diy led meter bridge
- Thread starter mr.Tom
- Start date
Help Support GroupDIY Audio Forum:
pucho812
Well-known member
yes. talk to dorrough ;D
I suppose you want a schemo, not a how-to guide for drilling 256 holes and putting LED's in?mr.Tom said:
There used to be several chips for this application, but today, there's mainly one, the 3914/3915/3916, and sometimes, it's not the most convenient chip in terms of versatility. personally, I don't like the choice of only 10 LED's.
So today I use a string of comparators (LM339); these chips cost next to nothing and I can build my string of references to get any law I want.
Attachments
peter purpose
Well-known member
Joe has something rather tasty... http://www.jlmaudio.com/shop/index.php?_a=viewCat&catId=9
![Soldering Iron Kit, 120W LED Digital Advanced Solder Iron Soldering Gun kit, 110V Welding Tools, Smart Temperature Control [356℉-932℉], Extra 5pcs Tips, Auto Sleep, Temp Calibration, Orange](https://m.media-amazon.com/images/I/51sFKu9SdeL._SL500_.jpg)
gemini86
Well-known member
abbey road d enfer said:I suppose you want a schemo, not a how-to guide for drilling 256 holes and putting LED's in?mr.Tom said:
There used to be several chips for this application, but today, there's mainly one, the 3914/3915/3916, and sometimes, it's not the most convenient chip in terms of versatility. personally, I don't like the choice of only 10 LED's.
So today I use a string of comparators (LM339); these chips cost next to nothing and I can build my string of references to get any law I want.
I've noticed that mouser only stocks the NTE substitute for the lm391X, and I'm seeing almost every commercial product uses the lm339 anyway. Is there a calc out there for determining resistor ladder values?
ruckus328
Well-known member
gemini86 said:
Tis because it's about 10 - 15x's more efficient. The LM391X links it's acurracy to LED current, plus the LED's are arranged in parallel, so to have full accuracy (with mind you only 10 LED's) it requires in the realm of 12mA per LED, so total of 120mA with all lights on, nearly the entire current availability of a 500 module.
You can do the same thing with a 339 arrangement with under 10mA total current consumption, and up to possibly 15 LED's depending on the type used and forward voltage drop of each LED.
Not to mention, as Abbey pointed out, you're not limited to the (generally ****** & limited) predefined scale they give you. It means you can have a very large scale range and still put some nice resolution where you need it most.
> The LM391X links it's acurracy to LED current
How?
The ref divider sets LED current, but we commonly use no divider (it is easier to divide the audio than to multiply Vref). 1mA per LED is quite practical.
We can also strap the LEDs in series, so total current is one-LED current, not 10X LED current. Page 18 in http://www.national.com/ds/LM/LM3915.pdf
The drawbacks of the '391x series are:
essentially single-source, a source which already abandoned us LM3909 lovers
10 outputs (not 5, not 16....)
three scale-law choices, and we always want something else
> light advertising boards with over 700 leds
Uh... study how the LEDs are connected. This will seriously limit your driver design, or waste the next year re-wiring 700 LEDs.
How?
The ref divider sets LED current, but we commonly use no divider (it is easier to divide the audio than to multiply Vref). 1mA per LED is quite practical.
We can also strap the LEDs in series, so total current is one-LED current, not 10X LED current. Page 18 in http://www.national.com/ds/LM/LM3915.pdf
The drawbacks of the '391x series are:
essentially single-source, a source which already abandoned us LM3909 lovers
10 outputs (not 5, not 16....)
three scale-law choices, and we always want something else
> light advertising boards with over 700 leds
Uh... study how the LEDs are connected. This will seriously limit your driver design, or waste the next year re-wiring 700 LEDs.
I personally don't like use of LM339's (or other comparators) except very special custom scale needed.
LM3914 for lin scale; lm3915 for 3db/led scale; lm3916 for analog meter-like scale.
They have good precision, LED current is easy to program
(in this case, use external reference, because the brightness depends on load at voltage reference out pin).
LM3914 etc are available at many places, except mouser - ebay or farnell for example, $2...4 per chip.
Easy to get 20, 30 or 40 led's string, datasheet shows the connection of 2 cascaded drivers,
with same mind possible to cascade more - just connect in series internal resistors strings, connect reference voltage to top of resistor's string;
mind 20k resistor used for cascading in bar/dot mode.
This way, each 10 led's can have different brightness. RTFM or (datasheet in this case
As it was told by PRR, check how led's are connected at your boards.
If they're connected as matrix, possible to use multiplexers for channels rectifiers/led's and one lm3914 string...but whole thing becomes bulk.
In old models, Dor**h used one string of LM3914's for one string of led's.
They used op-amps and transistors in log converters; IMO THAT2252 is better way to get rectifier in db scale and peak/avg/rms responce.
