Matching brightness across multiple 7-seg LED displays

I've actually made testing systems which bin such displays by human perception of hue and intensity. These involve optics, spectrophotometer system, some sense of your chosen "white point," and some CIE-based software to process the data and bin the parts.

If you want to put these into instrumentation and need to have multiple pieces "look the same" given the same driving currents, then you need to buy them already binned or bin them yourself or else calibrate the currents (and hope the hue variances are acceptable.)

That is, I think so if I understand your problem.

Jon

with

since

the

perception

some

the

pieces

hope the

Thanks for your reply Jon. So do you think the problem is variation between the display ICs themselves, or is it the transistors I'm using? I am thinking it is more likely to be the transistors, since each segment within a display has the same intensity, and they share the same transistor.

Steve

Well, what I was saying is that if the current is identical (and that means abstracting away from BJT or resistor issues -- I'm talking about identical current drive) then various LED displays will give different perceived hue and intensity. Even LEDs cut from the same wafer, I fear. Processes just aren't able to be controlled well enough -- or at least not affordably so.

You can drive two displays that have the same part number from the same manufacturer and even from the same manufacturing lot number and use the exact same currents (say, to within 0.1% for example) and get what looks to a user as kind of "funky" looking.

This is why manufacturers will sometimes offer a binning service. You might buy

1000 parts, let's say, but ask for them to be binned on hue and intensity (according to some spec.) Then you will get, let's say, four boxes: 130 parts, 540 parts, 220 parts, and 110 parts -- call them boxes A, B, C, and D. When you assemble instruments, you will select parts only from the same box. Kind of like that. Next time you order, you might get four more boxes, but now they are group A, D, E, and F -- no B or C types, this time. Etc.

Jon

Something crossed my mind that I forgot to mention... Binning on hue isn't so important for most red devices. Humans are lots less sensitive to wavelength changes near 700nm. The hue part is more important elsewhere.

Jon

so

wavelength

Thanks for your info Jon.

Steve

I think Don Lancaster suggested tweaking the on-times with software to even things out for Charlieplexing in one of his columns.

Good luck Chris

with

since

the

even

Thank you Chris - that's a great read.

It's a very good idea, but unfortunately in my case I've actually got 100 display boards to do in total (and more to come maybe) each with its own PIC. It could be done but I'd have to mod the code of each one by hand until it looked ok (there's no free code space, and no spare pins, to do it in a less painful way unfortunately.).

Steve

"Steve" wrote in news:1RN8d.30$ snipped-for-privacy@newsfe5-win.ntli.net:

I've this problem before and instead of findind what is a fault I simply insert a 1K trimpot set at minimum resistance in series with the driving transistors then tune all the displays to match the lowest brightness one . roma

Yes. Or, if one is able to consider the idea, PWM all the displays (they are being used three at a time in Steve's system, so I'm guessing they are already in a 1/3rd mux situation) and calibrate an adjustment factor for two of the digits. Something like: assume each digit at 100% of their respective time slots, select the dimmest (as you suggest), and then tweak down the mux time of the other two, leaving slight time gaps where no digit is driven at all.

Jon

On Wednesday 06 October 2004 01:15 am, Steve did deign to grace us with the following:

You say, "in some cases." Is it always the same digit that acts up? Is there any other correlation, like number of segments active? You could get that if you've got a common-anode display, for example, with only one resistor to +V, rather than the anode to +V and individual segment R's.

HTH! Rich

I can't say as I know or care what charlie plexing is. But if you are multiplexing the displays it might be that your program is doing some action other than running the displays that causes one or two to miss being run. Example: it lights the second one, goes and gets data, starts over with the first one....

----== Posted via Newsfeeds.Com - Unlimited-Uncensored-Secure Usenet News==----

The #1 Newsgroup Service in the World! >100,000 Newsgroups

---= East/West-Coast Server Farms - Total Privacy via Encryption =---

the

the

resistors

and

Hi Rich and thanks for your advice. With different boards, different digit displays appear brighter/duller, e.g. on one board it might be the leftmost is brighter, and on another board it might be the middle and the rightmoost that are brighter. It makes no difference how many segments are lit either. I guess that points either to differences in the 7-segment displays themselves or the drive transistors (2N3704). I'm going to swap some displays and transistors around to see if I can narrow it down further (my soldering is lousy, so I've held off this step for a while...!)

Steve

.

Hi Roma - that sounds a very practical solution. Space on the board is tight so I'll have a look for some small 1K trimpots.

Steve

.

are

already

the

time

time of

Hi Jon. Thanks for the idea. The software cal method would be good in that it won't add any more components. Only problem in my case is that I've totally run out of codespace, having already crammed in more than I bargained for. If I found a few bytes of ROM spare I could just about hard-code the cal factors, but that would mean each PIC would have different code, which is not great news. To do it a better way, such as holding the factors in internal EEPROM, would add more code that I really can't squeeze in as it stands. My fault for not allowing enough slack in the PIC!

Steve

being

Thanks for the idea. In fact, different digits have different intensities on different boards, so it can't really be the software loop in this case.

Steve

On Sunday 10 October 2004 01:14 am, Steve did deign to grace us with the following:

It wouldn't be the transistors, if you're reasonably sure that they're saturating, which they should.

So, I guess you're stuck either matching displays or designing some kind of individual brightness controls. ;-)

Have Fun! Rich

On Sunday 10 October 2004 01:23 am, Steve did deign to grace us with the following: ....

I didn't even know they'd ported Slack to the PIC! ;-)

Cheers! Rich

On Sunday 10 October 2004 01:16 am, Steve did deign to grace us with the following:

Like these?

Cheers! Rich

them -

an

transistors

IC,

could

one

digit

the

are

of

Yes, as you suspected it wasn't the transistors; swapping them over made no difference - the same digits were bright and dull. So yes, now need to decide on matching displays or using trim pots. Many thanks to all who've helped me out here.

Steve