Quad / 4 channel LED bargraph meter (audio-log reading)
Since I mentioned this circuit on the synth DIY list there has been a few requests. Unfortuantely I can't find the one I was looking for exactly so this will have to suffice. I did build an 8 channel version. That was 4 bar-graphs each with peak/hold. A very early version was a 2 channel with peak/hold. However this one is a straight up 4 channel with only average meter. However if nothing else it serves to demonstrate the principals involved.

Essentially speaking we are multiplexing the display through a National Semiconductor LM3916 VU led meter display chip. These chips are kinda expensive for what they are and so it's desirable to be able to multiplex them if a large number of bargraphs are required. The actual maximum number of displays possible with this chip is unknown. It's actually frequency response and propagation times are unknown. Obviously NSC had no intention of them being used this way. However they do make slight mention of using them as a peak/avarage meter. Which is what started all this.

This circuit was originally designed for a stereo enhancer based on an idea that came to me from the original alesis micro enhancer. The Enhancer it self wasn't entirely a success. I used the NE571 compander chips and these little guys are a bit noisy. Other than that it worked fine. Of the 4 input channels, only two have a rectifier stage. And a cheap one at that. This is because the other two channels were already reading DC control voltages. None the less these could be replaced with a rectifier stage as shown or even a better precision rectifier. In this case it's just a diode and a capacitor. Done for cheapness but a quick look inside many commercial designs will show that this is generally all that is necessary. In a peak/hold meter there would be two different rectifier stages. One would have a longer time constant. I'll get back to that.

The 4 inputs are multiplexed through a 4052 CMOS bilateral/analogue multiplexer. This is a dual 4 in 1 out device. (4 X 1 X 2) but I assume most people getting this far would already know that. There is a simple CMOS oscillator which is set in this case, roughly to 400 HZ. The speed isn't important as long as it's fast enough to avoid flicker and not to fast as to hit the maximum speed the LM3916 can cope with. Whatever that may be? The rule of thumb here is. "If it looks good, do it." The multiplexer selects one of the 4 inputs and presents this to the input of the LM3916. The LM3916 then does what it does best and makes some LEDS light up. However the other half of the 4052 switches a bank of transistors and thus a bank of 10 LEDS. Only the Bank corresponding to the selected input may be selected at any given time. Thus when input one is selected at the input, LED bank one is selected at the output. When the clock ticks over again, the next bank and input is selected and so forth. Do this fast enough and it appears as though there are 4 discrete displays. Simple really.

The LM3916 has a pin that selects between Bar and dot mode. This particular circuit is set to BAR mode giving the traditional line-o-LEDs. However in dot mode it's as one would expect, instead of a bar, A dot pegs out the level. IF one were to switch between these two modes very fast, and at the same time select between two different rectifier stages, one long and one short time constant, then you can produce two displays in one. A Bar showing the average instantaneous level and a dot with moves slower but shows the peak level. The Bar will appear to push the dot up. If the level remains low then the dot will gradually decay away. Just like a real meter should.

With this particular circuit you could modify it so that there is only 2 bars of LEDs. However there would still be 4 inputs. Simply selecting between two rectifier stages per channel. You would need to include a little extra logic to switch between Bar and Dot modes but the principals are the same. But remember, the more you multiplex though the chip, the slower it will need to be within some theoretical maximum. And we don't know what that is. I said in the beginning, I've done up to 8 and I'm sure 10 or 12 could be accomplished without too much trouble. You might even be able to get away with more however you may have to slow the clock down. The rule of thumb I use is that if you can't scan through the whole display 25 times a second then you're in trouble That's the number of frames that PAL TV uses per second. In this case however we're talking more like 100 times a second. Which in theory at least suggest that 16 bars of LEDS could be achieved without too much trouble. Or of course 8 bars with peak/hold each.

Oh I should make one final mention of powersupply issues. I found in the enhancer at least, that the power supply had to be decoupled through a pair of large, 10 ohm resistors and a huge cap. I think the resistors were 5 watt or so. This is because LEDs and their associated driver circuitry tend to dump dirty currents through ground and this propagates back though the power supply into the audio. This is true of digital systems particularly. When you start multiplexing LED with-in the audio range, IE 400Hz, this break through can become huge. It is a fairly standard technique where digital and analogue must co-exist, to run separate traces from the power supply to the different sections. Analogue and digital. Even then some times this doesn't help greatly and extreme measures as mentioned above need to be used. However this is largely dependant on the circuitry involved, the power supply scheme and how big the noise spikes are. As you may imagine, the spiked involved here are relatively huge. A Star distribution Powersupply is recommended.

Well I hope I've covered everything here. I'm sure there'll be some questions on this one.

Be absolutely ICebox.

blah blah View or Download
Schematic in
GIF file format