On 02/12/2010 08:47 AM, stef wrote: > Le jeudi 11 f?vrier 2010 19:50:25, vous avez ?crit : > >>> Hello, >>> >>> I am facing the same issues than you while adding support for the >>> HP3670. I eventually found that one of my decoding script was buggy and >>> hide most of AFE accesses. I have appended the fixed version. By decoding >>> usbsnoop again, you may find more AFE writes to be coded in backend. That >>> may fix the red tint issue you have. I wouldn't be surprised if only >>> write to red gain was showing in decoded log, leaving other channels >>> untouched. >>> >> Ok I have decoded another log file using this revision, you are correct >> there are many more writes to the fe showing up but they do not really >> mean anything more to me :(. I have attached the fe_writes.log file, >> which is just a straight grep of the fe writes with the new decoded log >> file. If you could look at it and tell me what you think is happening, >> the only reasonable things I can gather is that there are: >> >> * init style writes (0x00 -> 0x70, 0x01 -> 0x80, everything else -> 0x00) >> * possible offset settings (0x05-0x07 seem to be set to 0x80 and 0xff >> repeatedly) >> * offset (0x05-0x07) and gain (0x02-0x04) seem to be set arbitrarily at >> various times. >> >> I'm not sure if any of this is relevant but whenever I set the >> offset/gain values in genesys_devices they seem to make little (if any) >> difference to the scans. >> >> >>> You'll also have to look at the base gamma coefficient in the >>> Genesys_Sensor entry for your model. >>> >> I've had a look at this and I take it you mean this section: >> float red_gamma; >> float green_gamma; >> float blue_gamma; >> uint16_t *red_gamma_table; >> uint16_t *green_gamma_table; >> uint16_t *blue_gamma_table; >> >> I've tried changing the floats but there doesn't seem to be much >> difference, maybe I should wait until I fix the shading issue then tweak >> this? >> >> I assume the gamma_table entries are just initialisers for the tables. >> >> >>> When doing a calibrated scan, the REG01_DVDSET bit is set to activate >>> shading calibration. Clearing it may side step issues with shading >>> calibration, but defeat the purpose of calibration. >>> >> I was slightly worried that's what I had done here, i've restored it to >> the original values :P >> >> >>> When tuning shading calibration, you must >>> look at the black/white_shading.pnm and black/white_average pnm files >>> that are written when full debug is enabled. With proper offset and gain >>> values, you must have a full black picture (with some values slightly >>> above 0) and a full white one. If not, the computed shading calibration >>> coefficient will be bogus and you'll get black scans. >>> >>> Regards, >>> Ste >>> >> Ok i've also attached the black/white_shading/average files. If I >> understood you correctly the black_shading.pnm should be all white and >> not the pretty purple colour it is now and conversely the white should >> not look as it does. I have tried altering the offset and gain in >> gen_devices but this doesnt seem to do anything :(. Any other >> suggestions about how I move forward with the shading? I think as soon >> as the shading is sorted I might be there with this project and ready to >> commit to the dev version. >> >> Also I currently have OFFSET_CALIBRATION and DARK_CALIBRATION flags set >> for the scanner, how do I know if I should be using these (as opposed to >> COARSE_CALIB and DARK_WHITE_CALIB)? >> >> Sorry I have given you quite an essay to read there but at the minute I >> have a few assignments to deal with so i'm just trying to juggle >> everything but again thanks for all the help and time you have given me. >> One other thing I wanted to ask is about a new release? I've seen a lot >> of chat on the mailing list about this but I just wondered if you knew >> any more specifics? >> >> I only ask because I am taking this project on as part of my University >> course and the ideal end result for me is to have the drivers present in >> a SANE stable release,so if I could do this by the next release that >> would be perfect :) >> >> Thanks >> >> Chris >> >> > Hello, > > regarding the AFE writes, judging by the AD9826 data sheet the writes to > addresses 0x02-0x04 are R, G and B gains. The interesting thing is that red > gain value 0x32 is higher than the 2 others (0x1e). So a red tint seems > sensible. Personally, I would try to set them all to 0x32 (preferably) or 0x1e > and map the gain array to these addresses. The other writes at 0x05, 0x06 and > 0x07 are for offsets. For instance the sequence of AFE writes matches it, wee > see gains set to 0 and various offsets tried. Then gains are set half range > (0x80) and new offsets are tried. Looks like offset calibration for me. The 2 > remaining addresses are for AFE configuration. > > For shading calibration, GENESYS_FLAG_DARK_WHITE_CALIBRATION would > allow do > calibrate in one pass and generate 'black' data using dark areas. To my > opinion, GL842 scanners need really good black data to do proper shading > calibration, so I'd recommend using the GENESYS_FLAG_DARK_CALIBRATION flag > instead. Once gains are good, you'll have a better reference white scan. But > the white_shading.pnm shows there is a black line in the scanned area. > Scanners have often different calibration area (located under the roof, can be > seen by disassembling the scanner, or doing an uncalibrated scan with 0 y > offset in genesys_devices struct). While the packaging may slightly vary from > one model to another, this black line is always at the same distance from the > scanning area. See GENESYS_FLAG_SEARCH_START flag. It also 'tells' where the > white calibration is located. You'll have to get sure that the shading > calibration scan only a white area. > > One more model supported in the stable release would be a nice thing > indeed. > > Regards, > Stef > >
Hey Stef, Sorry I haven't been in touch in a while I have been busy. Ok i've looked at this and my understanding is that I have to do two things; sort the gain values out and the shading calibration. For the time being i am disabling the shading until i have the correct gain values for my scans, my understanding is that it will be easier to correct the shading once the gain is fixed. After trying some different values I find that altering both the gain and the gamma colours in the Genesys_sensor has an affect on the scanning image. My question is how do I find the correct values for them? When i take the fe values for the gain from a snoop they are way off, and i dont think i can counter this with the gamma values. Also do you know when a next release is likely to happen? Just need to know if I have to hurry up with this as i'd like to have it ready for the next one. Thanks Chris -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
