Spectrum Computing

I, Ball, has the introductory voice speech broken, because it was stored on the screen and then Firebird used their loader with a counter on the same area. I'll add a note to the title.

Some more:

- Sabrina
- City Connection (128kb version rediscovered recently, lots of speach)

Out of my own games:
https://spectrumcomputing.co.uk/entry/1 ... bie_Hunter
https://spectrumcomputing.co.uk/entry/3 ... abij_Ducha

Tristar has some speech, in Italian like "tutti i piloti sui caccia" (all pilots on the fighters).

Tombola Napoletana it's a neapolitan bingo with some speech and a sampled Jingle Bells intro.

Now Beeper and AY are separate groups:

Beeper Speech
AY Speech
Currah Microspeech Support
Fuller Orator Speech Support

Freedom Fighter has some speech on the presentation screen ("Freedom Fighter") and at game start ("Get ready", although as Chris Jenkins noted in his review on Sinclair User, it sounds more like "Geck Eddie" ).

I'd always assumed that speech uses the Beeper chip. I didn't know it could be done with AY. Is there any technical difference to how they sound?

Lee Bee wrote: ↑Sun Apr 16, 2023 4:57 pm I'd always assumed that speech uses the Beeper chip. I didn't know it could be done with AY. Is there any technical difference to how they sound?

Beeper = 1 bit
AY = 4 bit per channel digital playback (*)

So yes.

* not sure if volume levels are halved when using envelopes as they are on the SAM Coupé in which case it is 3

As promised I've been through existing entries (in latest ZXDB version) and all the games mentioned in this thread, to categorize them into Beeper / AY Speech. However for a few entries, I need a clue to where in the game it speaks

(328) Auf Wiedersehen Monty
(4426) Shadow Skimmer
(4541) Skateboard Kidz
(4767) Splat!
(4863) Starglider
(5092) SWIV
(12965) Mortal Kombat
(35410) Normal Life
(39293) City Connection

/Thomas

Einar Saukas wrote: ↑Sat Apr 15, 2023 6:32 pm Beeper Speech
AY Speech

This is fantastic! Though I notice not all the titles have been added yet, such as Ralf's games?

Stefan wrote: ↑Sun Apr 16, 2023 5:15 pm Beeper = 1 bit
AY = 4 bit per channel digital playback (*)

If AY speech can use the full 4 bits, that would be pretty decent quality (even cassettes and LPs are only around 6 bit). However, I'm not hearing this level of quality with AY, they still sound quite noisy to me. For example, the AY "180!" doesn't sound any clearer than the Beeper "Goal!". I don't hear the extra quality with AY. Is it possible that some AY speech is actually still 1-bit?

kolbeck wrote: ↑Sun Apr 16, 2023 8:49 pm As promised I've been through existing entries (in latest ZXDB version) and all the games mentioned in this thread, to categorize them into Beeper / AY Speech. However for a few entries, I need a clue to where in the game it speaks
(328) Auf Wiedersehen Monty

Every instance I know is on entering a new screen, and you'll hear "OW!", "MONEEEEEEY!" and similar.

kolbeck wrote: ↑Sun Apr 16, 2023 8:49 pm (4767) Splat!

Finish a stage. "YIPPEE!" It even prints that on screen.

Lee Bee wrote: ↑Sun Apr 16, 2023 9:22 pm If AY speech can use the full 4 bits, that would be pretty decent quality (even cassettes and LPs are only around 5 bit). However, I'm not hearing this level of quality with AY, they still sound quite noisy to me. For example, the AY "180!" doesn't sound any clearer than the Beeper "Goal!". I don't hear the extra quality with AY. Is it possible that some AY speech is actually still 1-bit?

That's because the sample in 180 is played through the beeper - it has been mislabelled in the database as AY speech.

For a good example of digital sound playback on the AY, have a listen to the title theme of "Byte Me" - the result is not dissimilar to Amiga MOD music.

For Splat!

TMD2003 wrote: ↑Sun Apr 16, 2023 9:27 pm
Finish a stage. "YIPPEE!" It even prints that on screen.

That seems to be the problem, I've never been good at that game

PROSM wrote: ↑Sun Apr 16, 2023 9:29 pm For a good example of digital sound playback on the AY, have a listen to the title theme of "Byte Me" - the result is not dissimilar to Amiga MOD music.

Oh that is impressive! Never heard that before! That needs adding to the list, Einar!

Lee Bee wrote: ↑Sun Apr 16, 2023 9:22 pm If AY speech can use the full 4 bits, that would be pretty decent quality (even cassettes and LPs are only around 6 bit). However, I'm not hearing this level of quality with AY, they still sound quite noisy to me. For example, the AY "180!" doesn't sound any clearer than the Beeper "Goal!". I don't hear the extra quality with AY. Is it possible that some AY speech is actually still 1-bit?

That it can, does not mean that it does.

The amount of memory required for 4 bit samples is four times that of a 1 bit sample, so by design the samples could just be 1 bit to fit in memory, and / or to be able to use the same sample for both AY and beeper.

Update: just read that it was in fact not AY speech, but my comment still applies.

