MSGEQ7 seven band audio filter device

You may need more capacitive decoupling across your power rails - as close as possible to the power pins on the chip.

Thanks for that. I do have good bypassing right at the power pins.

I found that I get output offset almost as low as with the internal osc, if I connect the external clock to the chip using a series 200K resistor! That is great, and I can prevent clock edge coincidence with the strobe since I control the clock and the strobe with the microcontroller. Clock edge noise is still present on the output, when using an external clock, but I can sample the output at any point I want to avoid the clock edge noise. A good ground plane might reduce clock edge noise too. I theorize that whatever inside the chip discharges the external C draws a spike of current from the external clock source and causes a higher offset on the output. A series resistor limits that current slug, and viola!

Note to users of an external clock source: Use a high value series resistor. 200K works fine.

Edit: Further experimentation using a second chip led me to discover that the series resistor needs to be 10K maximum. Higher values result in sometimes doubling or tripling of the frequency of the external clock applied because whatever discharges the external cap can cause voltage fluctuations at the clock input, resulting in false clocking. So, a stiffer source is needed. This has a small impact on the output offset.

ClydeCrashKop said:

The last one I made, Graphic equalizer club lighting in an infinity mirror
This one is 3 feet tall & 30" wide. It uses 5 columns of 53 Neopixels echoed to infinity. The Graphic equalizer separates the audio into seven bands. This light show displays the volume of the lower five bands. Bass is violet on the left. Treble is red on the right. It is fast enough to keep up with "Wipeout" and not miss a beat. Sometimes you can see the tremolo in a singer's voice. I had to dim the Neopixels down to about 1/3 for this video because full brightness overwhelmed the camera. I made this just for fun and to liven up parties.

Click to expand...

I'd really like to see that with a diffuser sheet mounted against the rear mirror, just to see the effect.

ccurtis said:

I'd really like to see that with a diffuser sheet mounted against the rear mirror, just to see the effect.

Click to expand...

The diffuser greatly reduces the path length of the light. You'll likely only see two or three reflections deep with increasing size of a diffused circle of light after each pass through the diffuser.

Also, the inside of the light box will no longer appear dark since finding a diffusion neutral density gray material will be a challenge (and transmits even less light (shorter light path). He white diffuser material will reflect room light and the box will look a much lighter gray to white in the off-state?

A few new discoveries.

1. If external clocking is used, the clock frequency can be swept to shift the center frequencies of the 7 filters. If this is done, before strobing the multiplexer, the peak detectors will store the highest magnitude over the swept frequency range. This is nice, because the filters are fairly narrow, leaving considerable gaps between them that sweeping can resolve.

2. The device operates well at 3V supply voltage. The offset voltage is lower at 3V than at 5V. Also the gain of the device scales with the supply voltage, which is good for applications with varying supply voltage and the supply voltage is also used as the reference for an ADC. If you input an audio tone with a fixed amplitude, the output of the device will change linearly with the supply voltage. That came as a bit of a surprise to me (although a nice surprise), as I would assume there is fixed gain ahead of the peak detectors.

3. I've clocked the device as high as 300 KHz with no problem. It may go higher, but that's as high as I want, or need. Clock duty cycle is not important, but should avoid extremely low or high duty cycles because the 10K series resistor in combination with the internal capacitance gives limited rising and falling clock transition slopes (1 us or so) at the device clock input.

Cheaper "MSGEQ7" chips I bought on ebay from China, and tried today, do not work, all ten of them. Output is solid high on all channels. What's up with that??? Penny wise...

If there is just something different about them, would be nice to know how to get them to work -- assuming they work at all. There should be a charge-discharge waveform at the clock pin with eternal RC connected and only power applied, but nada.

Thanks for documenting your continuing saga.
The MSGEQ7 is a very interesting device notoriously short of information.
We rely on people like you to map uncharted waters.
Thanks a lot.

I appreciate that ST.

One other thing I discovered may be interesting to some. The peak detectors in the device hold their value even if the clock signal input is halted, for those who may wonder, as I did, if halting the clock would bleed the peak detector level. I didn't determine for how long, or measure it exactly but not a significant change in output after one second of halting the clock, determined by observing a PWM driven LED controlled by the device output and visually comparing to the case with clock not halted. As stated earlier, the clock must run in order to read the output multiplexer.

What halting the clock will do, however, is effectively disable the input so it has little effect on the existing peak level.