understanding the EX1r custom white balance
 

I am noticing something strange

I am shooting interiors, and the correct white balance is very close to the 3500K preset

actually if I put the EX1r in full auto it will go to 3400K.....fair enough

now...if I instead try to do the custom preset (A or B), it goes to 2500K, which is way too cold and definitely wrong

how can this be? can anyone explain?

Describe your procedure for this step.

Exactly how are you doing it? What is the room/lighting setup? Where are you holding the card? What is the card made of? How are you setting the exposure before white balancing? How much of the frame are you filling with the card? Is the camera's white balance offset feature turned on?, etc.

How do you know the custom white balance is wrong? Are you checking the color with scopes or a properly calibrated professional HD monitor? Your perception of what looks right, or best, may be influenced by an incorrectly setup monitor or the ambient lighting where you're viewing the picture. I am reminded of pilots who don't believe their instruments and fly into the ground. :-) In my experience, it is more likely that it's the auto-white balance value that is wrong, unless you have not followed the proper techniques for setting a custom white balance.

Just to add to Doug's post. Custom Preset simply stores a value in the camera. There is a procedure in the manual (albeit poorly written) on page 48 entitled "Executing Auto White Balance". To set a value, you have to put a white card or piece of paper in the lighting, aim the camera at it with the switch in the preset you want to set and then you press the Wht Bal button. The camera then calculates a new value to store in that preset. If you use auto white balance on the same white card, I would expect it to read the same as when it calculates the preset.

I am actually not using any card, I am very casually pointing at a white wall, or even pointing at the light (not close-up)

my point is precisely that: regardless of the method I use (card or not card), I would expect the the custom white balance I put in A or B to give the same reading as in full auto mode, all conditions being equal, and it's not

it's just a judgment based on the playback.
it seems like the custom value is way off, whereas the auto balance looks more correct to my eyes, and anyway shouldn't the EX1r use the exact same mechanism to determine the white balance in both cases? the difference being that with full auto it can shift if the light changes obviously

I just came across this thread in the recent posts feed, I don't know if I can help, but I have a suggestion.

The reason you use a card is to match the light where the subject will be. If you point at the light, you will likely get a very blue-hued WB.

Try the process again using a card or even just a piece of printer paper. Make sure you fill the whole frame with the card/paper, and that the card/paper is under the same light as the subject. I'm probably going to miss your response, since I don't frequent the EX3 forum, but I think this should remove some of the variables from the issue here.

Excuse me, but it seems like your wasting our time by not disclosing this in your first post. If I had known that, I never would have posted. I don't really have the time to help someone who is "casually pointing at a white wall" and then wondering why he doesn't get good results. Please try to conduct some serious white balancing, then ask other people for help if you still have trouble.

And BTW, with auto-white balance, the camera is just making a guess as to what the color should be. Obviously it's going to be wrong more often that it is right. The camera can't tell whether your wall is pure white, bone white, teal, sky blue, parchment, or ivory. That's just one reason why auto-white balance should never be used under any circumstances.

Absolutely not.

It should be used if you want people to randomly change colors in the middle of your scene!

Sorry I couldn't resist.

-Garrett

I just had a good play around with the EX1's ATW.. That's a strange beast indeed.
White card lit with warm white fluro has ATW and manual agreeing at 3600K.
Change to very blue LED and manual WB gives 7000K and ATW will not budget from 3600K...at first.
If I go from my preset of 7000K ATW will hold that 7000K. If I came from the camera's preset of 3200K ATW still holds that 3200K. Changing ATW speed seemed to make no difference. Switching Off Shockless White finally got ATW to budge CT from whatever preset it started at but only if the seriously blue lit card didn't fill the frame, even when it decided to budge it never got anywhwere near 7000K.

I wish I could make some sense of the EX1's ATW, I've had a shoot where it might have been useful. The same shoot had a much lesser camera in ATW giving a better result. I had several goes at doing a manual WB of the EX1 off a white card and within a couple of feet the reading went from 2800K to 6500K and a few minutes later when the cloud cover outside the building changed it all changed. For the sake of an easier experience in post I gave up and left the EX1 in its 3200K preset.

I find your reply quite snappy Doug. I humbly want to point out I just used the procedure as suggested at page 48 of the manual, albeit it may no be the way pros do. I still do not understand why with all things being equal the White balance algorithm of the ex1r doesn't give the same reading whether taken automatically or "upon request" done pressing the WHT BAL button. That's the only thing I wonder.

