BackBit cartridge on the Intellivision

[8bitwidgets.com]

oh sorry right. need intellivoice. definitely

[+DZ-Jay]

5 hours ago, Lathe26 said:

I don't think @evietron was going to emulate an Intellivoice, just support Intellivoice games when the BackBit was plugged into a real Intellivoice.

 

5 hours ago, mr_me said:

It emulates Intellivoice cartridges, I don't think it emulates the Intellivoice.  You still need an Intellivoice if you want to hear speech.

 

5 hours ago, Caleb Garner said:

No it definitely does.  I've been in casual communication with @evietron and she confirmed B-17 bomber and others work with it.  I'm still waiting for mine to arrive to really test it, but I take her at her word if she says it does.  

Sorry, yes, it supports the Intellivoice, but does not emulate it.

 

   dZ

Edited August 6, 2022 by DZ-Jay

[+DZ-Jay]

20 hours ago, artrag said:

Is backbit going to emulate the intellivoice?

 

1 minute ago, DZ-Jay said:

 

 

Sorry, yes, it supports the Intellivoice, but does not emulate it.

 

   dZ

@artrag, do you need an Intellivoice?  Then send me a PM, I have spares.

 

    dZ.

[artrag]

1 hour ago, DZ-Jay said:

 

@artrag, do you need an Intellivoice?  Then send me a PM, I have spares.

 

    dZ.

Thanks but I was only curious 

Emulating the intellivoice would have been an amazing feature for a modern cartridge 

[carlsson]

I got the impression that BackBit operates on a KISS principle and since it is used with so many systems, it can't be too specific. However it has memory expansion (up to 32x8 kilobits it appears), can route joysticks through the cartridge port if the system supports it, has built-in SID player on C64 (using the computer's own hardware), some degree of floppy drive emulation, and a jumper to create a reset line on systems that normally can't be reset through the cartridge port. Emulating other hardware, whether an Intellivoice, secondary SID, adding an ARM or other co-pro etc probably would soon go into too system specific rabbit holes.

[8bitwidgets.com]

yea thankfully intellivoices are not too hard to come by which i think is pretty amazing..   and honestly it kinda blows my mind that so many of these intellivoice modules sold and no games (from what i know) beyond the 4 originals ever supported it.  i mean B-17 bomber is more than enough to justify it lol.. but seriously..   with so many people buying it, you'd have thought developers would have supported it..   maybe it came out shortly before intellivisions decline?  

[+DZ-Jay]

4 minutes ago, Caleb Garner said:

yea thankfully intellivoices are not too hard to come by which i think is pretty amazing..   and honestly it kinda blows my mind that so many of these intellivoice modules sold and no games (from what i know) beyond the 4 originals ever supported it.  i mean B-17 bomber is more than enough to justify it lol.. but seriously..   with so many people buying it, you'd have thought developers would have supported it..   maybe it came out shortly before intellivisions decline?  

B-17 Bomber is one of my all-time favourite titles, but when I bring out the Intellivoice, I find myself playing a few rounds of Bomb Squad.

 

My understanding is that the reason there weren't many titles was that people weren't really buying the device.  They had just spent several hundred dollars for the console and not many were keen on spending more for what appeared to be mostly a gimmick.  Making voice-enabled games was hard and costly, and designing a compelling title in which voice was a necessary key part, was not trivial.

 

Besides, there were plenty of games without voice already, and kids wanted action and arcade ports, anyway.

 

It is true that there appears to be plenty of Intellivoice units in circulation, but I think it is because they were sold at bargain basement prices, and sometimes with coupons that included a free game.  That all happened later on, when the crash caused Mattel to rethink their video game strategy, so there was no interest in making more voice-enabled games.

 

    dZ.

[+-^CrossBow^-]

Got my shipping notification yesterday so I too will be able to check out the cart in the future when I've time to do so.

 

The chip tester I'm especially excited to get as that will make some repair jobs much simpler.

 

[First Spear]

2 hours ago, Caleb Garner said:

yea thankfully intellivoices are not too hard to come by which i think is pretty amazing..   and honestly it kinda blows my mind that so many of these intellivoice modules sold and no games (from what i know) beyond the 4 originals ever supported it.  i mean B-17 bomber is more than enough to justify it lol.. but seriously..   with so many people buying it, you'd have thought developers would have supported it..   maybe it came out shortly before intellivisions decline?  

