Recreate the Votrax Type & Talk?

[rmelick]

Ok Rick, thank you for your explanation. Now I can get most fof the stuff that's going on with your experiments. I'll follow their evolution with interest .
BTW... Will all this ever give a commercial product, i.e. something I will surely buy?

I've ordered some parts today for a prototype. I'm striving for a simple, classic white, all-in-one design, with built-in speaker and volume control, that attaches directly to the User Port on the VIC/64/128. Driven by the $25 Raspberry Pi, it needs to also be reliable, upgradable and open.

I know we have two viable approaches in the works at this point, and I wanted to give a little update on where I'm at. I have reached out to the Festival Speech System researchers at the University of Edinburgh to get help on the creation of a Votrax module for their framework. In parallel, I'll working in Perl on the PETSCII to ASCII issues.

In addition, five prototype units are being assembled, but are not ready for sale. Denial members, for a preview you can see the item page here: http://goo.gl/nx64h. For goodness sake, please don't order anything! However, you can request to be placed in the back-order queue now if you are interested in one of these units.

I'd like to launch something by July 4th, and maybe offer a firmware upgrade later if we ever get the Votrax stuff figured out. We'll see.

[orion70]

Yep please, I'd like to back-order one

[Z0rb]

In parallel, I'll working in Perl on the PETSCII to ASCII

Can you elaborate a bit on this issue I think I might be have a similar issue in the Speakjet approach.

Also what sort of data are you seeing coming out of the Scott Adams game. At 1200,8,N,1 I am seeing characters above CHR$(127) and I do not believe that the type n talk uses anything above 127. When I set up my com port to 1200,7,E,2, I see data that is below 127 but when I correlation those characters against what the type n talk expects it appears to be the wrong phonemes for the sounds it should make.

Any thoughts? Please advise on your PETSCII issue.

Thanks

[rmelick]

In parallel, I'll working in Perl on the PETSCII to ASCII

Can you elaborate a bit on this issue I think I might be have a similar issue in the Speakjet approach.

Also what sort of data are you seeing coming out of the Scott Adams game. At 1200,8,N,1 I am seeing characters above CHR$(127) and I do not believe that the type n talk uses anything above 127. When I set up my com port to 1200,7,E,2, I see data that is below 127 but when I correlation those characters against what the type n talk expects it appears to be the wrong phonemes for the sounds it should make.

Any thoughts? Please advise on your PETSCII issue.

Thanks

Well, to be honest, I don't know if PETSCII it is a a real issue for sure, so I'm exploring it further to see if eliminating it adds any value or not. Either way it is irritating to me for some reason.

When I was tinkering with the SpeakJet I got lots of sound effects and pings. Every now and then it would interject and say "vertical bar" in-between the gibberish.

Later, when I hooked my VIC up to the Raspberry Pi and just echoed the input to the terminal I saw no printable text, just a bunch of smile faces and random looking characters with occasional spaces and newlines interjected occasionally.

More tinkering showed me that the British Pound key on the Commodore side results in a backslash on the Pi side. That might be useful later. I still don't know how to create a ~ character.

All this PETSCII nonsense is just a variable I'd like to try and eliminate. That will allow me to fine tune the STTY settings on the Pi side. STTY is a whole other mess because there are so many options to interpret escape codes for terminals and such.

You are actually a lot further along than me, and more scientific in your approach. I'm just hoping to get lucky here at some point.

[Vic20-Ian]

Rick,

this may have been considered but I think the discussion it is missing in this thread.

How about disassembling Scott Adams cartridge and examining the Text to Speech routine.

It is likely to be a reused routine e.g.

Send Setup String commands to Votrax
Send Text String
Send Close String commands to Votrax

Maybe it would be easier to patch Scott Adams cartridge to remove the Votrax commands and thus use your existing hardware without modification.

[rmelick]

Rick,

this may have been considered but I think the discussion it is missing in this thread.

How about disassembling Scott Adams cartridge and examining the Text to Speech routine.

It is likely to be a reused routine e.g.

Send Setup String commands to Votrax
Send Text String
Send Close String commands to Votrax

Maybe it would be easier to patch Scott Adams cartridge to remove the Votrax commands and thus use your existing hardware without modification.

