7D Ghost Images

egordon99 wrote in post #8904709
They should have used Java....No memory management needed for the user (unlike that nasty language called C :lol

If you are good at C, memory management isn't all that hard actually. You just have to remember to code it.

cbfabbri wrote in post #8904464
Let's believe it's just about a firmware update.
I really want to believe on that.

Just cancelled my BH order. I am too far away for a return/recall if needed.

Let's wait to see if it's just about a firmware fix, if ok, I'll get one near future.


Sorry guys, but it's a shame.

So do you do alot of 8fps bursts? Or at least planning on it?

I still have nightmares about pointers

egordon99 wrote in post #8904792
I still have nightmares about pointers

*iAgree=iAgree[0]='\0'​;

So, that is why we see all of those monstrous .NET and Java apps all over the place, which consume A LOT of memory and very slow!

I could see that the 7D may need a 3rd digic process just to run the jvm.

Okay, on 2nd thought this thread is getting pretty geeky!

cbfabbri wrote in post #8904464
Let's believe it's just about a firmware update.
I really want to believe on that.

Just cancelled my BH order. I am too far away for a return/recall if needed.

Let's wait to see if it's just about a firmware fix, if ok, I'll get one near future.


Sorry guys, but it's a shame.

You're entitled to do as you wish, but this is why Canon are always fearful of making announcements regarding rare issues. As soon as they try and let people know about a potential issue, people drop orders.

It's bad that it happened, but I'm glad Canon are able to come out and say it, and to keep us informed.

egordon99 wrote in post #8904792
I still have nightmares about pointers

How about arrays of pointers to functions that return pointers to other functions.

God, I can't even remember how to code that stuff any more, where my battered old copy of K&R....

Heh, I'm matlab man now, mostly code for maths models anyway.

Should try and find a reason to do some C before it all leaks out of these old ears and makes a mess on the floor....

TeamSpeed wrote in post #8904882
Looking at one right now in fact!

/* Transactional functions to invoke based on message */
INDEX_BROKER broker[] = {
/* Msg ID FROM MESSAGE, FUNCTION TO CALL (int func_xxx(MESSAGE *))
* --------------- --------------- */
IMG_MSGS, img_view_requests,
IMG_MSGS+1, img_reqt_requests,
IMG_MSGS+2, img_xref_requests,
IMG_MSGS+3, bnsf_wb_requests,
IMG_MSGS+4, scan_grp_requests,
/*********** ADD NEW TRANSACTIONS ABOVE THIS LINE ***************/
INDEX_EOB, NULL
};

This looks similar to the problems that we have in astro imaging with sensitive detectors. Residual images.

Not saying that it's the exact cause, but it sure looks similar in the images.

http://www.narrowbandi​maging.com …peaker_comments​_paper.pdf

Yes, I agree, it's a good point.
I just dropped the order since it was not shipped yet and I am really far away from the shop and in case of return/recall, if needed after the proper investigation on a fix, I would not like to have shipment/return/taxes costs, etc. Note: Canon has no international warranty valid in Brazil, so, no free warranty related fix here.

As soon everything is clear and fixed, I plan to get one since I believe it can be a great tool for me, considering its cost and benefits.

I have shot several thousand images with my 7D including BIF shots and have not noticed any ghosting problem. However, this situation is not something new or unique to this particular sensor. An erase pulse is used to clear all of the data in memory (memory refers to stored charge on the sensor and not some buffer), but unlike a read operation, it does not have to strobe each location individually in order to clear the whole chip which means that the erase operation takes far less time than a read operation -- but, it does require a finite amount of time.

The reason that erasing takes a finite amount of time is that each photodiode has a tiny capacitor in which its charge is stored. Theoretically, the stored charge never goes completely to zero no matter how long it is discharged, but there is a practical point that can be established as the zero reference floor. If the erase pulse happens to be too short there could be some residual charge remaining (which would sum with the next exposure to produce the "ghost").

It appears that there are several options to remedy the situation and I see no reason why they could not be implemented in firmware. The simplest would be to extend the length of the erase pulse by a few microseconds. This additional time line event is not significant enough to have an effect on the eight FPS high speed burst rate. Another possible solution might be to change the zero reference point for the sensor.

I suspect that the problem is showing up on some units because of normal manufacturing tolerances in the sensor along with timing tolerances that leave a bit too much residual charge -- but, apparently even then the exposure has to be pushed a couple stops in post processing for this to be noticeable.

cbfabbri wrote in post #8904949
Yes, I agree, it's a good point.
I just dropped the order since it was not shipped yet and I am really far away from the shop and in case of return/recall, if needed after the proper investigation on a fix, I would not like to have shipment/return/taxes costs, etc. Note: Canon has no international warranty valid in Brazil, so, no free warranty related fix here.

As soon everything is clear and fixed, I plan to get one since I believe it can be a great tool for me, considering its cost and benefits.

Reasonable I would say.

The secret behind the new sensor is that they have an old P40 CRT in there. It's burn in on the screen we see now.

TeamSpeed wrote in post #8904785
So do you do alot of 8fps bursts? Or at least planning on it?

Yes, just getting 7d to do what 5D2 cannot in terms of speed, using it exclusivelly for bifs, sports, and fast action photography.

I believe it's a great camera. I just need to wait to see how this fix will happen (if it is really just firmware related) and I'll get one as soon as possible.