deepDEEPpurple wrote:
It might appear that a faster SS could be provided and while SS and read out speed are related these are not the only considerations in an image pipeline. There are FW constraints, pixel reset, ADC/noise reduction/etc. Moving this amount of data at a faster pace would require even more engineering.
Sure, all possible, plus additional technical considerations that may not be apparent to us. Some notes about some of the points you raised...
The tight timings of even slower shutter speeds implies full hardware automation of the readout process, likely via the imaging ASIC. Many of the sensor application guides I've read require ASIC/FPGA timing for many of the readout modes. Based on this FW timing constraints shouldn't apply.
Based on my previous research, row reset times are very fast, effectively instantaneous relative to readout times.
The data throughput is constant irrespective of shutter speed because shutter speed is implemented by the time interval between the reset and readout of a row (ie, integration time) - the reset and read times themselves are fixed, thus the throughput out of the chip is constant.
tsdevine wrote:
I know there are times I can’t shoot wide open on a sunny day with my 50-150 for certain shots. I usually forget an ND filter, so there are times where 1/8000 is limiting. Not the end of the world, but 1/16000 would definitely reduce the situations I need a ND.
Of course it may be there may be enough headroom in the RAW file, but that’s hard to know absolutely while shooting in the field.
Being bothered by someone being bothered seems less defendable than someone being bothered by an actual situation a person runs into while shooting IMHO.
But at the end of the day, people will be bothered…....Show more →
The NiSi PRO Nano IRND Filter (95mm, 3-Stop) is $119.
Without the faster shutter speed and with these filter prices, I might just resort to f/2.8. It would take a number of instances of missing the shot at f/2.8 for me to buy the ND. So, A1ii shooters .. a plus for you.
Now, spending the money for a polarizing filter might get more use.
Tim:
Again, a heavy round of applause for your 50-150mm images. Do you find yourself using a polarizing filter with a degree of frequency? This might be the case of getting a bit of ND and better sky’s.. win .. win.
bwcolor wrote: tsdevine wroteI know there are times I can’t shoot wide open on a sunny day with my 50-150 for certain shots. I usually forget an ND filter, so there are times where 1/8000 is limiting. Not the end of the world, but 1/16000 would definitely reduce the situations I need a ND.
Of course it may be there may be enough headroom in the RAW file, but that’s hard to know absolutely while shooting in the field.
Being bothered by someone being bothered seems less defendable than someone being bothered by an actual situation a person runs into while shooting IMHO.
But at the end of the day, people will be bothered….The NiSi PRO Nano IRND Filter (95mm, 3-Stop) is $119.
Without the faster shutter speed and with these filter prices, I might just resort to f/2.8. It would take a number of instances of missing the shot at f/2.8 for me to buy the ND. So, A1ii shooters .. a plus for you.
Now, spending the money for a polarizing filter might get more use.
Tim:
Again, a heavy round of applause for your 50-150mm images. Do you find yourself using a polarizing filter with a degree of frequency? This might be the case of getting a bit of ND and better sky’s.. win .. win....Show more →
Thanks, yeah I do have a 95mm 3 stop ND filter, it's just for whatever reason I don't always bring it. It's in one of my other bags, etc. And when I forget, I just stop down a bit.
I do have a way to rig up the 50-150 to use my 82mm Maven magnetic filters. That gives me a full set of CPL and dark CPLs, but the adapter I rigged up only comes along if I think I'm doing waterfall shooting (which doesn't happen much with this lens). I also use a CPL when trying to increase the reflection on reflection shots, but there again that's not the use case where I'm trying to shoot wide open.
I just have to be more disciplined about checking that I have the filter before I walk out the door. Or I need to take the adapter and magnetic CPL set with me, and pick up a 3 stop ND in 82mm if I don't want polarization.
snapsy wrote:
Coming to this thread late (which based on Fred's post above is probably a good thing).
I was interested in a possible technical reason behind the 1/8,000 limitation. You would think any limitation would be due to sensor readout speed so I did some quick calcs:
Legend:
1 second = 1,000 milliseconds (ms)
1 second = 1,000,000 microseconds (us)
A7rVI sensor metrics (max ES shutter is 1/8,000):
Full-sensor resolution = 9,504 x 6,336
Full-sensor readout speed is 19.6ms = 19,600us
Readout speed per row is 19,600 / 6,336 = 3.09us/row
A7V sensor metrics (max ES shutter is 1/16,000):
Full-sensor resolution = 7008 × 4672
Full-sensor readout speed is 15.1ms (1/66) = 15,100us
Readout speed per row is 15,100 / 4,672 = 3.23us/row
Based on the above, the A7V sensor is slightly slower than the A7VI sensor yet supports a 1/16,000 shutter speed.