Its a great thought actually and may be worth pursuing. Can someone help with that?

It's a mystery what Commodore is doing for the setup. In addition, my concern at this stage is that I'm not seeing any evidence of text strings coming out the other end.

Z0rb may be on the right track tinkering with the framing. TNT expects 1 start bit, 8 data bits and 1 stop bit. They don't actually mention parity and I don't know if a start bit implies a kind of parity.

I had a Votrax Personal Speech System once and I don't remember having to change from no parity to even/odd when opening the device on the commodore side, but maybe TNT is different. (The kernal defaults to no parity unless it is explicitly set.)

[Z0rb]

The only thing I see when I look at what is coming out of the user port when playing Adventure Land at 1200,8,N,1 is what I would consider garbage.

[rmelick]

The only thing I see when I look at what is coming out of the user port when playing Adventure Land at 1200,8,N,1 is what I would consider garbage.

OK, so my PETSCII to ASCII conversion script didn't help much except to confirm that the adventures are sending characters codes > 0x7f. This is crazy for two reasons. There are no ASCII equivalents for any PETSCII characters greater than CHR$(127). Secondly, even if they were phoneme codes, Votrax's conversion chart also stops at the exact same place (0x7f).

RS-232 signals are inverted, could that be what's gong on? (All the 1's should be 0's and vise-versa.) We're TTL to TTL so that can't be it, can it? Did the RS-232 cartridge do anything other than transform the signal?

The adventure manuals state either the RS-232 cart could be used or a special Commodore cable from Votrax. I wonder what their cable did?

[Z0rb]

The adventure manuals state either the RS-232 cart could be used or a special Commodore cable from Votrax. I wonder what their cable did?

I'm wondering the same thing. I shot off a couple of emails. One to a guy that ripped apart and mapped out his Votrax Type and Talk, a fella named Kevin Horton. He said that maybe the settings for serial communication are not set up right. But I can definitely get my VIC to talk using a terminal program and my baud rate converter. When I set the VIC up for 1200,8,N,1. He also mentioned maybe ANDing bit 8 with bit 7 but I dont see how that would help matters.

I also shot off an email to Scott Adams himself to see if he has any insight.

What you are seeing is basically what I am seeing. Characters above CHR$(127). Votrax doesn't support these and the VIC don't like them either. Something has got to be going on in the conversion cable or the RS-232 interface that neither of us are seeing.

When I set my PC up to receive 1200,7,E,2 I got characters below CHR$(127) as could be expected but none of them followed any of the Votrax format. No '~'s and very few "?"s and all the alleged phoneme data was all none sense that wouldn't make even the most drunken slur.

I am going to keep plugging away at this but I think we should all keep an open dialog here to update one another on our findings so we can get both solutions working.

Also if anyone can comment on how the Votrax Vic interface cable works or the RS-232 interface works that would help greatly.

Thank All

[Vic20-Ian]

Are we sure that the Serial out or the RS232 cart used 1200 8N1?

Is there any spec sheet for the Votrax comm port also?

[rmelick]

Are we sure that the Serial out or the RS232 cart used 1200 8N1?

Is there any spec sheet for the Votrax comm port also?

Kind of sure. The adventure games manual only mention 1200 baud and don't specify the 8N1 directly. Indirectly, it could only be that, becaus the TNT was hard-coded for only those settings. They couldn't be changed on the TNT side.

Maybe Commodore mis-printed the baud rate?

[Vic20-Ian]

Manual - does this help any, especially the phoneme commands?


https://sites.google.com/site/analogvoc ... tnt-manual

[Vic20-Ian]

See the post below - we need 2400 Baud to read the output of the Scott Adams cartridges.

http://www.sleepingelephant.com/ipw-web ... c&start=15

See page 2 3rd from last post

[rmelick]

See the post below - we need 2400 Baud to read the output of the Scott Adams cartridges.

http://www.sleepingelephant.com/ipw-web ... c&start=15

See page 2 3rd from last post

Got it. 2400 baud works. Adventure Land is talking. I'm still testing...

[Vic20-Ian]

Please record samples for Jeff!