FAQ FAQ  Forum Search   Events   Register Register  Login Login

A7ii or A6500 for EF 400 5,6 & macro

Page  <12
Author
Miranda F View Drop Down
Senior Member
Senior Member

Joined: 11 January 2014
Country: United Kingdom
Location: Bristol
Status: Offline
Posts: 2185
Post Options Post Options   Quote Miranda F Quote  Post ReplyReply Direct Link To This Post Posted: 12 September 2017 at 10:48
Originally posted by QuietOC QuietOC wrote:

I have an idea. 1-bit serial DACs were the rage for a while in audio. Imagine an image sensor that just continually streamed bits. No physical shutter. No electronic front curtain. No reading groups of rows after the shutter like the A9. Just a serial constant bit stream from every pixel. When you hit the shutter the bits that were already captured however many you want are stored. Zero or negative lag. No rolling shutter effects. No global shutter needed. You could have multiple overlapping exposures sharing some of the same bits. Maybe the camera always saves a long exposure and you can chop whatever shorter exposures you want from it in post.


Love it!
It it a bit like how the human eye works, except that there the rate at which the bits come is proportional to light and the eye doesn't bother to send zeros ...
Apart from the problem of needing some distributed logic to count the bits, a possible issue is that the streaming rate would be enormous at high light levels. But the central concept does seem to be worth exploring.
A900, A58, 5d, Dynax 4, 5, 60, 500si/600si/700si/800si, various Sony & Minolta lenses, several Tamrons, lots of MF primes and *far* too many old film cameras . . .
 



Back to Top
Miranda F View Drop Down
Senior Member
Senior Member

Joined: 11 January 2014
Country: United Kingdom
Location: Bristol
Status: Offline
Posts: 2185
Post Options Post Options   Quote Miranda F Quote  Post ReplyReply Direct Link To This Post Posted: 12 September 2017 at 11:02
Originally posted by artuk artuk wrote:

In my experience, in good daylight, ISO isn't affected by exposure, since it can be managed by shutter speed - therefore at base ISO, larger formats are "better".
For available light work, the depth of field probably isn't the critical factor - in fact, many photographers shoot available light at open aperture. Again, larger formats will be better than smaller ones.
I understand this idea of "equivalence", but I feel the graph shown can be misleading to someone who does not clearly understand exactly what is being shown - where "real world" uses may favour the larger sensor.


Yes, I feel the idea of equivalence has little use beyond a basic guide to beginners, and as you point out on many occasions a user would neither want nor be able to achieve equivalence, and would need to know what differences there were where they were standing. Most of the attempts to discuss equivalence accurately fall down on the need for assumptions which either aren't realistic or which people don't agree on.

Having just got myself an FF DSLR (for the first time) I am interested, though. My main reasons for getting one were to explore Minolta's lens collection at their intended focal length, and to be able to take a film and a digital body out with the same set of lenses.

Since trying out the A900 after the A58, I find that apart from the obvious differences in viewing angle and DOF, the remaining optical/image differences seem to be ISO, noise, and saturation. And on these I haven't been able to say either camera is better or worse, just that they behave differently and need different settings or checks to give their best. The jpeg engine in the A58 is extremely good up to ISO800 but above that the A900 pulls ahead, and the A58 is more likely to saturate on whites.
A900, A58, 5d, Dynax 4, 5, 60, 500si/600si/700si/800si, various Sony & Minolta lenses, several Tamrons, lots of MF primes and *far* too many old film cameras . . .
Back to Top
artuk View Drop Down
Senior Member
Senior Member

Joined: 06 July 2007
Country: United Kingdom
Location: United Kingdom
Status: Offline
Posts: 3593
Post Options Post Options   Quote artuk Quote  Post ReplyReply Direct Link To This Post Posted: 12 September 2017 at 14:11
Originally posted by QuietOC QuietOC wrote:

DXOMark's dynamic range scores also doesn't really account for noise, and they seem to have problems measuring transmission accurately.

I just sent A65 data to Bill Claff of photonstophotos. Very nice of him to do that for an obsolete camera. I think his criteria for excluding noise is too strict, but it makes a bit more realistic dynamic range measurement.

Bill also worked out the Sony exif data for focus distance. It was magnification not distance. I am not sure if exiftool has been updated yet.


