Home · Register · Join Upload & Sell

Moderated by: Fred Miranda
Username  

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Sony Forum | Join Upload & Sell

1       2       3              7              9       10       end
  

Archive 2015 · A7RII 12-bit mode. How to avoid it.

  
 
snapsy
Online
• • • • • •
Upload & Sell: On
p.8 #1 · p.8 #1 · A7RII 12-bit mode. How to avoid it.


Fred Miranda wrote:
We still have not figured out if a combination of precision loss modes (13 bits to 12 bits) would degrade IQ even further.
For example:
Combining continuous bracketing + silent shutter
Bulb + LENR
etc...

Something came to mind when using continuous bracket today.
If you lose a full stop when using continuous bracketing mode, many bracketing settings could potentially provide less DR than shooting one image in the first place.

For example: when choosing continuous bracket [-0.5, 0, 0.5] you would end up with the SAME dynamic range as if taking a single shot. The stop you gain by bracketing you lose with the
...Show more

Fred, that would be true if that's the technique one uses to bracket. I guess I was speaking to how I bracket, which may be a bit different than many others. My darkest exposure actually starts with ETTR, spot metered + 2EV above the brightest highlight I want retained and then I bracket in 2EV brighter steps from there. So for example my typical bracket sequence relative to what many would consider a normal single exposure is +2EV, +4EV, and +6EV. I don't include any negative Exposure Comps in my bracket because they're just wasted exposures that don't add any information to the combined sequence.



Aug 14, 2015 at 11:26 PM
Fred Miranda
Offline
Admin
Upload & Sell: On
p.8 #2 · p.8 #2 · A7RII 12-bit mode. How to avoid it.


snapsy wrote:
Fred, that would be true if that's the technique one uses to bracket. I guess I was speaking to how I bracket, which may be a bit different than many others. My darkest exposure actually starts with ETTR, spot metered + 2EV above the brightest highlight I want retained and then I bracket in 2EV brighter steps from there. So for example my typical bracket sequence relative to what many would consider a normal single exposure is +2EV, +4EV, and +6EV. I don't include any negative Exposure Comps in my bracket because they're just wasted exposures that don't add any
...Show more

In this case, the 1-stop read noise/DR penalty is definitely negligible.

When I bracket, I use evaluative metering to avoid too many "blinkies" and then add 2EV and start my bracketing sequence (-2,0,+2). I should try your method and see if I get even cleaner shadows.



Aug 15, 2015 at 12:25 AM
saintiwari
Offline
• •
Upload & Sell: Off
p.8 #3 · p.8 #3 · A7RII 12-bit mode. How to avoid it.


Paedon the ignorance, but won't the highlights get blown at +2 +4 +6 ?

snapsy wrote:
Fred, that would be true if that's the technique one uses to bracket. I guess I was speaking to how I bracket, which may be a bit different than many others. My darkest exposure actually starts with ETTR, spot metered + 2EV above the brightest highlight I want retained and then I bracket in 2EV brighter steps from there. So for example my typical bracket sequence relative to what many would consider a normal single exposure is +2EV, +4EV, and +6EV. I don't include any negative Exposure Comps in my bracket because they're just wasted exposures that don't add any
...Show more



Aug 15, 2015 at 02:17 AM
snapsy
Online
• • • • • •
Upload & Sell: On
p.8 #4 · p.8 #4 · A7RII 12-bit mode. How to avoid it.


Fred Miranda wrote:
In this case, the 1-stop read noise/DR penalty is definitely negligible.

When I bracket, I use evaluative metering to avoid too many "blinkies" and then add 2EV and start my bracketing sequence (-2,0,+2). I should try your method and see if I get even cleaner shadows.


The frustration I have with my method is that most camera firmware bracketing implementations require me to jump through hoops to support it. For example, on stock Canon firmware if I want to do a 3-shot bracket for +2EV, +4EV, and +6EV here's what I have to do:

1. Common step for all camera brands: set camera to spot-meter mode and dial in +2EV Exposure Comp. I do the EC so that a centered meter reading gives me my initial +2EV ETTR reading.

2. Spot meter on the brightest highlight I want retained, then change the shutter speed so that the meter is +2EV above center. This brings my base exposure to +4EV above spot; +2EV from the previous EC which centers the meter for a +2EV reading and the next +2EV by using a shutter speed that the LCD meter display reads as +2EV.

