I just got out of trouble, so this is not a call for help, but I thought I would write a small summary to help people who have problems making the MSGEQ7 work properly.
In a project of mine, I started with one MSGEQ7 (the breakout board version). Mounted the reference circuit and software, printed the values-- results made no sense. So I triple-checked my circuit, and my components, altered the signal timing several times according to many articles on various forums: nothing worked. I read all the available forums and lost a lot of time. Then I saw somewhere that someone got a "bad" MSGEQ7, so I bought more chips from various vendors (why not, it's so cheap, right?) But still, nothing seemed to work as it should. Before going completely bonkers, I finally got an oscilloscope and a signal generator to see what was going on.
I made some measurements, and here are my conclusions:
Purchased directly from China (Amazon, AliExpress), almost all the MSGEQ7 I bought had one or more defects.
A few examples of the defects I observed:
1. Inoperative channels which constantly returned a fixed value, regardless of the transmitted frequency (between 0 and max voltage) [Example: on one of the chips, channel 7 (16KHz) remains stuck to HIGH regardless of the input signal]. Sometimes, the chip is entirely dead: the channels won't bulge, regardless of the input signal frequency or amplitude.
2. Two or more channels were merged together (they return the same value when one of them detects its frequency) [Example: on one of the chips, channel 2 and 6 always had the same value, which varied according to the frequency detected by channel 2].
3. The frequency effectively detected by each channel didn't correspond to its specification by more or less 50% [Example: on one of the chips, channel 1 (63 Hz) returns its maximum value around 40 Hz, 6250 Hz channel peaking at ~3000Hz etc].
4. The signal amplitude at which a channel reaches its maximum value is not the same for all bands. [Example: on the same chip, 0.24Vpp maximizes certain channels while other channels only rise to 50%-75% of their maximum value for the same amplitude and it is necessary to increase the amplitude of the input signal so that the correct value (max ) is reached].
5. On some chips, the RESET pin value bleeds into the DATA_OUT pin. It's not that bad, because when the strobe is performed, the RESET pin is off, but still, it's strange.
6. On the same chip, the behavior observed on 3V is radically different from the behavior on 5V.
The place I tried that finally gave me chips working on spec (i.e.: you place them in the same circuit and they behave completely as expected) is Sparkfun. Maybe stores like Sparkfun can provide quality because they test the stuff they receive from China, and resell only the chips that can pass their tests? I don't know. But in the future, I'll buy directly from China only when quality isn't important at all, or when I'm confident I can repair the broken junk myself. Otherwise, the cheaper prices just aren't worth the lost time and money.