The rest is easy to figure out from datasheets.
Another way to get precision meter with low parts count, is to do it with shift register, comparator, integrator, charge/discharge circuit, and good temp stable clock.
No resistors string, no comparators, but has some voodoo.
Possible to regulate whole meter's brightness (not in a big range) too.
I saw somewhere schemo of 64 led meter, with good stability and very low parts count. Same idea as in A/D converter based on counter/integrator.
Interestingly, if integrator's cap charged thru resistor, scale is linear. What happens if we charge it from constant current source?
(edit: found!!!!!!!)
http://www.eetimes.com/electronics-news/4164691/Simple-high-resolution-bar-graph-display-architecture-uses-analog-comparator
Elegant schem, aye!
Another idea is use 28 or 40 led matrix and microcontroller with internal DSP. Ho! So far so good This way the parts count is very low,
the level of fun is maximum, due to user accessable options and features like peak hold, BUT requires some hardcore with R&D.
LM3914 for lin scale; lm3915 for 3db/led scale; lm3916 for analog meter-like scale.
They have good precision, LED current is easy to program
(in this case, use external reference, because the brightness depends on load at voltage reference out pin).
LM3914 etc are available at many places, except mouser - ebay or farnell for example, $2...4 per chip.
Easy to get 20, 30 or 40 led's string, datasheet shows the connection of 2 cascaded drivers,
with same mind possible to cascade more - just connect in series internal resistors strings, connect reference voltage to top of resistor's string;
mind 20k resistor used for cascading in bar/dot mode.
This way, each 10 led's can have different brightness. RTFM or (datasheet in this case
As it was told by PRR, check how led's are connected at your boards.
If they're connected as matrix, possible to use multiplexers for channels rectifiers/led's and one lm3914 string...but whole thing becomes bulk.
In old models, Dor**h used one string of LM3914's for one string of led's.
They used op-amps and transistors in log converters; IMO THAT2252 is better way to get rectifier in db scale and peak/avg/rms responce.
The rest is easy to figure out from datasheets.
Another way to get precision meter with low parts count, is to do it with shift register, comparator, integrator, charge/discharge circuit, and good temp stable clock.
No resistors string, no comparators, but has some voodoo.
Possible to regulate whole meter's brightness (not in a big range) too.
I saw somewhere schemo of 64 led meter, with good stability and very low parts count. Same idea as in A/D converter based on counter/integrator.
Interestingly, if integrator's cap charged thru resistor, scale is linear. What happens if we charge it from constant current source?
(edit: found!!!!!!!)
http://www.eetimes.com/electronics-news/4164691/Simple-high-resolution-bar-graph-display-architecture-uses-analog-comparator
Elegant schem, aye!
Another idea is use 28 or 40 led matrix and microcontroller with internal DSP. Ho! So far so good
This way the parts count is very low, the level of fun is maximum, due to user accessable options and features like peak hold, BUT requires some hardcore with R&D.
PRR said:
PRR... Thanks...Cool! I didn't know it could do that.
But it looks like it does that by running the chip in the "Dot" mode, and sinking the current for that last LED in the chain that way. Do you happen to know if I can cascade them using this wiring as well?
I want to get 14 LED's running in Bar mode without the big current surges of running the LED's in parallel.
Using 10 green, 3 yellow and 1 red I need 31.7V in series and the chip is rated to 25 volts. Is there some trick to doing this?
I really don't like Dot mode (I wonder if anyone does).
I have been messing around with other alternatives . I did manage a log display with adjustable steps using 14 LED's that will run on 32Volts and use a constant 20mA. It uses discrete PNP's and a full wave rectifier and log amp from a TL074 which works pretty well (it got pretty complicated pretty fast, I had to add a thermistor to temperature compensate the log amp).
The LM3915 is a much better approach but I want the 14 dot bar. Anybody got that working?
The LM339 approach looks pretty good(is there a reason that the LED's are cascaded in groups of 4 (is this related to how much the lm339 can source/sink?) If I replaced the R77, R78, R79 resistors with a single current source and ran it off 32 volts could I do 14 LED's?
Andy Peters
Well-known member
mr.Tom said:
I'd do it with a cheap microcontroller with a built-in ADC and either enough port pins to drive the number of LEDs you want, or use an external shift register and the micro's SPI port (or bit-bang).
-a
Yes.bruce0 said:
The reason is the supply voltage limit of the 339. You could do groups of 8, or 6, or whatever number that does not exceed the maximum supply voltage of the 339 or your availbale supply voltage. [/quote]If I replaced the R77, R78, R79 resistors with a single current source and ran it off 32 volts could I do 14 LED's?
[/quote] I think that would be too close to the limit. Most current sources need some voltage across them, so you would probably need 35V. Why don't you do that with two current sources? You would need only 24V supply so the 339 would be safe.
