LM3915 clamps

Discussion in 'The Projects Forum' started by bribri, Feb 20, 2011.

  1. bribri

    Thread Starter Member

    Feb 20, 2011
    133
    5
    hi,

    so i made an array of 60 LEDs driven by 6 LM3915 (dot mode). they are driven by 2 audio channels which are each split into high-pass, band-pass, and low-pass via 2 state-variable-filters which are operating from a single supply. (everything is running of a +12V supply)

    the LED leads are fairly long so i've got a 10μF cap between pins 1 and 2 on each driver.

    i did observe the need for decoupling capacitors at the inputs, and presumably the outputs of the filters. but i've been plagued by some wacky DC offset kind of things at the inputs of the display drivers. when measuring with a scope, these levels gradually fall to about zero Volts, and then removing the probe, they climb back up again for some reason.

    i found elsewhere on these forums some recommendations for adding the peak-detection circuits which are mentioned in the datasheet. wanting to avoid extra work, i opted to try and clamp the audio signals with just a diode from pin 5 to ground.

    this didn't actually work with a 1n4148, but some germanium diodes i had kicking around worked really well. in fact they're performing some kind of peak-detection or averaging thing which makes the LEDs much brighter : )
    and a resistor in parallel with the diode adjusts how quickly the peak hold reacts. well i'm not completely sure what's going on there actually, but that's how i'd describe it. (i think i saw something similar on a video circuit schematic once)

    my low-pass channels were making their LED drivers really quite hot, but a resistor between the +12V and the LED anodes seems to be keeping that under control for the time being.
     
  2. bribri

    Thread Starter Member

    Feb 20, 2011
    133
    5
    here is schematic and build photos.

    comments?
     
  3. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    Like I said on the other website:
    1) The LM3915 survives inputs of plus and minus 35V so your clamp diode is not needed and just charges your huge 4.7uF coupling capacitor backwards for a moment which makes some DC offset. Good peak detector circuits are shown on the datasheet.
    2) The LM391x can use an input to ground resistor as high as 1M instead of your very low value 12k so the input coupling capacitor value can be small and will not have the input absorption problem (which causes a DC offset voltage) of your huge 4.7uF coupling capacitor.
    3) Your pin 7 and pin 8 resistors and wiring are messed up. See the other website and the datasheet.
    4) The LM391x has regulated output current for the LEDs but instead you are overloading pin 7 then using a 5.6k resistor.
     
  4. bribri

    Thread Starter Member

    Feb 20, 2011
    133
    5
    ah, thanks for having a look.

    oh, i thought 4.7μF electrolytics were pretty normal for audio with low frequenies. i just saw clamping as a way around having to set the low end of the voltage divider above zero. i also assumed that polarized caps would not charge backwards so much... but then i guess that would explain a few things.

    without the diode, the DC offset is 'high'. with the diode installed, the offset kind of bounces around zero. the lower the input to ground resistor, the less visible the bouncing, so 12k was chosen as a way of tuning this since keeping a bit of this bouncing effect has turned out to be desirable (at least aesthetically).

    yes, thanks for pointing that out. the 5.6k resistor was single solution before hard wiring the pin8 resistors to where they need to go (schematic was just copied from what was built).

    but yeah, i think i'll just try some smaller cap values, see what happens.
     
  5. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    I use 0.33uf film coupling capacitors that have no dielectreic absorption like electrolytic capacitors have. Their tolerance is 5%.
    Your 4.7uF capacitor has horrible tolerance so might be only 3.8uf. Into your very low value 12k input resistor its cutoff frequency is typically 2.8Hz. People hear as low as 20Hz, not 2.8Hz. My 0.33uf capacitor has a cutoff frequency of 1.8hz when it feeds a resistance of 173k ohms. You could use a 1M input resistor on the LM3915 and a 0.057uF coupling capacitor and still have a cutoff frequency of 2.8Hz.

    The input has nothing to do with the divider. The input ignors negative voltages.

    Whenever the input tried to go lower than 0.6V then the 4.7uf coupling capacitor was discharged a little (charged backwards). Then the LM3915 showed the changed voltage.