 

(shameless plug)

 

There is some good info about that in the Classic Intellivision FAQ, check hardware, section 2.6 https://www.intvprime.com/intellivision-faq/?date=20220806

 

(/shameless plug)

 

[+DZ-Jay]

15 minutes ago, First Spear said:

 

(shameless plug)

 

There is some good info about that in the Classic Intellivision FAQ, check hardware, section 2.6 https://www.intvprime.com/intellivision-faq/?date=20220806

 

(/shameless plug)

 

Thanks for that.  Here's the relevant information from the FAQ:

Quote

Because Mattel Electronics misread the market in late 1982 (buyers became enamored with arcade ports), and the games were complex and expensive to produce (dedicated months of salary people like voice actors which has never been done before), far more units were produced than sold. Mattel Electronics tried different cost-cutting bundles as well, flooding the market even more.

 

By the way, @First Spear, would it be possible to make the TOC at the top of the page link directly to the sections?

 

   -dZ.

 

[mr_me]

In 1982 when Intellivoice was new, there would have been reluctance for many programmers to develop for Intellivoice.  Why make a cartridge for something that has no install base vs something that has millions.  Plus much of the ROM space has to go to speech data at a time when rom space was already tight.  B17 Bomber and Space Spartans still sold about as well as other popular cartridges that came out at the same time like Night Stalker and Deadly Discs.  So Intellivoice sales were there, maybe they expected better.  Then in 1983, when Mattel Electronics was making cuts, hardware was the first to go including a few Intellivoice games they had in the pipeline.

[First Spear]

25 minutes ago, DZ-Jay said:

Thanks for that.  Here's the relevant information from the FAQ:

 

By the way, @First Spear, would it be possible to make the TOC at the top of the page link directly to the sections?

 

   -dZ.

 

Thanks for the idea/feedback. Will add to the backlog, it definitely needs deep links. Currently working on redoing the Intellibot, so anyone on the site can just ask a question and get a relevant piece of info.

 

[8bitwidgets.com]

3 hours ago, DZ-Jay said:

I find myself playing a few rounds of Bomb Squad.

I have to sit down and figure that one out..  like solar sailor..  i just haven't wrapped my head around them.  

 

3 hours ago, DZ-Jay said:

people weren't really buying the device.

Yea that must have been the case.  It just seemed like we had it when i was a kid and maybe we got it later on at a discount.  that wouldn't have surprised me but i know B-17 bomber was big in our house and my grandpa who was a WW2 vet and lover of aviation..  but obviously dev folks at the time would have supported it if it made sense and it didn't for some reason and number of units sold would likely be the biggest one!  I definitely respect the concern of voice eating precious rom real estate..  

 

3 hours ago, DZ-Jay said:

and designing a compelling title in which voice was a necessary key part, was not trivial.

Yea true, I would have though folks might try at least like KC's crazy chase where you don't NEED the voice, but it's a nice enhancement if you do.  I definitely could see games being reluctant to be designed with intellivoice as being a requirement as this would certainly cut down on the number of potential sales.  but yea i suppose it goes back to the rom space factor.  enhance a game for a few people sacrificing space for more game content that all players could benefit from. 

 

3 hours ago, DZ-Jay said:

I think it is because they were sold at bargain basement prices

Right, yea that makes sense.  Early adoption was poor and that's when it was needed most.  

 

 

[1980gamer]

The voice in WSMLB was well done.

 

I am hoping the upcoming Parsec will have the voices from the TI version.

 

Imagic's Microsurgeon had voice on the TI and I believe the DISC version of Demon Attack had voice as well.  To missed opportunists for Imagic. 

[carlsson]

43 minutes ago, 1980gamer said:

I am hoping the upcoming Parsec will have the voices from the TI version.

I believe the TI and GI chips are internally quite different, though principally alike, so it would take some recording of data. As far as I know there only exists one individual today (at least who has made it public to have tools for that) and that person seems 1000% busy plus that for some reason or not the tools may not be released to the public, 40 years after they were commercially viable. I'm surprised that no clever person reverse engineered the GI speech chips by now, or perhaps someone has but chooses to not post in public about their findings. Perhaps the General Instruments alumni has an active mafia hunting down anyone disclosing their secrets?

[8bitwidgets.com]

Mine arrived but ack..  no microSD card reader to transfer roms too.. had to order one.  the one i had was built into my iMac which is no longer available..  looks fantastic though.  can't wait to load it up and finally be able to test my builds and other homebrews on real hardware!  