Another important factor when playing samples is the playback speed (KHz) and that the playback speed is constant (beware of contended memory).

On my YouTube channel (see link in signature) there are quite a few oscilloscope videos of Amiga mods being played back through the SAM Coupe's soundchip (3 bits per channel), there are a few non-oscilloscope SAMdac (= printer thing DAC that allows 6 bits per channel) playback videos, the difference (3 bits, so 8 times the sample resolution) is huge.

kolbeck wrote: ↑Sun Apr 16, 2023 9:30 pm For Splat! That seems to be the problem, I've never been good at that game

My brother had a Game & Watch of a similar concept, so we both had plenty of practice. I'd never make it all the way through Splat! but the first three or four stages are within my range.

Stefan wrote: ↑Mon Apr 17, 2023 3:54 pm That it can, does not mean that it does.

The amount of memory required for 4 bit samples is four times that of a 1 bit sample, so by design the samples could just be 1 bit to fit in memory, and / or to be able to use the same sample for both AY and beeper.

Would there be any benefit to 1-bit samples on the AY. It seems like it would massively complicate any code and you wouldn't really benefit from the AY being asynchronous because the Speccy doesn't have suitable timers for playing back samples on an interrupt. If you were going down the 1-bit route, wouldn't you just go with the beeper?

I could see some benefits to using another sample quality, although I suspect most developers would've gone with 4-bit just because it's easier and more likely to sound better.

It's definitely "Meteor Alert"

AndyC wrote: ↑Mon Apr 17, 2023 5:15 pm Would there be any benefit to 1-bit samples on the AY. It seems like it would massively complicate any code and you wouldn't really benefit from the AY being asynchronous because the Speccy doesn't have suitable timers for playing back samples on an interrupt. If you were going down the 1-bit route, wouldn't you just go with the beeper?

I could see some benefits to using another sample quality, although I suspect most developers would've gone with 4-bit just because it's easier and more likely to sound better.

The only possible benefit I can think up is that you have only /one/ sound source, so you do not have the volume discrepancies between AY and beeper that manifest on some models, although I doubt that has ever been anyone's real world argument.

I've also long forgotten if both beeper and AY are sent to the television speaker.

Yeah, both the AY and beeper go to the television. The beeper being about the same volume as all three AY channels together. I guess if you wanted quiet samples there might be a reason to do it.

On the AY chip, there's a something like enabling tone but setting the cycle period to 0 sets the output high. This means that the output signal then simply moves up and down with the volume control. So that's how samples are replayed; by rapidly altering the volume.

On the ST I know that the three channels don't quite output at the same volume, so programmers use look-up tables that disperse an 8-bit sample value across the three 4-bit channel volumes to get the best quality output. I dont know of any demos that played four-bit samples on separate channels. Most sample trackers would add four concurrent samples together, then play the result in the same way as a single channel sample. I don't know which approach has been used on the Speccy.

(You can also set the volume control to be the triangular wave envelope, and run the envelope fast enough to be audible. It gives you a triangular instead of square waveform, which sounds much more natural, at least for the bass frequencies it can reach).

Joefish wrote: ↑Tue Apr 18, 2023 11:47 am I dont know of any demos that played four-bit samples on separate channels. Most sample trackers would add four concurrent samples together, then play the result in the same way as a single channel sample. I don't know which approach has been used on the Speccy.

I'm pretty sure the norm for Speccy sample /music/ was to use 3 channels, see for example Digi Synth (press shift+space to skip the intro waffle). I "converted" this to the SAM Coupe back in the day (see Fred 46).

Oscilloscope video from SimCoupe (since I have nothing set up for creating AY oscilloscope videos) demonstrating that three channels were used:



For the SAM, one of the channels was doubled to left + right.

Satellite Killer says "DANGER ATTACK!"

Btw. do you any tool to convert wav to raw 1-bit sample ?

catmeows wrote: ↑Wed Apr 19, 2023 3:26 pm Btw. do you any tool to convert wav to raw 1-bit sample ?

You can convert a sample to 1-bit in Audacity, using a Nyquist script called Broadcast limited III. After installing it, select your audio, run the script, turn the Threshold down to 0, and the resulting sound will be 1-bit. However, the file itself will still be 16-bit and I don't know how to export as 1-bit. There's a utility called PCM2PWM which may do it?

catmeows wrote: ↑Wed Apr 19, 2023 3:26 pm Btw. do you any tool to convert wav to raw 1-bit sample ?

That'd be fun! Depends also how you're going to replay it. The old programs that let you capture and replay samples on the Speccy itself used to just detect high/low transitions on the EAR port and save the time between them. So it really was a 1-bit sample. And they sound awful.

Later stuff would have been sampled on something like an ST at 8-bits, then converted into 1-bit pulse-width-modulation, where the 1s and 0s come so fast they actually juggle the speaker into intermediate positions, like how multi-channel sound routines work. That might take some fiddly code though to run that fast.

And that's just for a beeper. On a 128K you could take an 8 or 16 bit sample and round it off to 4 bits. Then it's a question of whether you can afford the time to prise two samples out of one byte as you replay it or if you just have each byte of data only uses four bits.