If you compare the per-row readout speeds to the 1/16,000 and 1/32,000 exposure times it would appear there is plenty of margin in the readout speed to keep up with the faster shutter speeds, all else being equal....Show more →
Readout speed and max shutter speed is not necessarily related.
Fuji X-T5 readout is 1/20s but electronic max shutter speed is 1/180,000s.
Consider a row of pixels on a sensor. On a rolling shutter camera, each row of pixels is triggered to start and then stop, one row at a time.
To support 1/16000s, each pixel in that row needs to be able to expose itself for 1/16000s. this is the bottleneck to increasing shutter speed.
Each pixel receives a reset followed by a readout instruction. "Reset" clears existing charge and starts integrating, "readout" dumps existing charge to memory and stops integrating
If the physical silicon of the pixel cannot clear its charge fast enough during the reset, or if the transfer gate cannot open/close cleanly within that microsecond window, the pixel cannot physically achieve a 1/16,000s exposure.
The pixel size on the A7RVI is smaller. the analog components clearing the pixel's charge and the transfer gate are miniaturized compared to the A7RV. On top of this readout speed was increased despite having more pixels total, and additional circuitry is probably used to readout each pixel twice to boost dynamic range in mechanical shutter mode.
Its not that surprising to me that something had to give to achieve this.
A9iii can do 1/80,000s but the pixels are way bigger and the well size is smaller. To reset the pixel less charge needs to be removed since the well size is smaller (no iso 100), and the far bigger pixel size means you can double or triple down on analog components related to clearing charge during a reset, and components to readout the charge.
aCuria wrote: snapsy wroteComing to this thread late (which based on Fred's post above is probably a good thing).
I was interested in a possible technical reason behind the 1/8,000 limitation. You would think any limitation would be due to sensor readout speed so I did some quick calcs:
Legend:
1 second = 1,000 milliseconds (ms)
1 second = 1,000,000 microseconds (us)
A7rVI sensor metrics (max ES shutter is 1/8,000):
Full-sensor resolution = 9,504 x 6,336
Full-sensor readout speed is 19.6ms = 19,600us
Readout speed per row is 19,600 / 6,336 = 3.09us/row
A7V sensor metrics (max ES shutter is 1/16,000):
Full-sensor resolution = 7008 × 4672
Full-sensor readout speed is 15.1ms (1/66) = 15,100us
Readout speed per row is 15,100 / 4,672 = 3.23us/row
Based on the above, the A7V sensor is slightly slower than the A7VI sensor yet supports a 1/16,000 shutter speed.
If you compare the per-row readout speeds to the 1/16,000 and 1/32,000 exposure times it would appear there is plenty of margin in the readout speed to keep up with the faster shutter speeds, all else being equal.Readout speed and max shutter speed is not necessarily related.
Fuji X-T5 readout is 1/20s but electronic max shutter speed is 1/180,000s.
Consider a row of pixels on a sensor. On a rolling shutter camera, each row of pixels is triggered to start and then stop, one row at a time.
To support 1/16000s, each pixel in that row needs to be able to expose itself for 1/16000s. this is the bottleneck to increasing shutter speed.
Each pixel receives a reset followed by a readout instruction. "Reset" clears existing charge and starts integrating, "readout" dumps existing charge to memory and stops integrating
If the physical silicon of the pixel cannot clear its charge fast enough during the reset, or if the transfer gate cannot open/close cleanly within that microsecond window, the pixel cannot physically achieve a 1/16,000s exposure.
The pixel size on the A7RVI is smaller. the analog components clearing the pixel's charge and the transfer gate are miniaturized compared to the A7RV. On top of this readout speed was increased despite having more pixels total, and additional circuitry is probably used to readout each pixel twice to boost dynamic range in mechanical shutter mode.
Its not that surprising to me that something had to give to achieve this.