    Maybe your DC offset was caused by mains hum pickup if you built the circuit on a breadboard. Maybe the polarity of your 4.7uF coupling capacitor was backwards. Or maybe it was caused by dielectric absorption from your electrolytic coupling capacitor.

    With a 12V supply, pin 7 overloaded so the IC does not regulate the output current then the LED current with your 5.6k resistor is only 1.7mA for red LEDs or 1.4mA for blue or white LEDs. Pretty dim.

    Do you understand how pin 7 and pin 8 and their resistors should be connected? When pin 8 is connected to ground then the LM3915 is the most sensitive to signal level. The resistance from pin 7 to pin 8 or to ground causes the LED current to be 12 times more.

    Smaller cap values will reduce bass frequencies unless you re-calculate and increase the value of the resistance the capacitor feeds.
     
  6. bribri

    Thread Starter Member

    Feb 20, 2011
    133
    5
    yes 20Hz for hearing. . .
    but in defense, i have read that 10 times the desired cutoff is normal (and presumably a good idea with such wacky tolerances). plus, perceptibly visible frequencies go from about 50Hz and lower, and part of the whole deal of visualizing sound is to aid the perception of things which hearing can't readily get a hold of. i did fail to mention that, though my basic set up involves a fairly standard audio playback device, lower than 20Hz frequency response is actually a requirement.

    no i mean to adjust for the unintended DC offset.


    well definitely not the 1st two, so probably the latter.

    well it's not too bad due to the averaging effect of the cap-diode-resistor thing, but 5.6k was all i had laying around in a pinch. i figured a 1/4 resistor wouldn't deal with 6 drivers all that well.

    well yes and no. building from memory, being a bit dyslectic, whatever ; )
    my approach in the past was to just leave the current full on (dot mode)

    if the LED current issue is dealt with i'd like to just stick more or less with how the results actually look in terms of DC offsets which are a bit bouncy. i guess the alternative is to build in the peak-detectors and source some other caps, but in this particular case i'm keen to just stick with the lower part count and monitor stability.

    i appreciate the input. the absorption issue must surely be what i was looking for and i'm looking forward to checking out the strategies for mitigation, and maybe perhaps some exploitations of it.

    i've also been experimenting with the lm3915 in higher frequency applications... 240KHz or so would be nice... or higher. haven't quite really tested though. i like to assume that the analogue nature of the chip would be happy with a few MHz.
     
  7. bribri

    Thread Starter Member

    Feb 20, 2011
    133
    5
    forgive me for not knowing, but how would you go about calculating the cutoff frequency here?

    is it cutoff in Hz = 1/2πRC ?
     
  8. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    That is correct. 1/(2pi RC). If the source resistance is fairly high then it must be added to the R.
     
    bribri likes this.
  9. bribri

    Thread Starter Member

    Feb 20, 2011
    133
    5
    okay, this is good to know.

    but now, to lower the Q of the filter, i'm assuming that i need only change the 180K resistor, is that right?
     
  10. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    I have never seen that type of bandpass filter before.
     
  11. bribri

    Thread Starter Member

    Feb 20, 2011
    133
    5
  12. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    Yes I have seen a State Variable Filter before (I forgot about it) but I never made one.
    Reduce the value of R7 to reduce the Q but this filter is supposed to have a high Q.
    With a low Q then the slopes are very gradual which makes it useless as a bandpass filter.

    For a wide bandwidth low Q filter then use second-order or third-order Sallen and Key highpass and lowpass filters.
     
  13. bribri

    Thread Starter Member

    Feb 20, 2011
    133
    5
    i actually checked what a ceramic would do...
    way more absorption than the aluminum electrolytic.
    i just assumed that it was because of polarization.

    i guess coupling capacitors don't have to be polarized no?

    what were yo using, panasonic ECQE series maybe?
    multicap?
     
  14. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    I used Philips and Siemens film capacitors for years. Then they got together to make EPCOS film capacitors. They are excellent and are inexpensive.
     
Loading...