April 14th, 2005, 12:58 AM | #361 |
New Boot
Join Date: Apr 2005
Location: Riga, Latvia
Posts: 13
|
did some reading here. Looking for good sensor choice to use in HomeMade HD. I have found, that the only one 2/3 that do electronic snapshot is the ibis5a. Am i wrong here?
And and what do you feel about using rolling shutter sensors with mechanical shutter like http://sunex.com/News/Press8.htm or even homemade rotating shatter? |
April 14th, 2005, 09:48 AM | #362 |
Inner Circle
Join Date: May 2003
Location: Australia
Posts: 2,762
|
Yes, for cmos that we consider Ibis5 is the one for snapshot shutter (but lupa series, fast motion micron, and any other expensive ones have snaphot too).
For mechanical shutter, I think noise and reliability VS cost issues (I think you will have to buy an expensive one to get reliability, not best to be homemade). But I see reference to rotating mirror shutter. I have not seen these, but hopefully simpler, more reliable and able to be made at home. But I am sure others here will have better opinion than mine. Thanks for coming Konstantin Wayne. |
April 14th, 2005, 04:16 PM | #363 |
Silicon Imaging, Inc.
Join Date: May 2004
Location: Troy, NY USA
Posts: 325
|
Konstantin,
Watch whether the sensor can overlap the exposure and readout times. With the IBIS-5A, they are sequential. If the max readout time is 1/40 sec (25msec) and you want a frame rate - integration + readout of 1/24 sec (41.7msec) then your max integration time is 16.7msec. Since the sensor isn't too light sensitive, you might have low light problems This is not the case with some of the other sensors Wayne talks about. Steve
__________________
Silicon Imaging, Inc. We see the Light! http://www.siliconimaging.com |
April 15th, 2005, 06:11 PM | #364 |
New Boot
Join Date: Apr 2005
Location: Riga, Latvia
Posts: 13
|
well, i've got more stupid questions:
Steve: 60 fps chip has lower readout than 40 fps but, how to calculate from sensor data, what readout time is? I'm curious how white balance is achieved in single chippers. Lets say we cant exposure r-g-b pixels separately isnt it? Rai wrote they doing wb in two steps: first in-chip, second while 10->8bit conversion. mean its more range of 12 bit adc converts into better wb capabilities. right? and, sensors, i tryed to contact sensor manufactures to get eval kits, ie sensor + board + drivers + docs. At least FF and Altasens offer- still no reply. How do you "testing" this or that sensor, how to get em? Thought i'm a private developer and would not buy more than 1 unit anyway. May be there is some global sensor supersore :)), like bhphoto for video gear? By the way, i'am not an electronics specialist, more a C programmer with video as parttime biz/hobby. So I like idea of Rai that we can take sensor board, put it into linux mc then write to hdd. I also like to spent some freetime to learn how to make these things to work. But if sensor kits are so rear beasts (oem only?) it may be practical to go with camerahead. |
April 16th, 2005, 05:44 AM | #365 |
Inner Circle
Join Date: May 2003
Location: Australia
Posts: 2,762
|
I suggest you get (and everybody else) the documentary DVD "The Corporation". Watch through the film, then watch through the special features, then watch through the second disc interviews (around 7 hours worth all told). You will be progressively more shocked by each of these. This bias and collusion that keeps small people (low profit customers) out of large business is common. You are much more likely to get somewhere if you have big money, or established company that can order big. The market is getting more and more sown up for big players.
What people here do is get cameras/sensor boards from various camera companies to test the sensors. Steve is from Silicon Image, there is Sumix (one of the cheapest), and a bunch of others. Be careful with Ibis5a, there is various ways to implement it. The SI version appears much better than the Sumix version that Ben had last year, and Rai's implementation is better again. You might find this with other sensors. Everybody is on the Micron, Ibis, Altasens bandwagon (no 3chip yet) so it might be good to find an new alternative sensor that is cheaper/better rather follow what other's are doing. Wayne. |
April 19th, 2005, 09:40 AM | #366 |
Inner Circle
Join Date: May 2003
Location: Australia
Posts: 2,762
|
Here is some interesting tech:
This is the 300GB optical holographic disc due next year: http://www.extremetech.com/article2/...129TX1K0000532 The future is here, but only at 20MB/s so far. At least it is cheap for backup storage or RAW footage. Notice that it uses a camera chip from FillFactory as it's sensor ;) This is the competitor: http://www.extremetech.com/article2/...099TX1K0100486 New HD Mac's: http://www.betanews.com/article/New_...ase/1113497814 |
April 20th, 2005, 04:54 AM | #367 |
Inner Circle
Join Date: May 2003
Location: Australia
Posts: 2,762
|
Nab reports: The New Apple stuff:
http://www.tomshardware.com/business...418/index.html The new Panasonic (note: they are talking about 1440*720p mode, I think I have heard of such a thing, but not on this camera?): http://www.tomshardware.com/business...191/index.html |
April 20th, 2005, 08:53 AM | #368 | |
Silicon Imaging, Inc.