3. I can now shoot my 3-exposure -:0:+ bracket. The "0" exposure will correspond to +4EV above spot; the minus exposure will correspond to +2EV above spot; the plus exposure will correspond to +6EV above spot

Magic Lantern makes this easy - in addition to supporting the normal -:0:+ bracket it also supports a ++:++:++ bracket. This means I only have to do my initial ETTR metering and then the exposures of the bracket will be +EV beyond that. This is partly why I've been reluctant to give up my 5D classics - the bracketing is so easy with ML

Nikon supports something similar to the Magic Lantern feature although it has a quirk in the implementation that requires knowing when to set the base exposure relative to when bracketing is enabled. I wont go into it here unless someone is interested.

Final point: some MILC don't accurately meter at +2EV above spot, because an MILC meters off the image sensor and +2EV brings it close to saturation and this leads to incorrect meter readings.



Aug 15, 2015 at 02:49 AM
snapsy
Online
• • • • • •
Upload & Sell: On
p.8 #5 · p.8 #5 · A7RII 12-bit mode. How to avoid it.


saintiwari wrote:
Paedon the ignorance, but won't the highlights get blown at +2 +4 +6 ?

My base exposure is the darkest and is metered to put my important highlights at the right edge of the histogram. The +4 and +6 exposures are done to yield brighter shadow exposures.



Aug 15, 2015 at 02:51 AM
kwalsh
Offline
• • • •
Upload & Sell: Off
p.8 #6 · p.8 #6 · A7RII 12-bit mode. How to avoid it.


Fred Miranda wrote:
In this case, the 1-stop read noise/DR penalty is definitely negligible.

When I bracket, I use evaluative metering to avoid too many "blinkies" and then add 2EV and start my bracketing sequence (-2,0,+2). I should try your method and see if I get even cleaner shadows.


I shoot m43 and I do what snapsy does though my ETTR parameters are slightly more conservative and I use flashing highlights instead of spot meter to get there. But the results for shadows and even mid-tones like a polarized sky are very impressive when your shortest exposure in a bracket is close to ETTR and you go four to six stops up from there:

http://www.kenandchristine.com/photos/i-PLKNkQL/0/XL/i-PLKNkQL-XL.png


Obviously m43 has way more of a noise problem than FF does so it is more critical that I do this. But the point is if you are going to HDR stack then a bracket that starts with the middle exposure at what the camera thinks is "right" means the HDR merge is likely to not use one of your exposures at all and only small regions of your middle exposure. You'll end up with most of the final image being not much better than a single ETTR frame would have been.

Cheers!



Aug 15, 2015 at 07:44 AM
hiepphotog
Online
• • • •
Upload & Sell: On
p.8 #7 · p.8 #7 · A7RII 12-bit mode. How to avoid it.


Wow, very good point guys. I have to try this out to get a hang of it.


Aug 16, 2015 at 03:28 PM
timballic
Offline
• • •
Upload & Sell: Off
p.8 #8 · p.8 #8 · A7RII 12-bit mode. How to avoid it.


Fred Miranda wrote:
This is not something new to the A7RII as all previous A7 series bodies have the same limitation. At least Sony documented this issue instead of just reducing the bits and stay quiet about it. Photographers can now decide if this camera or any other A7 series camera suit their needs.




These details in the A7rII Help Guide, do not appear in the original A7/A7r Help guide:
"RAW images recorded with this camera have a resolution of 14 bits per pixel. However, resolution is limited to 12 bits in the following shooting modes:
[Long Exposure NR]
[BULB]
Continuous shooting (including continuous shooting in [Superior Auto] mode.)
[Silent Shooting]"



I assume the noise reduction and Bulb settings reduction to 12 bit, still apply to the A7?



Aug 20, 2015 at 02:55 AM
Fred Miranda
Offline
Admin
Upload & Sell: On
p.8 #9 · p.8 #9 · A7RII 12-bit mode. How to avoid it.


Yes, without a doubt.

timballic wrote:
These details in the A7rII Help Guide, do not appear in the original A7/A7r Help guide:
"RAW images recorded with this camera have a resolution of 14 bits per pixel. However, resolution is limited to 12 bits in the following shooting modes:
[Long Exposure NR]
[BULB]
Continuous shooting (including continuous shooting in [Superior Auto] mode.)
[Silent Shooting]"


I assume the noise reduction and Bulb settings reduction to 12 bit, still apply to the A7?




Aug 20, 2015 at 10:25 AM
RustyBug
Offline
• • • • • • •
Upload & Sell: On
p.8 #10 · p.8 #10 · A7RII 12-bit mode. How to avoid it.


Long Exposure, BULB, Continuous ... not a big surprise that shaving some bits is employed as a tradeoff.

But ... Silent shooting

Why would that need to reduce from 14 bits to 12 bits?



Aug 20, 2015 at 10:39 AM
DamonJoyce
Offline
• •
Upload & Sell: On
p.8 #11 · p.8 #11 · A7RII 12-bit mode. How to avoid it.