A9iii can do 1/80,000s but the pixels are way bigger and the well size is smaller. To reset the pixel less charge needs to be removed since the well size is smaller (no iso 100), and the far bigger pixel size means you can double or triple down on analog components related to clearing charge during a reset, and components to readout the charge....Show more →
The X-T5 full-sensor readout speed is 1/27, which calculates as:
Full sensor readout speed 1/27 = 37.037ms = 37,037us
X-T5 resolution is 7728 × 5152
37,037 / 5,152 = Readout speed 7.18887us/row
Shutter speed of 1/180,000 = 5.55us integration time
Thus even 1/27 (37ms) can support a shutter speed of 1/180,000, all else being equal (ie, no other technical limitations).
Everything you wrote about other technical limitations is certainly possibly but seems highly unlikely to be the cause of the 1/8000 restriction considering how much timing margin there is for a 1/16,000 shutter speed (62.5us).
Btw, here's my writeup on how rolling shutters work:
snapsy wrote: aCuria wrotesnapsy wroteComing to this thread late (which based on Fred's post above is probably a good thing).
I was interested in a possible technical reason behind the 1/8,000 limitation. You would think any limitation would be due to sensor readout speed so I did some quick calcs:
Legend:
1 second = 1,000 milliseconds (ms)
1 second = 1,000,000 microseconds (us)
A7rVI sensor metrics (max ES shutter is 1/8,000):
Full-sensor resolution = 9,504 x 6,336
Full-sensor readout speed is 19.6ms = 19,600us
Readout speed per row is 19,600 / 6,336 = 3.09us/row
A7V sensor metrics (max ES shutter is 1/16,000):
Full-sensor resolution = 7008 × 4672
Full-sensor readout speed is 15.1ms (1/66) = 15,100us
Readout speed per row is 15,100 / 4,672 = 3.23us/row
Based on the above, the A7V sensor is slightly slower than the A7VI sensor yet supports a 1/16,000 shutter speed.
If you compare the per-row readout speeds to the 1/16,000 and 1/32,000 exposure times it would appear there is plenty of margin in the readout speed to keep up with the faster shutter speeds, all else being equal.Readout speed and max shutter speed is not necessarily related.
Fuji X-T5 readout is 1/20s but electronic max shutter speed is 1/180,000s.
Consider a row of pixels on a sensor. On a rolling shutter camera, each row of pixels is triggered to start and then stop, one row at a time.
To support 1/16000s, each pixel in that row needs to be able to expose itself for 1/16000s. this is the bottleneck to increasing shutter speed.
Each pixel receives a reset followed by a readout instruction. "Reset" clears existing charge and starts integrating, "readout" dumps existing charge to memory and stops integrating
If the physical silicon of the pixel cannot clear its charge fast enough during the reset, or if the transfer gate cannot open/close cleanly within that microsecond window, the pixel cannot physically achieve a 1/16,000s exposure.
The pixel size on the A7RVI is smaller. the analog components clearing the pixel's charge and the transfer gate are miniaturized compared to the A7RV. On top of this readout speed was increased despite having more pixels total, and additional circuitry is probably used to readout each pixel twice to boost dynamic range in mechanical shutter mode.
Its not that surprising to me that something had to give to achieve this.
A9iii can do 1/80,000s but the pixels are way bigger and the well size is smaller. To reset the pixel less charge needs to be removed since the well size is smaller (no iso 100), and the far bigger pixel size means you can double or triple down on analog components related to clearing charge during a reset, and components to readout the charge.The X-T5 full-sensor readout speed is 1/27, which calculates as:
Full sensor readout speed 1/27 = 37.037ms = 37,037us
X-T5 resolution is 7728 × 5152
37,037 / 5,152 = Readout speed 7.18887us/row
Shutter speed of 1/180,000 = 5.55us integration time
Thus even 1/27 (37ms) can support a shutter speed of 1/180,000, all else being equal (ie, no other technical limitations).
Everything you wrote about other technical limitations is certainly possibly but seems highly unlikely to be the cause of the 1/8000 restriction considering how much timing margin there is for a 1/16,000 shutter speed (62.5us).
Btw, here's my writeup on how rolling shutters work:
aCuria wrote: snapsy wroteaCuria wrotesnapsy wroteComing to this thread late (which based on Fred's post above is probably a good thing).