To be honest, I don't understand why their "screen" and "print" dynamic range values are different - since I assume a screen or print will have a dynamic range potentially limited by the medium, not the image. As is often the case with DXO, there is little explanation that really makes it clear - probably deliberately.
Art
Back to Top
artuk View Drop Down
Senior Member
Senior Member

Joined: 06 July 2007
Country: United Kingdom
Location: United Kingdom
Status: Offline
Posts: 3593
Post Options Post Options   Quote artuk Quote  Post ReplyReply Direct Link To This Post Posted: 12 September 2017 at 14:22
Originally posted by QuietOC QuietOC wrote:


I have an idea. 1-bit serial DACs were the rage for a while in audio.


I never really understood "bitstream" DACs.
The analogue signal is digitally encoded into 16 bit words, at a certain sample frequency.
To recover that analogue signal, you must process each of those words, at an accurate sample rate, and convert each one back to an analogue value.

You cannot convert a digital signal back to an analogue signal one bit at a time - you require 16 bit words to generate a value.

I always assumed that what they meant was that instead of sending an entire 16 bit word into the DAC, and converting it, and then processing the next one, single bits were streamed into the DAC and then when 16 had arrived, it produced an output value. This in theory removes some of the critical timing issues where the timing of words into the DAC became slightly out of correct sync ("jitter"). Ironically, tests of jitter with bitstream DACs showed no benefit over well implemented full word DAC systems, and in many cases were much worse - so it clearly didn't solve the timing issues.

Current sensors are architected to read out entire rows of pixels as a stream. They already do this many times a second for live view. What you describe is effectively electronic front curtain, except that it's architected at row level. Making a sensor read out architecture at pixel level would complicate the circuitry by a factor of the number of pixels per row - which would be 6000-8000 depending on sensor resolution. I don't fancy a sensor with 8000 times more wiring - the pixels are going to be really small compared to the wiring!

Edited by artuk - 12 September 2017 at 16:05
Art
Back to Top
Miranda F View Drop Down
Senior Member
Senior Member

Joined: 11 January 2014
Country: United Kingdom
Location: Bristol
Status: Offline
Posts: 2185
Post Options Post Options   Quote Miranda F Quote  Post ReplyReply Direct Link To This Post Posted: 13 September 2017 at 09:34
Originally posted by artuk artuk wrote:

Originally posted by QuietOC QuietOC wrote:


I have an idea. 1-bit serial DACs were the rage for a while in audio.


I never really understood "bitstream" DACs.

It sounds like you still don't - no offense!

Bit stream works by switching the digital level between 0 and 1 at a very high speed, such that the average value of the stream, over 65536 bits, represents a 16-bit decimal. So for audio this means many Mb/s.

Why bother?
There are several reasons. Firstly you don't have to worry about ADC/DAC non-linearity, because it's only 1 bit. Secondly, the actual bandwidth of the signal is much higher (at low resolution) than you would get from a basic 16-bit datastream, so you can get an approximate value much more often (subject to the filtering you do on it). Thirdly, provided the coding of the conversion is done right, it pushes the sampling noise up to a high frequency where it can be filtered more easily.
A900, A58, 5d, Dynax 4, 5, 60, 500si/600si/700si/800si, various Sony & Minolta lenses, several Tamrons, lots of MF primes and *far* too many old film cameras . . .
Back to Top
QuietOC View Drop Down
Senior Member
Senior Member

Joined: 28 February 2015
Country: United States
Location: Michigan
Status: Offline
Posts: 1533
Post Options Post Options   Quote QuietOC Quote  Post ReplyReply Direct Link To This Post Posted: 13 September 2017 at 11:34
It sounds like these people already made one.

nVidia's Tesla Volta V100 graphics processor has a 4096-bit wide memory bus. That is wide enough to cover every row on a 24 MP pixel sensor. It certainly wouldn't be easy to cool a V100 strapped to the back of a image sensor with enough HBM memory to store the data, but it is technology existing today.

Edited by QuietOC - 13 September 2017 at 14:27
A65 30M 35 50 16-50 16-80 16-105 18-135 18-250 55-200 55-300
A5000 LA-EA1 16 20 16-50 18-55
Maxxum 70: 20 24 28 50 85 100M 135 28-135 35-70 35-105 35-200 70-210 75-300 100-200 100-300D
 



Back to Top
Dyxum main page >  Forum Home > Equipment forums > Camera Talk Page  <12

Forum Jump Forum Permissions View Drop Down



This page was generated in 0.078 seconds.

Monitor calibration strip

Dyxum.com - Home of the alpha system photographer

In memory of Cameron Hill - brettania

Find us on Google+

Feel free to contact us if needed.