My guess would be that full electronic readout (silent shutter uses two electronic "curtains"), requires a lot of data flowing on the bus, and dropping to 12-bit is the only way to handle all that data.


Aug 20, 2015 at 10:42 AM
RustyBug
Offline
• • • • • • •
Upload & Sell: On
p.8 #12 · p.8 #12 · A7RII 12-bit mode. How to avoid it.


Snapsy ... interesting. I would have thought many would already be bracketing like that. Funny, I get bashed for suggesting the very similar thing @ -1, +2, +5. Suddenly, you're a hero.

Oh well, maybe the light has finally come on and folks will understand much better from you than I was ever able to convey regarding bracketing. Carry on.

Although, I'm not sure why all the convoluted approach to setting your brackets is necessary.



Aug 20, 2015 at 10:46 AM
kwalsh
Offline
• • • •
Upload & Sell: Off
p.8 #13 · p.8 #13 · A7RII 12-bit mode. How to avoid it.


RustyBug wrote:
But ... Silent shooting

Why would that need to reduce from 14 bits to 12 bits?


---------------------------------------------

DamonJoyce wrote:
My guess would be that full electronic readout (silent shutter uses two electronic "curtains"), requires a lot of data flowing on the bus, and dropping to 12-bit is the only way to handle all that data.


Silent shooting (or any electronic rolling shutter) is typically run with a read out faster than when using a mechanical shutter to cover the sensor during read out. This is to reduce rolling shutter artifacts. When you have a mechanical shutter covering the sensor (as it would in EFCS) you can "take your time" reading out the sensor.

Because of the faster read out in "silent shooting" you run into two problems. One is the digital data bus bandwidth as Damon suggests. The other is that the faster you run the ADC on read out the noisier it is. Said another way, read noise increases as you read the pixels faster. So even if you have the digital data bus speed to handle the fast read out you may in fact end up with more noise in the conversion than at "normal" read speeds such that there is no point in going to 14-bit.

This is a pretty common problem with electronic rolling shutters (i.e. "silent" shutters) and many of the m43 cameras that implement them step down to 10-bit rather than their usual 12-bit read out. Most obviously this was the case with Panasonic that earlier models read out slower but at 12-bit. In an attempt to reduce rolling shutter artifacts they increased the read out speed but with the result that read out was at 10-bit instead resulting in a pretty significant hit to base ISO dynamic range.

In general - faster read speeds mean noisier data. That's the nice thing about EFCS, you don't get shutter shock from the first curtain but you still have the second curtain covering the sensor so you can optimize the read out speed for low noise rather than having to rush it to combat rolling shutter artifacts.



Aug 20, 2015 at 11:28 AM
secondclaw
Offline
• • •
Upload & Sell: On
p.8 #14 · p.8 #14 · A7RII 12-bit mode. How to avoid it.


Silent shooting drop to 12-bit makes sense. But what about bulb mode?


Aug 20, 2015 at 11:38 AM
Jman13
Online
• • • • • •
Upload & Sell: On
p.8 #15 · p.8 #15 · A7RII 12-bit mode. How to avoid it.


This is wild speculation, but is it possible that the photon counter (if you will) in Sony cameras dumps the data every 30s to memory during bulb mode, and then the values are added at the end (or more likely added and then the new value is stored in its place), and as such the dump needs to happen very quickly in order to reset the sensor with minimal interruption? Perhaps this produces better noise characteristics even with the 12 bit drop than a 14-bit read with a single 'exposure' as far as the camera is concerned?

Pure conjecture, of course.



Aug 20, 2015 at 11:47 AM
hiepphotog
Online
• • • •
Upload & Sell: On
p.8 #16 · p.8 #16 · A7RII 12-bit mode. How to avoid it.


Jman13 wrote:
This is wild speculation, but is it possible that the photon counter (if you will) in Sony cameras dumps the data every 30s to memory during bulb mode, and then the values are added at the end (or more likely added and then the new value is stored in its place), and as such the dump needs to happen very quickly in order to reset the sensor with minimal interruption? Perhaps this produces better noise characteristics even with the 12 bit drop than a 14-bit read with a single 'exposure' as far as the camera is concerned?

Pure conjecture, of
...Show more

This resetting mechanism is still under R&D so I doubt that's what Sony is doing. See MIT's Modulo sensor.

I think bulb mode drop is there because it happens to be outside the usual single-shot exposure (from 1/8000 to 30sec). It must be a switch of sort in the ADC as soon as it's not the single-shot mode.