Join Date: May 2004
Location: Troy, NY USA
Posts: 325
|
Konstantin,
The min readout time is: (horiz size + horiz blanking) * (vert size + vert blanking) * pixel clock On many sensors, you can increase the vertical blanking time to keep the readout time fast but the frame time slower (what you want for rolling shutter artifact reduction). On others, you just drop every other frame. On single chip cameras, some have separate analog gains for each of the 4 (R,G1, B, G2) color channels. These are only rough gains to keep the dynamic range. You will still need to do a color matrix to get accurate white balance after capture. Sensor companies do not want to support single unit sales. It costs the same to support a single unit sale as an OEM design-in that yields high volume. That is why we will sell board sets to integrators - the volumes can be low but that is OK. Some companies will only do business in the 10K units per year volumes. Our volumes aren't that high but we do have some large commitments to the sensor manufacturers. Quote:
__________________
Silicon Imaging, Inc. We see the Light! http://www.siliconimaging.com |
|
April 20th, 2005, 08:53 AM | #369 |
Major Player
Join Date: Jul 2004
Posts: 587
|
The Tom's Hardware info on the Panasonic is very inaccurate. I like Tom's Hardware but I wouldn't trust anything in that specific article.
|
April 22nd, 2005, 05:27 AM | #370 | |
Inner Circle
Join Date: May 2003
Location: Australia
Posts: 2,762
|
Quote:
|
|
April 22nd, 2005, 05:33 AM | #371 |
Trustee
Join Date: Apr 2005
Posts: 1,269
|
Which software are you talking about?
|
April 22nd, 2005, 07:03 AM | #372 |
Inner Circle
Join Date: May 2003
Location: Australia
Posts: 2,762
|
Rob Scott's capture software for the SI cameras, he was the original one to volunteer to do software. He has been busy for a while. The link to his Wiki is in my first post in this thread.
Thanks Wayne. |
April 22nd, 2005, 01:14 PM | #373 |
New Boot
Join Date: Apr 2005
Posts: 12
|
its me Konstantin Serafimov, must be smth wrong with this forum, my old password is not valid anymore, and so does my real name and email :)
thank you Steve, very informative. So i havent got it, is SiliconImage a wholesaler or retailer? Should i consider this respectable company as a potential seller of camera for me? As far as I still have got no hardware, i'll continue theoretical talk. How image from cmos sensor could compared to ccd one. I remmember Rai used "video-ish" word (or alike) to describe image from ccd, even thought its a larde and quality sensor. Is it there that big difference? Wayne, i looked for alternative, even in russia, here near my country, but with no paticular result. It seem that FF and micron has simply best technology, (and altasens' webpage is similar to zero, beatifully colored zero, to hell with em). |
April 24th, 2005, 11:12 AM | #374 |
Regular Crew
Join Date: Dec 2003
Location: San Luis Obispo, CA
Posts: 38
|
Konstantin, regarding the wholesaler/retailer issue, that is a very good question. As I have been reading these forums for the last 6-8 months with a great deal of excitement I have however kind of had to temper it with the notion that these cameras may not be available to us on a single unit basis. For example are SI or Summix wholesalers that require you to be a dealer or require one to purchase multiple units? Wayne or Obin maybe you guys can shed some light on this issue. If not SI or Summix are any of the other chip makers disscussed or that either of you guys know about, willing to sell single units. If not then perhaps the best idea is to put together a group who are willing to buy (say 10-15 units total)....i think that someone was trying to do that a while ago if i am not mistaken. Anyhow please forgive the ignorance....i've just come across conflicting info and would like to know the answer once and for all.
thanks, Omar |
April 24th, 2005, 05:34 PM | #375 |
New Boot
Join Date: Apr 2005
Posts: 12
|
Not sumix, they are one of two companies that ever replied me about price and availability of their cameras. Second was elphel which is very, very helpfull if they only had 2/3 sensor cams. They claim their micron 1/2 gives beter picture than FF ibis5.
So sumix, there was additional questions to them. Especially, concerning global shutter implementationa and framerates/bitdebth. As far as they write at their SMX-150M webpage, 24bit 24fps even 1280*720 dont go via usb2. So its either lower resolution, framerate or bits. So i hope ill get clear these questions. concerning idea of group order, i doubt its possible. Lets say we have 10 ppl that agree to order 10 ideal camera heads, which is at least 2/3 1980*1080 100fps through cheap interface with all api etc etc at $1500. But It doesnt exist. So here start compromise, and everyone has its own priorities. For example, my dream solution would be to buy 10 Microns M413 (large hi-speed sensor used for rapid cameras with 4Gigs of ram onboard). So then hire somebody to develop sensor board with any common computer interface to put raw data directly to ram buffer, and make it an opensource project (or cooperate between owners) to quickly produce drivers, c-libraries etc. It was written here the micron costs around $900 (which qantity?). |
| ||||||
|
|