Driving PrimeLed bargraph meters?

[DrFrankencopter]

I'm looking at using some of primeled's 26 segment LED bargraphs to serve as meters for a variety of projects. But, I don't really understand how to drive them.

Here's the details Primeled provides on their site: http://www.primeled.com/bargraphdetails.htm

Lets say I want this to work like a standard bargraph, and want the lights to cumulatively light up from the bottom up as level increases. So, I understand that I can hold pin 8 low, and send voltage(current) on pin 14 and get LED 26 to fire red, and then put voltage on pin 15 and get LED 25 to fire, but lets say I want to turn on LED 23 + 24 + 25 + 26. It seems to me that as soon as I hold pin 7 low that LED 22 will come on as well since I'm driving pin 15 to get LED 25 to turn on. Sounds confusing, but thats how I interpret the drawing.

Can anyone check this out, and see if maybe theres a better way to drive these LED arrays....surely there must be because primeled sells fancy driver circuits. I hope that its not some pulsed thing, where thechnically only one LED is turned on at a time.

Thanks,

Kris

 

[CJ]

Us help you out?
Shouldn't it be the other way around, or am I missing something here?
Gettin cold up there yet?
Freezin down here. 68. :grin:

 

[Gus]

A simple drive circuit will have a problem with the current changing with the amount of leds driven. If one X axis is on and you want two or more Y axis leds to fire the current will divided up between the leds and unevenly

I would guess the red and green for one segment are driven off a flip flop and Q and Q not are selectively enabled for the three colors wanted.

I also guess it only fires one segment at a time (two leds red and or green) and increments along the matrix at a rate faster than the eye firing only the leds wanted.

Why do I think the above

Peak current is noted

Duty cycle is noted

match between the leds is given

The adjacent led brightness might be due to the turn on and turn off time of the LED drive circuits

You could build a driver for it but maybe their chip(s) driver would be better.

 

[DrFrankencopter]

Crap, so they're pulsed.....maybe I can program a PIC to drive them. I think that using SHARC DSP to do it is overkill. They're just LEDS after all.

They sell driver circuits for the 56 and 103 segment LED bargraphs only, not for the 26 and 10 segments. I ordered 10 of the 26 segment ones....guess I'd better come up with a solution.

Cj, the weather here...around 80 today, but leaves are starting to turn, so fall is definitely on its way in....

Cheers,

Kris

 

[TomWaterman]

Roger's right, I also tried to find this info from them a while back as I was interested in the 53-segment meters and some smaller 10 segments for the Forssell opto.

They seem pretty bad at replying to emails and the tech stuff they sent to me was incomplete. Can't quite remember what is in it, but I can try and scan it if I can find it......

The sharc may seem like overkill but these bad boys can display VU, PPM and other parameters like RMS simultaneously. They are very nice meters and they 'look' stunning in action.

Good luck with it anyway.
Tom

 

[PRR]

> how to drive them.

SHARC????????

It is just an LED array. They are everywhere. You never designed for one?

It is just multiplexed. Each column is live 1/6th of the time. Repeat fast so the eye does not notice a flicker. Drive with the rated continuous current: average current will be 1/6th of that, but 2 or 4 mA average is probably ample. (The "10kHz minimum rate" for peak rating seems odd.)

Use a 3915 "VU meter" chip to drive the pins on the side.

Use a 300Hz clock into a 4017 CMOS counter, strapped to count to 6, to feed the pins across the top. For simplicity, ignore the red LEDs, so you just need a 3-counter and 3 wires.

Build a 3-step switched-gain amplifier to put before the 3915. If you actually use 3915, the gains should be 0dB, +30dB, and +60dB. This is controlled by the 3-counter.

The counter runs through 3 states 100 times a second. In the first state, the lowest group of LEDs gets fed from the top, and the amp is gain of 60dB. Then the middle bunch is fed, and gain is 30dB. Then the top bunch, and gain is 0dB. In each state the 3915 feeds the left side pins according to its steps and the swiched-gain boost.

Adding Red is a minor frill. If you want two signals, one red one green on one row, just count to 6 and add a 2-input switch in front of the switch-gain amp. Getting only the top LED, or just LEDs above reference level, to light red is a little trickier.

Oh, wait. Those guys wired the LEDs inconveniently. In my way of thinking, we need a 3-level version of the 3915 and a 9-step switched-gain amp: just switch the top and left wires and change the drivers for the proper number of wires. Also the 3915 is now the wrong polarity, but we probably would not use 3915 anyway.

I say it can be done with a small number of CMOS chips costing a couple bucks.

It could also be done with the larger Basic Stamp or a PIC, except you won't find 60+dB A/D converters in those.

 

[DrFrankencopter]

Hi PRR, thanks for the insight!

I've never designed for a LED array betfore...my first exposure to LEDs for metering was the Forssell opto compressor, which uses a 3916 to drive a bank of 10 LEDs in bargraph mode.

Since I'm only trying to drive a 26 LED segment, I don't tneed that much dynamic range on my converter. I figure 26 steps will be well covered by any converter with more than 5 bits. Of course, that works for linear steps (which works for gain reduction metering), true dB levels will require higher dynamic range. Maybe I could use multiple A/D's and boost the low signals into one to get more effective range. I'll have to take a look at the PICs out there.

The CMOS approach sounds interesting too....I wish I knew more about digital hardware design.

