whether the extremes are 17-20 or 28-35 makes little difference. That said, I suspect I will be better with a zoom than a prime, particularly as I feel a 20mm fast prime is pointless for most landscape work and, generally, less useful than a fast 35/45....Show more →
I like 20mm. I first started to like it when shooting weddings with Nikon and when I bought the Nikon 20mm/1.8G. With Sony I used the Loxia, which was sold a few days ago; a fantastic lens but I found it cumbersome to use in its double role, some landscapes for fun and using it to get images of the location etc. et weddings. Awkward to mount and demount, MF, I longed for AF and faster aperture. The new Sony looks to be a great lens. I like what I saw so far, so preordered it. Mine ships tomorrow.
