Go Back   Steve's Digicams Forums > Digital SLR and Interchangeable Lens Cameras > Canon EOS dSLR

Reply
 
Thread Tools Search this Thread
Old Jul 26, 2004, 9:36 PM   #21
NHL
Senior Member
 
NHL's Avatar
 
Join Date: Jun 2002
Location: 39.18776, -77.311353333333
Posts: 11,547
Default

eric s wrote:
Quote:
The switch to Digic II wouldn't be for the faste write time. It would be to reduce parts and inventory overhead. Get as many cameras off of Digic II is a win for them, as they'll be making it for awhile into the future.
This only benefits Canon... what's in it for me or for the average Joe who rarely fills up the buffer plus most folks don't even use RAW -> fix the buffering algorithm so it can store more JPEGs instead of the same number of frame as RAW for the same amount of internal memories (another firmware upgrade)!
NHL is offline   Reply With Quote
Old Jul 26, 2004, 10:04 PM   #22
Senior Member
 
hedwards's Avatar
 
Join Date: Jul 2004
Posts: 190
Default

NHL wrote:
Quote:
This only benefits Canon... what's in it for me or for the average Joe who rarely fills up the buffer plus most folks don't even use RAW -> fix the buffering algorithm so it can store more JPEGs instead of the same number of frame as RAW for the same amount of internal memories (another firmware upgrade)!
Not that i know anything more than you guys, but i don't think it is a foregone conclusion that when the future jpegs are in the buffer they are jpegs. Before they are converted to jpegs they have to be in the buffer. The compression is presumably on the fly as the camera writes to disk. If they were jpegs at that point canon would allow more jpgs per burst than crws.
hedwards is offline   Reply With Quote
Old Jul 26, 2004, 10:17 PM   #23
NHL
Senior Member
 
NHL's Avatar
 
Join Date: Jun 2002
Location: 39.18776, -77.311353333333
Posts: 11,547
Default

hedwards wrote:
Quote:
Not that i know anything more than you guys, but i don't think it is a foregone conclusion that when the future jpegs are in the buffer they are jpegs. Before they are converted to jpegs they have to be in the buffer. The compression is presumably on the fly as the camera writes to disk. If they were jpegs at that point canon would allow more jpgs per burst than crws.
True, but check the Minolta D7Hi/A1/A2 -> It has been done, one can accumulate more JPEGs shots than RAW from the same buffer! The limitation is the bursty and slow traffic to the CF cards, remember the long flush time (ie the resulting files need to be designed for the worst case anyway to be parsed to the slowest serial memory available on the market)?

Also why can they "reduce parts and inventory overhead by the Digic II" and not apply the same logic with the 8Mp sensor? More volume = lower price right? We could even see a DRebel/300D mrk II with the same sensor and the cost for the device is even cheaper still... correct?
NHL is offline   Reply With Quote
Old Jul 26, 2004, 11:12 PM   #24
Senior Member
 
hedwards's Avatar
 
Join Date: Jul 2004
Posts: 190
Default

NHL wrote:
Quote:
Check the Minolta D7Hi/A1/A2 -> It has been done, one can accumulate more JPEGs shots than RAW from the same buffer! The limitation is the bursty and slow traffic to the CF cards, remember the long flush time (ie the pictures need to be buffered anyway to be parsed to the slower serial memory)?
Apparently you can add the 1d mkII to that list. Looks like Canon may consider that to be a pro feature at the moment. As to whether that means they couldn't add the feature through a firmware update or not i know better than to speculate on. Even if they could and did it isn't going to affect my shooting though I don't use the depth of the current buffer arrangement. Digic II would make a bigger difference to me.

NHL wrote:
Quote:
Also why can they "reduce parts and inventory overhead by the Digic II" and not apply the same logic with the 8Mp sensor? More volume = lower price right? We could even see a DRebel/300D with the same sensor and the cost for the device is even cheaper still... correct?
I agree the price for each sensor would go down. The question is will the price per sensor go down enough to justify the hit on the 1d mkII sales and be inexpensive enough to put in a 1400-1500 dollar camera. Digic II may help with battery life as well as the speed of the bursts. Since Digic seems to be used reducing noise that might have allow for a proper 3200 rather than gain, and adjustment.
hedwards is offline   Reply With Quote
Old Jul 27, 2004, 6:54 AM   #25
NHL
Senior Member
 
