Thu Nov 22 18:59:33 CET 2007

the circuit again

because the input impedance is so low, the 10uF cap is really not
neglegible! in fact, this gives a 10ms time constant with a 1K
resistor. that's 100Hz, but the filter cuts of at 1kHz, so it's ok..

it's not ok for what i wanted to do, which is to use only a single
transistor to drive the speaker without a capacitor. this might
reverse polarize the cap: it's probably best to keep the switching
frequency high enough so this doesn't need to happen.. i wanted to
replace it with a ceramic one, but that needs at least 1uF.

wait a sec.. maybe it's just not possible to drive the cap to a
negative voltage? yup.. the + side of the cap will be at saturation

                  Rg 220K
               |  __     |
        Rs 1K  | |  \    |
    /--/\/\/\--o-| - \___o     o Vdd
    |           _| + /         |
   === Cs      | |__/           >|
    | 10uF     |                 |---/\/\/\---o SPK
    |          o Vbias          /|
    |                          |
|\  @
| ||@
|/  @

i wonder if it's possible to make the circuit such that the transistor
doesn't blow up if SPK is driven low for too long. check this:

somehow the DC path needs to be blocked or at least limited.

hmm.. i think there's really no better way than to use switching: that
produces the least amount of heat in the transistor. just be careful
to not drive it too long, and use minimal DC: start wave 'touching'
ground instead of symmetric around 2.5V.

i'm ordering some BC640, which are TO92(A), 1A dropin for A1015. i had
a BC516 PNP darlington on my list, but the BC639 is 1A. for the
switching loads i care about, i don't need high beta.