Here's a shorter answer, later I will post a schematic.
A better title could say: " Cascaded COB Chimes make for a novel , unexpected time delay filter".
When this set-up worked, hooking an audio output to the DC power input of a second 'chime' circuit, I started speculating, how(?) was that possible. I've sent audio thru a fixed 2 khz buzzer, successfully, so started thinking...
One C.O.B. is marked "PT-313", has a surface mount transistor, probably speaker driver, and is used in a both forward power mode (toy puppy barks through the little speaker) but also passively tolerates a reversed supply (that be -3V). That way, this puppy toy can have two buttons on a wired hand control. First push button causes the toy to walk forward, while second push button causes the puppy to wag tail, barking (that be conventional +3V).
Apparently, the arrangement saves wire costs, but the C.O.B. is not harmed, or diode protected (?). So first I had been trying some audio outputs for direct connection to an LED, no resistor (risky). That worked, and I measured about 700 mV AC, relative to ground. Being an 'H' output, this AC voltage went up, to 1.3 V when both speaker output wires were measured, against each other (both outputs switch, in an 'H' driver.) Ok, then somehow I tried to see what would happen if I tried to power that barking puppy C.O.B. and it served to pass through the audio! That is, a cascade of a bird chirp audio circuit to 'power' a next puppy bark circuit actually works: You hear chirping even though expecting barking sounds.
Theory was, the processor had not yet begun running enough for barking, but the output transistor served to respond to the (unconventional) application of AC to the DC supply. Kind of a supply ripple.
Of course this could be done in software timing, a conventional way to go, but I'm doing salvaged parts exploration, and this has no added components, to get the little 'dog' toy to chirp like a bird.
The cascades of two circuits does not work when single sided audio output is tried: remember, that was 700 mV AC. So I pondered how to best cascade the C.O.B. pair, as grounds cannot be common.
But also I realized that cascaded pair has a time-delay filter aspect. Suppose your circuit will send out significant audio level, only after 300 milliseconds. Any short lasting input will be blocked, while an AC waveform with positive half of longer duration can make it through. But I STILL expected barks, not short chirps.
I hoped to later ask eeVblog: What size capacitor would one of those epoxy covered ICs include, under there? Does the processor execute code, yet, or stay in power up reset?
So I have a couple directions here:. One way is study of useful salvage, no changes. The other is with my own written code / compiler/ burner to experiment with the whole set of possible cascaded circuits.
More tomorrow...