[+DZ-Jay]

13 hours ago, carlsson said:

I believe the TI and GI chips are internally quite different, though principally alike, so it would take some recording of data. As far as I know there only exists one individual today (at least who has made it public to have tools for that) and that person seems 1000% busy plus that for some reason or not the tools may not be released to the public, 40 years after they were commercially viable. I'm surprised that no clever person reverse engineered the GI speech chips by now, or perhaps someone has but chooses to not post in public about their findings. Perhaps the General Instruments alumni has an active mafia hunting down anyone disclosing their secrets?

LOL!
 

For the record, the Intellivoice chip has been fully reverse-engineered and documented by Frank Palazzolo and Joe Zbiciak.  What is "black magic" (as Joe calls it) is how to encode speech in a usable way to the chips.  Mattel used proprietary encoders and hardware to do this back in the 1980s, and none of that (as far as we know) is currently available.

 

My understanding is that it is a highly specialized technology.  The problem is not that it is secret, it's that the speech is encoded in a way that it tries to reproduce the human vocal tract.  This is entirely different from what we use today for sampling and synthesizing speech, so it is very arcane knowledge.

 

As far as I understand, the information to do this is out there, not secret; it just takes quite some effort to assimilate (as you know yourself, at least one person did it).

 

@intvnut shared with me once some high level information about the technology.  It is not a secret, it is just that I am not too smart to do anything with it.

 

My understanding is that the the orator chip uses a 12-pole linear predictive coding (LPC) process, in which the coefficients map directly to the parameters of the speech synthesizer.  What I remember is that intvnut found some research from the Naval Academy describing an algorithm, and some code from a university (most of it proprietary) that he hacked all together until it sort-of, kind-of worked.  Essentially, you transform PCM samples into LPC coefficients. (Actually he uses 10 Poles, althought the Intellivoice supports 12.)

 

He told me all practical LPC code he could find at the time was either proprietary, or derived in some way from the same original LPC-10e DoD vocoder implementation, which he used himself.

 

What he came up with is not open source, and apparently not very reliable either, which is the reason he has stated for not sharing it yet.

 

I'm sure anybody with some interest and motivation can figure it out, but it appears that LPC encoding is an industry in itself with very specialized military and commercial applications, and so I guess, it is not the sort of thing you find in GitHub from some bored university student (although it's been 10 years, so maybe now it is?).  PCM waveform sampling seems to be so much easier for what we use it now.

 

Like I said, I looked into it 10 years ago, and got stuck in a rabbit hole that I didn't think I was smart enough to go through, or for which I did not have the stamina, interest, and time to dedicate.

 

   dZ.

Edited August 7, 2022 by DZ-Jay

[+DZ-Jay]

Here's some information, straight from the horse's mouth.  I hope @intvnut does not mind me sharing this.  From the context of our conversation, this was not intended to be a secret.  It's been over 10 years (this is from an e-mail conversation in April, 2011), and I believe it may be useful to others.  Knowing his generosity of spirit and his enthusiasm for bringing these technologies back to life, I imagine that he wouldn't mind if someone took the mantle with this effort and was able to learn from his own ideas.

 

Quote

This page gives a short overview of the LPC-10e encoder: http://www.otolith.com/otolith/olt/lpc.html

 

The LPC-10e encoder actually has two phases:  LPC analysis (incl. pitch tracking, voiced/unvoiced determination, etc.), and then quantization/encoding.  See, LPC-10e not only does LPC analysis, it also defines a specific bitstream format for packaging up all the speech parameters, as well as an algorithm for how to (lossily) compress all the computed coefficients.

 

My encoder only uses the first phase (analysis), and skips the DoD quantizer/encoder.  I take the full-resolution pole/zero information and instead encode that in the Intellivoice's bitstream format.  Currently, I don't do any compression at all beyond quantizing to the Intellivoice's pole/zero ROM values.  I use the Intellivoice's "load everything" opcode to update all speech parameters completely, every frame, at full resolution.  (You can read up on the Intellivoice's opcodes here:  http://spatula-city.org/~im14u2c/intv/tech/sp0256_instr_set.html)

 

If you're keeping score, though, you'll see that the Intellivoice is 10kHz LPC-12 and the encoder I started with is 8kHz LPC-10.  That's OK.  I just take a 10kHz audio file and feed it to the LPC-10 coder.  It just thinks everyone's speaking with a lower pitch than they're actually speaking with.  It still makes pretty good decisions.   The LPC-10 coder also only gives me 10 poles to work with.  That's also OK.  I just fill in the missing poles with 0, running the Intellivoice in LPC-10 mode at 10kHz.

