Camera for in-car touchscreen video

dth122

I have an upcoming project and could use some advice on finding a suitable camera.

I need to take live video of the touchscreen on a car's dashboard, send it over SDI to display on a projector.

I've been experimenting with a Marshall CV350 (Marshall Electronics - CV350-10XB / 10X, Compact 10X Camera (Full-HD) with Precision Autofocus) which isn't bad, except I have problems getting the exposure set correctly so it doesn't blow out the light parts.  More troubling, it's grainy at low light levels and I have problems with the picture rolling or flickering at lower brightness levels, I assume it has something to do with the refresh rate of the touchscreen but adjusting the shutter speed doesn't seem to help.  Note that the camera is limited to 29.97 and 59.94fps.

Has anyone done this successfully before?  Or can anyone recommend a source for more information on the best gear and setup to make this work?  It's somewhat unique so I'm having trouble finding people to ask.

It's preferable that the camera have SDI out, but I can work around it as long as I can convert the output to SDI for a 100' run.
- Dave

12harry

Hi,  It's not clear to me if the car will be moving (hope not)... It is possible the car's screen is also about 60fps as this is considered "standard" in some parts.... any discrepancy results in a bad case of flicker. ( or moving-band, etc. ).
Some pro gear allows the shutter=speed to be adjusted, so this could help and might have SDI out. But you'll need built-in ND filters.  What I'm not understanding is why there are instances of low-light . . . normally any display is bright (for the driver) or dimmed with car-lights ON at night.
Is it possible the camera is monitoring the whole-screen to get an average exposure? This will vary with the amount of data being displayed...   Again a pro-camera will be able to overate in Manual, so there is no change (unless you want it ).

Finally, what has been done in the past?  In-car displays are nothing new.
Hope that's some help. . . . .   Cheers.

dth122

Car will not be moving... that would be tricky!

The refresh rate issue is a bit of a mystery because auto makers don't publish this on their spec sheets.  I assume it could be anything depending on the display itself.  That's why I'm looking for a camera that has some flexibility in this area, but I haven't found anything suitable yet.  Everything I've found is limited to the standard frame rates - 25, 30, 50, 60.  

The exposure isn't isn't with the screen itself, but the contrast between the screen display and the surrounding controls, which are usually very dark.  Even with the screen brightness at the lowest setting I have to bump the exposure up so that the controls around the screen don't disappear in a black blur.  Doing that causes blowouts on the screen even at the low setting (with headlights on.

Your question about what has been done in the past is exactly what I was hoping to get answered here.  Surely someone must have done this before.
- Dave

12harry

The camera is "almost" seeing what is there; - excepting that "designers" of car-dashboards like to have their input, which usually means it's non reflective, dull and so on. The screen-makers only sell to the car Mfr if their screen is daylight viewable ( eg in a car with the roof down ) - hence there is a wild difference in brightness levels.   But I'd expect there is some dimming when the lights are on...?
To that extent you can't fix the difference....
......although some ND filter, applied to the screen would mean the dash would appear brighter.  You'll have to go to Pro-Suppliers for a large sheet of this ND stuff - and apply it with care so as to minimise reflections.  You could alter the "colour" by using a coloured filter - My alarm clock-radio (display) is covered in deep RED so the blue LCD hardly shows, making it easier to read at night...of course it's trickier in daylight, but I can lift the filter as it's hinged.
I'm not clear "why" you want the dash-Scree to be equally exposed - other than it just looks wrong, otherwise.
However,this filter-suggestion may prevent "touch" operation - so you may have to handle that separately e.g. by interposing cutaways of most action areas. If these are done as a close-up then the exposure issue may be reduced with a ND filter on the camcorder itself.
In the past it was possible to get camera filters with a central zone of a different colour . . . not exactly what you want (I suspect)- but might be worth trying if the camera is rigidly mounted some distance from the filter-holder. ie suspend this ND filter material so the LCD screen is darkened, allowing the "touch" to remain good.
I fear a camera-rig is fast-approaching, so you have the necessary play-room.... very important should a re-shoot be needed.
That LCD refresh-rate may be your prime difficulty - and only by altering the shutter-speed can you remove some of this, if it isn't exactly matched. An alternative might be to use stills - if the LCD display only changes  slowly ( e.g when some feature is happening ), then stills at 1/4 second intervals will not lose anything.  If you apply "touch" action very slowly your audience might not notice enough to complain ((It's only a hand getting in the way of the view anyway)).... Pro camcorders are not usually very well-spec'd for stills - and some don't do stills as that's another person's job ( So I've been told!).... which is silly IMHO since stills can be useful. Some cameras will take 2, or 3 stills quickly at different exposures - so this might allow some cropping/inserting to balance the screen/dashboard.  However, that involves more work later-on.... whereas fixing a piece of filter is a 1-off faf and you can see it's right-enough at the time...
Combing both filter approaches with movie-stills should fix all of the above issues and if done within the "Story-telling" the results (technique) might be unnoticed by the Boss and the Audience, too.

Good luck.

dth122

The reason the rest of the dash needs to be exposed, besides it looking wrong, is that the demonstrator will be pressing buttons on the dashboard that affect what's on the screen and we want the viewer to see those interactions as well as the touchscreen presses.

I might not have been clear in my original description, but this is for live demonstrations.  There is no option to re-shoot.  The rig will be used over again many times, but only for live events.  There's no recording planned.

This is also going to make the still approach challenging unless part of the demo is pre-recorded, which can be done but will be hard to intersperse with the live content.

In any case, your idea of using ND filters is a good one.  I'm going to do some experimentation with these and see what I can get done.

Thanks.
- Dave

12harry

With that extra-info, I wonder that you won't need some additional lighting so the buttons (etc.) show up. In particular, you could add brightness to the one that needs operation, then the audience can see the correct operation. The operator's hand will need to be in a black glove to avoid upsetting the exposure. You may need to build a custom light-support panel so the lights are close to the dash. but just far enough away to allow clear access, minimise reflections etc...
Trying to interleave live and recorded will need discipline and good timing - If there is someone else to operate the pre-recorded part - that will make it much easier,  - esp. if the "Routine" is always the same.

If this works, you might move to a"Green-Screen" where the operator's glove is a striking green: so it is ignored in the Edit - as though the hand doesn't exist ( or is a partial "ghost" ), that would allow the audience to see what's happening without the fingers destroying the view. The underlying image is the recorded but without hands, then the hand operations are recorded too . . . somewhat complicated at this stage... maybe Next Year?

FWIW.
I suspect you'll have enough work just getting the basic rig-setting correct . . . . but once that's fixed, the rest will be easy!

Cheers.

dth122

I never thought about cutting in pre-recorded segments or running green screen on the fly.  I suspect timing will become very important in those scenarios, but it should be achievable.

Thanks for the additional ideas.  I have a lot to think about and test.
- Dave