Aug 20, 2015 at 12:15 PM
Fred Miranda
Offline
Admin
Upload & Sell: On
p.8 #17 · p.8 #17 · A7RII 12-bit mode. How to avoid it.


snapsy wrote:
The frustration I have with my method is that most camera firmware bracketing implementations require me to jump through hoops to support it. For example, on stock Canon firmware if I want to do a 3-shot bracket for +2EV, +4EV, and +6EV here's what I have to do:

1. Common step for all camera brands: set camera to spot-meter mode and dial in +2EV Exposure Comp. I do the EC so that a centered meter reading gives me my initial +2EV ETTR reading.

2. Spot meter on the brightest highlight I want retained, then change the shutter speed so that the meter
...Show more

I was in Puerto Vallarta for a couple of days and tested this out. My bracketing method is actually similar to yours.
The difference it that I use evaluative metering to expose the entire scene (instead of spot metering the highlights).
The camera evaluative metering gives me a 'good' exposure saving highlights as much as possible but not entirely. Then, I add 2EV to that exposure reading and start my silent shutter/continuous (-2,0,+2) bracketing. I will post some samples soon.



Aug 20, 2015 at 12:29 PM
Jman13
Online
• • • • • •
Upload & Sell: On
p.8 #18 · p.8 #18 · A7RII 12-bit mode. How to avoid it.


The Modulo is about infinite DR, though, not about this. That mechanism requires instant resetting or precise timing of well fill: this simply needs to bleed the data off and store it, then reset. If that can be done in 1/30s, you wouldn't miss much on a 30+ second exposure. The Modulo though, needs to work for even short exposures in the hundredths of a second.

hiepphotog wrote:
This resetting mechanism is still under R&D so I doubt that's what Sony is doing. See MIT's Modulo sensor.

I think bulb mode drop is there because it happens to be outside the usual single-shot exposure (from 1/8000 to 30sec). It must be a switch of sort in the ADC as soon as it's not the single-shot mode.




Aug 20, 2015 at 01:40 PM
kwalsh
Offline
• • • •
Upload & Sell: Off
p.8 #19 · p.8 #19 · A7RII 12-bit mode. How to avoid it.


Jman13 wrote:
The Modulo is about infinite DR, though, not about this. That mechanism requires instant resetting or precise timing of well fill: this simply needs to bleed the data off and store it, then reset. If that can be done in 1/30s, you wouldn't miss much on a 30+ second exposure. The Modulo though, needs to work for even short exposures in the hundredths of a second.


Pretty much all CMOS sensors can do this already and it actually doesn't matter what the read speed is to do it (the bulb case I mean, as you say the fast read/reset for the Modulo is something new). You are doing a rolling reset and read, so the only "missed" time is the length of time for a single row read/reset which is on the order of (total sensor read time)/(number of rows) which puts it in the few microsecond range at worst.

You can also non-destructively read a CMOS sensor during exposure. Most astronomy sensors have done this for ages and again it is something pretty much every CMOS sensor out there could do if desired. I'm not sure if this is what Olympus is doing on their LIVE BULB settings or not, but it is one way they could be doing it.

As to reading/resetting multiple times and combining the exposures in the digital domain that is certainly possible. I'm not sure what the effect on noise would be. As far as kTC reset noise goes you'd actually probably be making things worse and not better by doing multiple resets and reads, but of course that is just one source of read noise.

So anyway, it is certainly possible they could be doing bulb as multiple reads but even if they did they wouldn't need to increase the read speed to do it. Not clear exactly why it is 12-bit in that case (nor why it is for long exposure NR for that matter).

One thing it is pretty safe to say is that for very long exposures the DR doesn't justify 14-bits to begin with, you get plenty of noise from dark current by the time you are getting out to 30 second exposures without a cooled sensor I suspect. And same case for anything the requires long exposure NR. So it is possible that they do 12 bit in those cases because there is no point in doing 14 bit.





Aug 20, 2015 at 01:59 PM
Fred Miranda
Offline
Admin
Upload & Sell: On
p.8 #20 · p.8 #20 · A7RII 12-bit mode. How to avoid it.


Here is one of the bracketed images I took from my trip.
I used these bracketing settings and the shadows looked very clean.

So, to snapsy's point, as long as your bracket gets a lot of data for the shadows, it does not matter if you use continuous and silent shutter to achieve great DR.

____________

Shot in Continuous Bracketing with self-timer in Silent Shutter mode (so in 12-bit)

PS: the longest shutter speed was 30 seconds




Sony A7RII and FE 16-35/4 @16mm, f/10, ISO 100






Shadow detail (1:1)






30 second frame(+2) from (-2,0,+2) bracket






Another one from another day there




Aug 20, 2015 at 04:43 PM
1       2       3              7              9       10       end




FM Forums | Sony Forum | Join Upload & Sell

1       2       3              7              9       10       end
    
 

You are not logged in. Login or Register

Username       Or Reset password



This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.