Again lm3915, this time slowly decaying peak value (30 leds....).

[fezder]

I hope audioguru doesn't spill nuts when I yet again ask of this IC....
So, I made 30 leds version, last led is for 100w/4ohms (not peak value though) and leds are laddered 3db apart.
Following circuit does what I'm looking for, and can be modifed for either slow-decay or as in normal on/off fashion what lm3915 does. First of these behaviours I'm more after. So that is why I can't, or can't find suitable way to use your peak detector circuit that you supplied earlier audioguru.
Idea behind this circuit I came up on is simpe; comparator changes that small change in led on/off for proper pulse, this then controls pulse strecther, that is finally fed to opamp (lm358 works well in this application....). Due the amount of parts needed for only one led, you can see why I ask this, maybe there is better way? Please no microcontrollers, or double-supply, I can buy quite pile of components with price of another PSU...
I left those 2 other stages of lm3915's out from circuit as they aren't issue.
Any ideas/questions?


There is possible issue of oscillation with comparator, and I could omit last opamp, but then I'd need bigger capacitor to keep led live and that in turn could lead bigger inrush current of capacitors...

 

[audioguru]

If the LM3915's are set for DOT mode with no peak detector then the first set of LEDs will show both peak and average levels as explained in the datasheet.
Will you make 30 of those pulse stretching circuits? Then most lower LEDs will be lighted all the time unless there is no high level input for more than about half a second.

The inputs of the LM393 comparator do not work above 1.5V less than the positive supply voltage so they might mess up in your circuit and an LM358 has the same inputs with the same problem. I would use a PNP transistor with a series base resistor (and a resistor across the first LED) and it drives the pulse-stretching RC that feeds the NPN LED driver.

 

[fezder]

Will you make 30 of those pulse stretching circuits? Then most lower LEDs will be lighted all the time unless there is no high level input for more than about half a second

Yes, I'm aware of this phenoma, but that's okay, I'm after that effect where leds shut off one by one fading.

The inputs of the LM393 comparator do not work above 1.5V less than the positive supply voltage so they might mess up in your circuit and an LM358 has the same inputs with the same problem. I would use a PNP transistor with a series base resistor (and a resistor across the first LED) and it drives the pulse-stretching RC that feeds the NPN LED driver.

Oh, didn't think of those, they did work O.K so didn't think of any problems. I could of course power them from main-power line to counter that possible issue with vcc-1.5v margin. MAin power is close 24v. Also I could switch lm358 for mcp6002 (might buy quad versions in the future), or tle2142.
Can you draw circuit consisting of that PNP transistor? Thankfully I got plenty of those, at least bc557 comes to mind.

Actually now that I thinked bit, this circuit I came up on from your description and it does make sense (for clarify I used 2nd led):

 

[fezder]

Oh yes audioguru, yet again you helped, that circuit works like charm!. I never used pnp transistors before as for some reason I found them complicated to use. Indeed, that paraller resistor seemed to be unnecessary. Small value resistor (~100 ohms) before capacitor could be added to take bit of burden off bc557 during inrush-current, as you did with your peak detector?
Here's the final circuit I used:

 

[audioguru]

The base of a transistor should never be left floating. The resistor that turns off the PNP transistor that is parallel with the LED can be 100k. It prevents the PNP transistor from being turned on by leakage current in the output of the LM3915 and negative voltage developed by the LED when it behaves like a solar diode if it has some light shining on it.

I think the NPN transistor has a high voltage gain which causes poor and fast fading. Add an emitter resistor to slow it down.

Without having an input peak detector on the LM3915 then all the outputs will drop at the same time when sounds end causing all LEDs to turn off at the same time, not one after the other.

Yes, the 100 ohm resistor prevents zapping the PNP transistor when it tries to feed a few Amps into the capacitor.