Archived from groups: rec.games.video.arcade.collecting (
More info?)
Well... Did some research...
There are 36 used sound addresses. Sounds 17 - 22 are the garbled voice
sounds. Unfortunately, things are more complicated than I originally
thought. Apparently, the choice of phonemes does appear random. I haven't
looked into the code to confirm, but I may. The difference between those 6
sounds is not necesarily the selection of phonemes (still not sure), but it
is the pitch at which they are played. The pitch is controlled by U19, a
1408 DAC that sits on the data lines (through a couple of 74LS75s) and
outputs the clock to the SC01 (via an LM741 and a couple of transistors).
So, the selection of those sound addresses appears to play a random set of
phonemes at a pitch controlled by the specific address.
This complicates the circuit I had in mind 2-fold. First, to make a
reasonable showing of it I would have to sample multiple garbled sounds at
each of the 6 pitches. This is basically what is done in MAME.
Unfortunately, this would significanly increast the sample storage
requirements and would still just be an approximation of what the real board
could do. Second, as opposed to simply translating the address select lines
to indicate the memory location of the appropriate sample, I would have to
implement some algorythm (random?) to choose from amongst the samples for
the given pitch. This would clearly indicate that the circuit would require
at least a small CPU (PIC) to perform that function.
Well, this probably puts this project on the back burner, at least for now.
I'll look into how much memory an appropriate set of samples (more than
those in mame) would consume and then determine if it could be done in a
cost effective manner.
Thanks, all, for the input. I will post once I've researched this further.
-roy-
"Chumblespuzz" <nospam@cademo.com> wrote in message
news:vrj4e.16441$8P2.4996@fe10.lga...
>I think you are slightly mistaken in your statement, but I will have to do
>a bit of research to be sure.
>
> You and Marcel are both correct, that trying to sample all the phonemes
> and then reconstruct them into the Qbert garble would be difficult.
> However, I believe the "randomness" of the Qbert sounds is not that it is
> random at run-time, but that the selection of phonemes was random at
> design-time.
>
> According to the documentation, Qbert has 36 sounds. These sounds are
> selected on the soundboard via the address lines coming from the 74LS377
> on the main board. Note I used the term selected. I do not believe there
> is any randomness involved, except perhaps the randomness of which of the
> 36 predefined sounds is selected.
>
> So, my plan was to sample the 36 sounds produced by the soundboard and
> play the appropriate sampled sound based on the sound select address
> lines.
>
> However, I have not done the research into this yet. The purpose of this
> post was to determine if I should do the research. Now I'm curious for
> sure, so when I have the time I will setup a test rig and cycle through
> all of the Qbert sounds. I will post the results when complete.
>
> -roy-
>
>
> <rgvac_junk@mgcap.com> wrote in message
> news:1112624848.141885.33690@z14g2000cwz.googlegroups.com...
>> Unfortuantely, sampling the Votrax SC-01a doesn't work as well as one
>> might hope. The samples done for MAME are only a small approximation of
>> all the different sounds the chip can make. The Q*Bert code relies on
>> the random voice playback which, if sampled, would take up a large
>> amount of space, and even then might not be enough...
>>
>>
>> tm
>>
>
>