That's interesting Bob. You seem to confirm some of my doubts although what I experience is different: my manual WB results seem to be constantly "colder" than the ATW. in your case it seems like ATW doesn't work properly whereas in my case it seems like the manual WB doesn't, or at least works in a way I don't understand

can someone just do this test, and point the camera to one specific area in a room, do a custom white balance, then switch to auto white balance (not using full auto mode, or else other parameters would be changed by the camera), and see if the values are close or the same?

Federico,
at the same time can I suggest you try repeating you test with the camera stable e.g. tripod or on a table, and point it a an evenly illuminated white card. A piece of A4 / foolscap paper would be good enough for this test. Light the card with a daylight source and then repeat your experiment.

I agree with you 2500K does seem very low, the only source I've had that low is candlelight. I'm just very suspicious based on my own tests that something else is going on here.

Good. I meant it to be snappy. :-)

I was being diplomatic... take it easy : ) especially you should treat better one of your customers he he
my question is still on BTW : )

Several things going on here I expect.

ATW: Generally the electronics looks at the image as a whole and will use the brightest object in the scene as white and try to balance the image so that the brightest parts look white. This is easily fooled by bright walls that may be coloured, windows with exterior light, small lamps casting pools of coloured light etc. The indicated Kelvin reading will be for the area of the scene that the camera has chosen to consider as "white"

Manual White: You determine the white object, so have more control. However with white cards exposure is important, never overexpose as this pushes the signal up into the knee and will give erratic results. For this reason a grey card which matches skin tone brightness is preferred. Also consider where the light falling on the card (or wall) is coming from. You want to ensure the card is illuminated by the light source(s) that you are balancing for. In addition any coloration of the paper or wall will create an offset in the white balance as it tries to eliminate that tint. White printer and copier paper is often bleached or has a very small amount of blue dye as this makes it appear brighter to the naked eye, this is also true of many white paints. Beware of this as it will upset your white balance.

These are two different systems that determine white in different ways. Because of this the end result can be different and to be honest I would be more surprised if they were the same.

thanks but are you then saying that the area the EX1r uses to determine the white balance is different whether you use the manual or Auto balance (kind of the same way DRSL use different methods to determine the exposure)? in what way using the manual white balance "you determine the object"?
I will try tonight but in my opinion if you point the camera to a card, in the exact same position, and ask the EX1r to take the reading first manually and then automatically, the reading should be the same.
why shouldn't it be? I always think that auto white balance helps if you have gradual changes in white (say shooting outdoors for long periods) and you want the camera to slowly adapt to those conditions, but in the case of my example you are comparing like for like.

once again I am not disputing at all that manual readings are the way to go and much more accurate, I am just trying to understand if my EX1r is faulty

One difference is that ATW considers the entire screen area so may choose a corner or anywhere in the screen for the white balance. Manual white only uses the center 15 to 20%.

thanks. that's very interesting. that could explain my problem

As Alister noted, ATW looks for the area that it can find that has the closest thing to bright, even colors and assumes that to be white, The camera then sets that as white which then affects all other colors in the frame. When manually setting the WB you are telling the camera what to assume is white (as mentioned on the EX cams it's the center of the frame). So by you taking the camera and pointing it in the same direction as when you took an ATW reading, and then trying to manually WB, you are actually not balancing the camera to white, You're forcing the camera to make whatever is in the center part of the frame equal levels of RGB.

The only way to tell if your camera is not functioning correctly is to use the correct procedure for manual WB.

A grey card is a very inexpensive instrument that should be an essential part of your kit, even for someone doing it as a hobby. It always amazes me when someone will spend $5K on a camera but won't spend $50 on a grey card to be able to properly white balance.

-Garrett

Grey vs. White? This thread from last year is worth reviewing. Check out the tests in the post on the 2nd page.
http://www.dvinfo.net/forum/sony-xdc...better-wb.html

Just for the record, I still stand by my posts in that other thread. My F800 doesn't lie.

I just re-did the test using a White A4 paper in a tungsten lit room.
The manual reading is 2500K
The auto reading is 3400K

The camera was in the exact position on a tripod and all parameters were the same

I suggest you try to white balance in daylight and with colored light.
Just to determine whether the costum WB does actually change.
If it does change, record your white balance process (make sure you tell us what you are doing) and post it. I have a feeling that would be a lot quicker way to find out what is going wrong. Is it an electronic failure or human error?

