lm3915 with clamps

[irbirb]

hello,

i just wanted to get some comments an this thing i built.
2 audio inputs are each split into highpass, lowpass, and bandpass via state-variable-filters. these operate on a single supply.

each of the six resulting audio channels drive six lm3915's.
the filtered audio had some weird DC offset things going on. the levels would be too high, but when measuring with a scope, the levels gradually drop to where i think they should be, centred around zero volts. i'm not really sure why this is.

i decided to try and clamp the driver chip inputs with a diode. it didn't work with a 1n4148 but a some kind of germanium diodes i have are doing the job well. they end up working as a sort of peak/average which makes the LEDs brighter in general and a resistor in parallel allows for a level of adjustability... a higher value holds the peaks longer... at least i think that's what is going on.

 

[crutschow]

The reason your output is initially at a high voltage is because there's no DC path for the output cap, so its output rises to about 1/2 the supply voltage when you apply power. The scope probe provides the DC path which discharges the cap until its output settles at 0V.

 

[audioguru]

The LM391x doesn't need a clamp diode since its input can be plus or minus 35V without any protection device. Your clamp diode charges the 4.7uF coupling capacitor backwards which causes a momentary DC level shown on the LEDs. The datasheet shows some good peak-detector circuits.

The input impedance is high enough that a 1M resistor to ground can be used with a small input coupling capacitor value but you have a very low value of 12k which requires your huge 4.7uF cap that probably has dielectric absorption that also messes up the DC level.

You have pin 7 and pin 8 messed up. They are supposed to be an adjustable reference voltage like shown on every circuit in the datasheet.
The current from pin 7 to ground sets the regulated output current in the LEDs but you have the pin 7 current overloaded then you try to reduce the LED currents with a 5.6k resistor.

 

[irbirb]

The scope probe provides the DC path which discharges the cap until its output settles at 0V.

ah, yes. seems logical. thx.

The current from pin 7 to ground sets the regulated output current in the LEDs but you have the pin 7 current overloaded then you try to reduce the LED currents with a 5.6k resistor.

ah, yes i see. i wanted to program the current with pin 8 and use 7 to adjust the sensitivity for some reason 0.o

 

[audioguru]

If the pin 5 signal input of the LM3915 does not have a resistor to ground then the input bias current of its PNP input transistor will cause the input to float up to the supply voltage which will cause the input voltage to be very positive and the IC will turn on the highest LED or all LEDs when in the BAR mode. The input bias current is very low (typically 25nA) so it will charge your huge 4.7uf coupling capacitor very slowly.
The max value of the input resistor is 1M ohms. You used nothing and only 12k ohms.

 

[irbirb]

If the pin 5 signal input of the LM3915 does not have a resistor to ground then the input bias current of its PNP input transistor will cause the input to float up to the supply voltage which will cause the input voltage to be very positive and the IC will turn on the highest LED or all LEDs when in the BAR mode.

okay, i wouldn't have guessed that from the data-sheet, just reading "high-impedance input buffer" and not really thinking twice about it. the only input resistor reference i noticed had was in series and is meant to protect the inputs from higher than 35V over voltage.

The input bias current is very low (typically 25nA) so it will charge your huge 4.7uf coupling capacitor very slowly.
The max value of the input resistor is 1M ohms. You used nothing and only 12k ohms.

i made one in the past but it had passive filters and, perhaps as a result, has not shown similar issues.