Cheers,

KRis

 

[DrFrankencopter]

Just found a pic:

http://ww1.microchip.com/downloads/en/DeviceDoc/30275a.pdf

This thing has 33 I/O lines (enough to drive 2 26 segment meters), and more than enough 12 bit A/D lines. That should be good for 60 dB (well depending on the real performance of the A/D). Published cost is around $5.00. I'm not sure what the supporting components would be...PICs are totally new to me. Probably need a clock, and maybe some kind of memory (probably depends on the PIC), also will need some sort of programmer.

Cheers,

Kris

 

[Rochey]

Don't forget to put a buffer inbetween the PIC and the LED's, as the outputs of the PIC probably won't have enough current to drive an array.

I discussed a thread somewhere with the aim of driving relays, but the theory should be the same.... http://www.groupdiy.com/index.php?topic=2296&highlight=relay+driver
cheers

R

 

[Gus]

You can sometimes use the limited current drive of CMOS ICs to make the drive simpler. You might not need resistors for current limiting to the LEDs.

I was thinking simple shift reg drive but like PRR wrote the x,y setup of the displays is not for this style. I would guess they made the display the way it is to sell their drive circuits.

I looked around at the site and noted the built display part gets a digital word for the display to convert to the drive circuit. I did not see the word length

I am guesss this display strong point is that it can be remote from what ever is being sensed being a digital word

 

[DrFrankencopter]

I would have used a panasonic array, but they are not suitable for flush mounting. I like the primeled array since it can be made to just poke through a front panel with no edges or other kind of bezel required.

For driving the LEDs, how about the digital I/O's turn on a transistor for each pin. That way the I/O doesn't have to drive lots of current, just enough to turn on the transistor which will allow current to flow to the array.

I'd really like to get this to work out and be reasonably priced. The LED arrays (26 segment) are $12 each, so thats pretty reasonable. Its the DSP that's expensive, and I think not required. I'm pretty sure a PIC would be more than adequate for a meter driver (even with some multi-colour action).

Cheers,

Kris

 

[Gus]

maybe a 8 bit AD with a log converter circuit before the input to the AD. Then use a lookup table to map to the LED(s) you want to drive. 8 bits, 0 to 255 steps then compare and light the red or green or both in each segment. Simple and crude but it should work. This would be more of a peak read with no hold time on the top led.

Read the input, compare, display for many clock cycles start again, more of like a PLC program.

I would think the DSP is used for trends and fancy ways to fire the caculating to display peak, rms, VU, monostable hold times etc.

Darn you can't use two, 3 to 8 decoders to cut down the lines to the drive to use all of the leds but if you give up the 25 and 26 leds you can.(EDIT I think to 3 to 8s would work )

You could also use a analog switch for different inputs for many displays from one pic etc

My idea would be CMOS 3 to 8's for the X and Y You turn on the 12 then scan 15,1,14,2,13,3 then 4 and then 15,1,14...... etc.

CMOS for the built in current limiting I need to look at some of my old notes.

The pic or whatever would need only 6 lines this way to the display.

Just some fast ideas.

 

[DrFrankencopter]

This could have potential to be a great Lab project! I'm pretty sure that coding the ballistics functions would be fairly easy. I've got some experience coding digital filters. I could probably fake VU, and program in peak hold (and washout) functionality as well. It all seems like pretty simple stuff, just a few lines of C code, so what's that, like a page of assembler???

Now, how fast would the meter really have to sample at? Is 22khz bandwidth essential? Do most meters respond to 20kHz?

Hmmm, now I'm thinking....

K

 

[EZ81]

[quote author="Gus"]maybe a 8 bit AD with a log converter circuit before the input to the AD. [/quote]

Maybe AD636? I have a schematic for a AF voltmeter using this IC somewhere on my harddisk, can't find it, though... A bit $$$, but with a 3$-µC (like Atmel ATMega8, for which there is a free gcc available) and a few driver transistors this might be enough.

http://www.analog.com/en/prod/0%2C2877%2CAD636%2C00.html

 

[Gus]

Some of the pics have a 8 bit AD inside. Now the other 12 bit AD pic might have enought power to do a linear to log conversion and be good enought for a meter.

The reason I brought up 3 to 8s is the possablity of multplexing the inputs and drive outputs with an enable signal to the input and output chips and protecting the microcontroller from the outside world..

 

[TomWaterman]

Prices look about right, I had a quote for $375 for an analogue-in 53 segment job....

Looks like you got better tech info than me too. Wouldn't mind a copy of those PDFs if possible Roger?

Cheers Tom

 

[Carl_Huff]

Take a look at LEDTronix at http://www.ledtronics.com/ds/LBG280TX/

They are a little larger and easy to work with. You can buy a 90 degree version that sets the LEDs off at right angles from the end of your PCB. You can easily drive these from a National LMM3914 bar/graph display driver. I have a design somewhere in my archive that does 40 segments of rms and peak concurrently.

You could decode the Primes with a common addressable latch like the MC74HC595 and an inexpensive dsPIC from www.microchip.com. You'd have what ... 50 bucks invested per channel after the PCB cost???

 

[Carl_Huff]

Roger,

They can be flush mounted, but you end up 'wider' than the width of the LED's because of the plastic carrier that the LEDs are snapped into.

When I have used them I have always mounted them immediately behind the panel with a thin sheet of frosted polycarb plastic plastic between the LEDs and the panel. The plastic sheet makes for a more aesthetic presentation.