The part I've bolded might be a clue. If the card is only being lit by reflected light then it could be any CT. To conduct a meaningful test you need the light source directly lighting the card.

Under your current scenario ATW is probably getting the right answer and so is manual WB. The problem isn't with the answer, the problem is the question :)

Also, do it with no PP selected so you eliminate those menus entirely from the white balancing.

I hold an A4 paper opposite to the light source then point the camcorder so that all the image I am seeing is white then press the white balance button. will try tomorrow morning to do the test in daylight and post the results

I have the light source directly lighting the card. the reading doesn't really change much though. and always stays around 2500K (2400/2600)


sure...all these test are being done with no PP selected

Try to hold up a few pages of the white paper to ensure you are not getting any bleeding through from a light source behind the paper. Most paper is think enough where colors coming from behind the page start to show through. Also make sure your exposure is set correctly before taking your WB reading.

-Garrett

hi Walter
this morning I tried the daylight test
it's a cloudy day so not full sunshine

I used the A4 paper in different positions and every-time took the manual reading and then the automatic reading for comparison. I also took different measurements using the white wall (sorry Doug! : ) ) - all measurements were taken with no Picture Profile selected, and never using Full Auto

1) paper placed near the window so that it gets direct daylight: manual 9200K, automatic 9200K
2) white wall near the window: manual 6600K, automatic 6700K
3) wall inside the room (not directly lit): manual 5000K, automatic 5000K
4) looking outside of the window: manual 6500K, automatic 6800K

afterwards I went into a tungsten lit room and redid the test with the A4, and once again the results were quite different: manual: 2800K, automatic 3500K

so my point is that from my tests it seems like my EX1r shows very little variance when measuring white balance from natural light, and results are either identical, or vary in the order of 100K,

when measuring white in artificial light, I keep seeing a far wider difference, in the order of 600/1000K, and often I have the impression the manual reading is too cold

9600K from your paper and 6600K from the white wall suggests the paper is not white, but blue or contains dyes designed to make the paper look bright and white to the naked eye. 6600K is a much more sensible reading from an overcast sky.

I don't know why you are seeing such variations under tungsten light, but I'm still not surprised that there are differences. They are two different systems.

Whilst they are indeed different systems it's not too hard to setup a test that cancels out the differences. A reasonably evenly lit surface correctly exposed surface should yield the same results in auto and manual. Whilst the values maybe wrong they should at least be consistently wrong.

Still I have to agree (again) with Federico, 2500K is very low, well into candelight. Unless the lamps are on dimmers or the building has a wiring fault something is wrong here.

Hi Bob
that's precisely the point I am making: they values might be wrong, but should be consistently wrong

the value I get from auto (around 3400) seems much more correct to my eyes (at least)

my EX1r is still under warranty so I am not worried, but just trying to understand if there's a fault or not

The best thing you can do to establish whether the camera is faulty or not is to do a proper manual white balance and then look at the output on a waveform monitor and look at the RGB parade or vectorscope. On a vectorscope the white should just be a spot right in the center of the scope or on RGB parade the white R, G and B levels should be equal.

This eliminates the monitor or other variables and is more accurate than judging by eye

You could record a clips and look at it in your edit software.

If you lack the gear or experience to do what Alister is suggesting take the camera to a large Sony dealership in London and do a comparison between your camera and another one.
Maybe as others have speculated you're doing it wrong, hopefully someone in the shop will set your right.
Maybe your camera does have some strange fault, again a trip to the shop would resolve that.
Short of one of us having your camera in our hands all we can do is speculate.

I didn't read through this whole thread, but I thought I'd add this:

1) Don't forget to set the camera to Auto Iris before you hit the white balance button. This has tripped me up a few times.

2) Make sure that the light falling on your subject is the same EXACT light that's on your white card. Sometimes you have tungsten lights coming from fixtures in the ceiling, while you have daylight pouring in through windows - all happening in the same scene. This can make white balancing very tricky. (In this situation I've learned to either close the blinds to eliminate the daylight or set up some portable tungsten lights to over power the daylight coming in the windows)

Just my $0.02. Sorry if this has already been covered....

that's a good idea, but believe it or not it's quite tricky to find a shop in London that sells the EX1r
mine, I had to go well out of town to buy it and I already contacted the shop, who very unhelpfully responded they weren't sure, and want to charge me £125+VAT just to have it checked!!

any fellow DVInfo member in Central London can be around for a quick test?