NHL's Avatar
 
Join Date: Jun 2002
Location: 39.18776, -77.311353333333
Posts: 11,547
Default

hedwards wrote:
Quote:
I agree the price for each sensor would go down. The question is will the price per sensor go down enough to justify the hit on the 1d mkII sales and be inexpensive enough to put in a 1400-1500 dollar camera. Digic II may help with battery life as well as the speed of the bursts. Since Digic seems to be used reducing noise that might have allow for a proper 3200 rather than gain, and adjustment.
1. It doesn't have to be $1500... -> $2000 - $2499 MSRP is still OK to "feel" out the market demand... and I'm sure they will sell out everyone of them initially (me included) at that price.
2. They can "improve" the DRebel/300D back to the features level found on the 10D and still undercut the D70 -> This is all firmware upgrade (heck you can even get an "unofficial" one already) and hold the competition @ bay. Also most of us still have the 10D or will kept it as a backup :roll:
3. Remember there's a Minolta coming with internal IS for the same price, ie they need to protect their lenses lines with an attractive body not available anywhere else...
4. You guys are spoiled... When is the last time anyone used ISO 1600? I was happy with ASA25, 64 or even 160 slides, and 36 shots was the max on a roll. Now everyone can shoot all day on a 4G Microdrive and still complain about battery life... Maybe one should go easy on that IS lens (ie turn it off to save battery)? :-)
NHL is offline   Reply With Quote
Old Jul 27, 2004, 7:10 AM   #26
Senior Member
 
Join Date: Dec 2002
Posts: 5,803
Default

Quote:
Originally Posted by NHL
This only benefits Canon... what's in it for me or for the average Joe who rarely fills up the buffer plus most folks don't even use RAW -> fix the buffering algorithm so it can store more JPEGs instead of the same number of frame as RAW for the same amount of internal memories (another firmware upgrade)!
Come on! This is just a little self/Customer -centered! Having worked designing consumer electronic devices, I can guaranty you that companies make some decisions which only benefit them while planning features on a device.

If Canon switched to Digic II, they would get faster performance, maybe some noise reduction algorithms (are those in the chip or more in hardware/sensors design?) Maybe better jpg compression algorithms? But the big win would be to reduce the need for the Digic I. If they build it, it lets them scale back their fab plant. Working towards freeing up that fab plan space is a worth while goal. If they contract the building out then it saves them money, which can be better spent on something more profitable.

I seem to remember, somewhere back in the merky recesses of my brain, that Canon didn't have a separate buffer for the unprocessed jpg data and that it took a couple of processing/copying steps to get from RAW to jpg. Also, that it was all done in the same buffer that RAW uses for its buffering. Maybe the Minolta added extra (maybe faster?) memory just for this purpose? Just thinking creatively.

If I remember correctly most sensors are made with multiple data pumps to get data off it. I bet the same is the true of the 1D-II's sensor. So an obvious difference (if they reused the sensor) could be to just make it slower by not enabling the 1/2 the pumps. Heck, a chip with a defective pump is now no longer thrown away but used in the 10D-II. Considering the cost of a sensor, this would be a win for both Canon and Us.

But I would be surprised if they moved away from the 1.6x crop sensor size. The physically smaller sensor is cheaper to make... is it "cheaper enough" to justify two sensor lines? I don't know. On the other hand, it looks like increasing the MP of a sensor that size will start to have non-linear noise characteristics (at least, that seems to be true looking at other sensors & their noise.) So it cuts both ways. They will eventually have to stop using that type & size of sensor due to increased noise... but when? Can other suppression techniques over come it? (will those methods be too expensive in themselves?)

This speculation is fun! :-)

Eric
eric s is offline   Reply With Quote
Old Jul 27, 2004, 7:22 AM   #27
NHL
Senior Member
 
NHL's Avatar
 
Join Date: Jun 2002
Location: 39.18776, -77.311353333333
Posts: 11,547
Default

eric s wrote:
Quote:
If I remember correctly most sensors are made with multiple data pumps to get data off it. I bet the same is the true of the 1D-II's sensor. So an obvious difference (if they reused the sensor) could be to just make it slower by not enabling the 1/2 the pumps.
This an inefficient way to use silicon:
1. CCD sensors are obviously just analog memory elements
2. CMOS sensors already have extra "non-light sensitive" processing circuit to change theses charges to digital hence reducing the available area for the actual light collecting element -> Remember the "fill-factor"?
If they incorporated the "multiple data pumps" into the sensor wouldn't it reduce the available area even further for the actual light collecting elements and increase the noise of CMOS even further and even make this sensor larger than 1.6x?

Now if the dual data pumps above are outside the sensor, then there would be plenty of opportunity to cut cost wouldn't it? If you need the speed get the 1D mrkII and pay for it otherwise a de-featured 8Mp will do nicely...
NHL is offline   Reply With Quote
Old Jul 27, 2004, 9:42 AM   #28
Member
 
Join Date: Mar 2004
Posts: 54
Default

The issue with larger sensors, which has already been encountered in the past with CPU chips, LCD screens etc., is that the larger you go, the cost increases exponentially.