 

After not having looked into this in 10 years, it all goes way over my head.

 

     -dZ.

[+BBWW]

On 8/6/2022 at 2:12 PM, 1980gamer said:

The voice in WSMLB was well done.

 

I am hoping the upcoming Parsec will have the voices from the TI version.

 

Imagic's Microsurgeon had voice on the TI and I believe the DISC version of Demon Attack had voice as well.  To missed opportunists for Imagic. 

Keep hoping...

 

[8bitwidgets.com]

OK not sure if anyone else confirmed this but JLP games don't appear to work.  @DZ-Jay Christmas Carol wouldn't play.  it does show the opening screen with the xmas lights twinkling but no audio and it won't go beyond that screen.

 

other than that very solid product.   

 

Was there a final word on bank switching and JLP being incorporated?  @evietron  

[mr_me]

Christmas Carol shouldn't need any JLP functions, it runs in old emulators from the 90s.  Are you running a bin/cfg file or an intellicart .rom version?  The backbit cartridge doesn't support cfg files.

[+DZ-Jay]

30 minutes ago, mr_me said:

Christmas Carol shouldn't need any JLP functions, it runs in old emulators from the 90s.  Are you running a bin/cfg file or an intellicart .rom version?  The backbit cartridge doesn't support cfg files.

 

The Christmas Carol ROM published on the web site does not use the JLP, but it does rely on bank-switching to overlay the ECS EXEC ROM address space and use it as an additional data segment. (It uses 38K of the known 42K memory map.)

 

I suppose that if you do not have the ECS plugged in, this should not matter.

 

     -dZ.

[8bitwidgets.com]

here is a video (unlisted, not for the public) I made showing what happens. 

 

Bank switching is probably the issue as i don't think this is supported yet, but hopefully i'll get word on when / if this is happening soon. 

 

 

[mr_me]

Looking at the Christmas Carol cfg file, I don't see bank switching but I do see cartridge ram.

 

[mapping]
$0000 - $0EFF = $2100
$0F00 - $0FFF = $4800
$1000 - $2FFF = $5000
$3000 - $34FF = $7100
$3500 - $51FF = $A000
$5200 - $91FF = $C000
[memattr]
$8000 - $9EFF = RAM 16

 

edit:

If you have an ECS computer module, you can try World Series Baseball to test ECS bank switching.

Edited August 13, 2022 by mr_me

[+DZ-Jay]

1 hour ago, mr_me said:

Looking at the Christmas Carol cfg file, I don't see bank switching but I do see cartridge ram.

 

[mapping]
$0000 - $0EFF = $2100
$0F00 - $0FFF = $4800
$1000 - $2FFF = $5000
$3000 - $34FF = $7100
$3500 - $51FF = $A000
$5200 - $91FF = $C000
[memattr]
$8000 - $9EFF = RAM 16

 

edit:

If you have an ECS computer module, you can try World Series Baseball to test ECS bank switching.

 

I think that's Intellicart Bank-Switching.  The old "cart-mac" library used by Carol employed the ECS bank-switching, I believe.  Or the other way around.

 

        IF _m.edis = 1
        ORG     $4800
        ; Disable ECS ROMs so that they don't conflict with us
        MVII    #$2A5F, R0
        MVO     R0,     $2FFF
        MVII    #$7A5F, R0
        MVO     R0,     $7FFF
        MVII    #$EA5F, R0
        MVO     R0,     $EFFF
        B       $1041       ; resume boot
        ENDI

 

("_m.edis" is set to "1" for the 42K memory map.)

 

I do not really recall how all that stuff works right now (it's been a while), but I know that other emulators have had issues with the old ROM of Christmas Carol, and the guys developing the Android emulator also had to employ some work-arounds to get the ROM to work.

 

As I recall, Christmas Carol was the first game to use the "cart-mac" memory allocation and structure library, and so I believe there may be some issues on the way all that stuff was configured back then in a non-standard way.

 

    -dZ.

Edited August 13, 2022 by DZ-Jay