7800 sound

[Bruce Tomlin]

Hmm, interesting idea.  That would reduce the 40 pin POKEY down to 18-20 pins:

 

3 power, ground, clock

1 audio out

4 address

8 data (write only)

2 chip select (one high, one low, might be able to make do with only one)

(might still need a R/W line to latch the data properly)

 

But I suspect to do so would have cost Atari $$ in retooling, instead of just dipping into the parts bin.  Remember that POKEY was used for a lot of different things, including some arcade titles.

It shouldn't have needed anything more than a different packaging, like how the 6507 was a 6502 with fewer address lines bonded out to pins. Even if the die was too big for a .3" 20-pin chip, it would have fit in a .6" 24-pin chip.

 

But Atari was too cheap by then to even care. Jack just saw everthing that wasn't from the computer line as something to be clearanced.

 

EDIT: I wonder if a PIC could be programmed to emulate a POKEY chip?

Edited October 31, 2005 by Bruce Tomlin

[+Mitch]

Hmm, interesting idea.  That would reduce the 40 pin POKEY down to 18-20 pins:

 

3 power, ground, clock

1 audio out

4 address

8 data (write only)

2 chip select (one high, one low, might be able to make do with only one)

(might still need a R/W line to latch the data properly)

 

But I suspect to do so would have cost Atari $$ in retooling, instead of just dipping into the parts bin.  Remember that POKEY was used for a lot of different things, including some arcade titles.

It shouldn't have needed anything more than a different packaging, like how the 6507 was a 6502 with fewer address lines bonded out to pins. Even if the die was too big for a .3" 20-pin chip, it would have fit in a .6" 24-pin chip.

 

But Atari was too cheap by then to even care. Jack just saw everthing that wasn't from the computer line as something to be clearanced.

 

EDIT: I wonder if a PIC could be programmed to emulate a POKEY chip?

957417[/snapback]

 

Well, you can in an FPGA anyway. Check out this.

 

Mitch

[supercat]

EDIT: I wonder if a PIC could be programmed to emulate a POKEY chip?

 

A PIC could probably produce all the sounds of a POKEY, but unless the POKEY has strange timing requirements I would doubt that a PIC could adequately respond to the CPU bus (it's possible and practical for two POKEY writes to occur less than 3us apart, right? No way the PIC can respond that fast).

[gdement]

It just seems short sighted to me that Atari expected companies to go thru the extra expense of adding the pokey chip.  I could only see that happening if the 7800 was as big as the VCS in sales so the added cost was made up in volume.  Especially since they didn't even do it themselves on 99% of the games so Atari didn't even want to be bothered with it.

 

Here is a theory I had on this...

 

If you have looked at the 7800 PCB it's pretty crowded, so adding a pokey to the design probably would have increased the size of the PCB, the size of the system, and the overall cost of the system.

 

A lot of the 7800 carts had one of three "added features", extra RAM, extra ROM, or a POKEY, so the Pokey cart would not have been much more expensive then the other carts to make. Even if you had to tack on a few dollars to the price of the cartridge, this may have actually been more palatable to the consumer then increasing the price of the console by say $10.00.

 

Dan

955276[/snapback]

 

I don't think the Pokey cart situation was the original plan, because it really doesn't make any fiscal sense. If they were concerned about pricing, they could just as easily have swallowed the cost of integrating a POKEY in the console, and jacked up the price of some "Supercarts" that use it. The consumer wouldn't know the difference, and it would save Atari money. They didn't have to charge the Pokey-tax on the component where it actually was located.

I think it's more likely that they either didn't think the sound upgrade was needed, or GCC just didn't realize they were short of board room until it was too late. They thought they were in a hurry to ship, so they punted.

 

It seems to me that integrating sound in the MARIA would have been too much complexity for one new chip, but I don't know much about such things. I wonder if they could have instead designed a downward compatible version of Stella or RIOT that had some enhanced sound features integrated. But I suppose the RIOT might have been problematic to imitate since it's not an Atari chip to begin with.