First, consider size. Using geometry, the increase in area is determined by the (smaller crop/larger crop)^2. That is (1.6/1.3)^2 = 1.5. This means that the area of a 1.3 crop sensor is actually 50% larger than a 1.6 crop sensor.

Second, consider the issue with yield with an increases sensor size. Bill posted a fine explanation on another forum:
Quote:
The major problem is yield. This was highlighted with the old Pentium Pro which, at the time, was significantly larger than other CPUs available. Basically an error rate can be reduced to the probability of a single transistor being defective. That chance is very very small, say 1 in 300 million. Assuming for a second that a defective transistor kills the chip then you end up with an exponential curve of number of transsitors vs yield.

Basically the probability of a chip being functional, if 1 transistor in X is non-functional, is:
Yield = (1-1/X)^T

where T is the number of transistors. Now this curve has interesting properties. Basically the failure rate is almost zero up to a certain point and then drops off fairly rapidly. A 50 million transistor chip could have a 90% yield whereas at 75 million the yield is 10%.

Lower yield has two effects. The first and most obvious one is that it reduces the number of chips per wafer.

The second is less obvious but nonetheless important. Each chip produced needs to be verified. This testing is time-consuming (even though automated) and expensive and is a significant part of the cost basis. Lets say it takes N seconds to test a single chip (assume for a second that full testing is required for the full duration before viability can be determined; there are no early fails) with yield Y (1 = 100% yield, 0.5 = 50%, etc).
Average testing time = N / Y

So 500 seconds and a yield of 90% = 555seconds. Reduce yield to 80% and time jumps up to 675 seconds. At 50% its 1000 seconds. At 20% its 2500 seconds.

Now some simplifications have been made here. For example not all faulty transistors make the product unviable (eg dead pixels on an LCD are acceptable to a point). Testing too could provide early fails. These change the formulae but not the underlying principle.

Which is that chips with a large wafer area become more expensive far in excess of the rate of growth of wafer area. Twice the wafer area could mean it costs ten times as much (or more).

Lower costs mean lower volume which means a bigger slice of the R&D budget must be recovered on each chip which again makes them even more expensive.

There might be a new DSLR from Canon thats cheaper than the 1D2 and has a 1.3x crop (eg 3D) but I don't think it will (yet) be at the 10D price point. The 10D I think will remain at the 1.5-1.6 crop factor level.
Bottom line, moving up to the larger sensor isn't a "slam dunk" or "small cost increase" by any means. It isn't impossible, but there are strong arguments against it.

Graeme
Graeme Shiomi is offline   Reply With Quote
Old Jul 27, 2004, 10:10 AM   #29
NHL
Senior Member
 
NHL's Avatar
 
Join Date: Jun 2002
Location: 39.18776, -77.311353333333
Posts: 11,547
Default

Graeme Shiomi wrote:
Quote:
Bottom line, moving up to the larger sensor isn't a "slam dunk" or "small cost increase" by any means. It isn't impossible, but there are strong arguments against it.
Excuses... If Canon doesn't do it, may be Nikon, Minolta or other OEMs will: http://www.atmel.com/dyn/resources/p...ts/doc2133.pdf :idea:
NHL is offline   Reply With Quote
Old Jul 27, 2004, 12:23 PM   #30
Senior Member
 
hedwards's Avatar
 
Join Date: Jul 2004
Posts: 190
Default

NHL wrote:
Quote:
4. You guys are spoiled... When is the last time anyone used ISO 1600? I was happy with ASA25, 64 or even 160 slides, and 36 shots was the max on a roll. Now everyone can shoot all day on a 4G Microdrive and still complain about battery life... Maybe one should go easy on that IS lens (ie turn it off to save battery)? :-)
Actually i used the iso 1600 last night. And before you ask i did use the iso 3200 last week. I personally won't mess with microdrives, but still being able to cram 150-160 photos onto a flash card is a bit spoiled. A 4gb drive would probably last me all month. As for my last roll of film, it was fuji reala 100.

I would really like to see a 6 mp 1.3x crop factor camera. The low light performance should be astounding. Not sure what that would do to the cost, but the light collectors on it could be huge.

Unless Canon is planning to add some big features i wouldn't expect for the price to be much different than the 10d that we know and love. The are not going to place a camera at $1000 to compete with the rebel, and honestly how many people are going to buy the upgrade if it costs and additional 500 dollars, when the current generation does tons already.
hedwards is offline   Reply With Quote
 
Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



All times are GMT -5. The time now is 12:15 AM.