I was interested in a possible technical reason behind the 1/8,000 limitation. You would think any limitation would be due to sensor readout speed so I did some quick calcs:
Legend:
1 second = 1,000 milliseconds (ms)
1 second = 1,000,000 microseconds (us)
A7rVI sensor metrics (max ES shutter is 1/8,000):
Full-sensor resolution = 9,504 x 6,336
Full-sensor readout speed is 19.6ms = 19,600us
Readout speed per row is 19,600 / 6,336 = 3.09us/row
A7V sensor metrics (max ES shutter is 1/16,000):
Full-sensor resolution = 7008 × 4672
Full-sensor readout speed is 15.1ms (1/66) = 15,100us
Readout speed per row is 15,100 / 4,672 = 3.23us/row
Based on the above, the A7V sensor is slightly slower than the A7VI sensor yet supports a 1/16,000 shutter speed.
If you compare the per-row readout speeds to the 1/16,000 and 1/32,000 exposure times it would appear there is plenty of margin in the readout speed to keep up with the faster shutter speeds, all else being equal.Readout speed and max shutter speed is not necessarily related.
Fuji X-T5 readout is 1/20s but electronic max shutter speed is 1/180,000s.
Consider a row of pixels on a sensor. On a rolling shutter camera, each row of pixels is triggered to start and then stop, one row at a time.
To support 1/16000s, each pixel in that row needs to be able to expose itself for 1/16000s. this is the bottleneck to increasing shutter speed.
Each pixel receives a reset followed by a readout instruction. "Reset" clears existing charge and starts integrating, "readout" dumps existing charge to memory and stops integrating
If the physical silicon of the pixel cannot clear its charge fast enough during the reset, or if the transfer gate cannot open/close cleanly within that microsecond window, the pixel cannot physically achieve a 1/16,000s exposure.
The pixel size on the A7RVI is smaller. the analog components clearing the pixel's charge and the transfer gate are miniaturized compared to the A7RV. On top of this readout speed was increased despite having more pixels total, and additional circuitry is probably used to readout each pixel twice to boost dynamic range in mechanical shutter mode.
Its not that surprising to me that something had to give to achieve this.
A9iii can do 1/80,000s but the pixels are way bigger and the well size is smaller. To reset the pixel less charge needs to be removed since the well size is smaller (no iso 100), and the far bigger pixel size means you can double or triple down on analog components related to clearing charge during a reset, and components to readout the charge.The X-T5 full-sensor readout speed is 1/27, which calculates as:
Full sensor readout speed 1/27 = 37.037ms = 37,037us
X-T5 resolution is 7728 × 5152
37,037 / 5,152 = Readout speed 7.18887us/row
Shutter speed of 1/180,000 = 5.55us integration time
Thus even 1/27 (37ms) can support a shutter speed of 1/180,000, all else being equal (ie, no other technical limitations).
Everything you wrote about other technical limitations is certainly possibly but seems highly unlikely to be the cause of the 1/8000 restriction considering how much timing margin there is for a 1/16,000 shutter speed (62.5us).
Btw, here's my writeup on how rolling shutters work:
You compute a readout speed of 7.18887us/row ~= 1/139,000 s but this is neither an upper nor lower bound to the shutter speed.
The XT-5 camera’s shutter speed can clearly be set faster (1/180,000s) or slower (1/8,000s)
To me the analogue pixel reset mechanism and transfer gate is what bounds the shutter speed. Readout speed is tangential
There’s no rule saying you need to read out one row at once btw, even on rolling shutter cameras. My 1D read out at least 2 rows at once for example… ...Show more →
The shutter speed can't be faster than the row readout speed after reset, otherwise the readout can't keep up with exposure integration, ie the exposure time is less than the row readout time, since integration is ongoing during (at least a portion) of the readout. You're right about the XT-5's 14-bit readout being too slow to support 1/180,000 - I overlooked that in the summary of my calculations. That must mean the XT-5 drops to 12-bit readouts for those fastest shutter speeds, like it does with continuous shooting - its 12-bit readout rate is 1/41.72 (as measured by me here). This translates to a readout rate of 5.99us/row, which is within the measurement margin of error for achieving its 1/180,000 exposure time of 5.55us.
The measured full-sensor readout speed accounts for sensors that can readout multiple rows; in fact, sensors that readout multiple rows concurrently have a slower individual row readout speed, such that the fastest shutter speed the readout supports is the per-row readout rate divided by the number of concurrent rows readout.