[sukotsu9]

This scenario keeps going through my head since I started reading 7800 history: what if it had released on its original date, thus pre-empting the Famicom/NES console by 2 years? That 2 year lag is where I think all the 7800 weaknesses lie. I don't know about POKEY development, but I would guess it was developed somewhat after the processor board was completed (too small = no additions). Perhaps someone consolled themself with the idea that carts would contain POKEYs, increasing cost for cart makers while keeping the ProSystem cost down. And, as mentioned, the smug idea that gamers wouldn't notice the difference. Which may have been so in '84, but not by '87.

 

Having said all that, I can't say I'm really impressed with POKEY sound that much. I dunno, some non-POKEY titles are better and many are worse. It almost seems like it's more dependent on programming, RAM, etc. than having a better sound chip.

[EricBall]

The main advantage of POKEY is it has more bits for the frequency divider, which makes it much easier to make music that doesn't sound like an out-of tune electronic piano. However, when it comes to that actual sound (waveforms), POKEY doesn't have that much more than the TIA. So maybe it's better to say that POKEY means better music rather than better sound. (Well, it does have four 8 bit channels while the TIA only has two 5 bit channels.)

[supercat]

So maybe it's better to say that POKEY means better music rather than better sound.  (Well, it does have four 8 bit channels while the TIA only has two 5 bit channels.)

960809[/snapback]

 

Too bad Atari couldn't have spared two latches two NAND gates, and two NOR gates (or four latches and a few more gates) to allow a divide-by-two or divide-by-four prescale option in the TIA.

 

Basically, what I would have liked to have seen would have been to have the horizontal timing generator produce two or four pulses per scan line (ideally spaced about equally, but it doesn't matter too much--there should be something in the timing circuitry that generates the necessary pulses, e.g. start of blanking, start of sync, end of blanking, and middle of line (score mode). If the puses fire in the order p1, p2, p3, p4, then each audio input would be driven with:

 

p1 or ((p2 or p4) and L2) or (p3 and L1)

 

with whatever inversions were convenient. Just that little bit of circuitry would have greatly improved the available intonation accuracy on the TIA.

 

BTW, where does the TIA get its clock from? It's not HSync--what is it?

[EricBall]

BTW, where does the TIA get its clock from?  It's not HSync--what is it?

960837[/snapback]

 

The TIA is driven by the 3.58MHz colorbust clock, which is then divided down by some cascaded LFSRs. There's a document (TIASOUND) by Ron Fries which gives the details. (Or have a look at the schematics in the AtariAge archive and puzzle it out from there.)

[supercat]

The TIA is driven by the 3.58MHz colorbust clock, which is then divided down by some cascaded LFSRs.

961095[/snapback]

 

I should have clarified: the TIA is driven by 3.579545MHz colorburst, but the sound channels seem to have a pre-divisor which is not the horizontal sync rate. The piece I read suggests that that may be driven by either colorburst or by colorburst/3, but that would suggest that each sound channel has its own pre-divisor which would seem rather wasteful. I wonder why they'd do that instead of using hsync or hsync*2 (or *3, or *4), both of which are already produced elsewhere and wouldn't need dividing down.

[supercat]

It shouldn't have needed anything more than a different packaging, like how the 6507 was a 6502 with fewer address lines bonded out to pins.  Even if the die was too big for a .3" 20-pin chip, it would have fit in a .6" 24-pin chip.

957417[/snapback]

 

I wonder if Atari ever manufactured any POKEY chips after the 7800 came out, or of they were only using up old stock? Throwing a newly-made POKEY in a 28-pin package shouldn't have been a particular problem (looking at the die map, smaller packages might have been) but if all the POKEYs in existence had already been thrown in DIP40 packages there would have been no reasonable way to change that.

[Bruce Tomlin]

The TIA is driven by the 3.58MHz colorbust clock, which is then divided down by some cascaded LFSRs.  There's a document (TIASOUND) by Ron Fries which gives the details.  (Or have a look at the schematics in the AtariAge archive and puzzle it out from there.)

Here's part of a scan of the Coleco reverse-engineered documentation for the TIA. The waveforms are on the second page.

 

http://atari7800.net/files/coleco-tia-figs.pdf