Forums

For the past couple of hours I've been doing some research on the SN76489 PSG chip and there's some things that I truely don't understand. Mostly out of it's design. What were the targed electronics that Texas Instruments were designing for when they created these chips? Were these processors made for engineers who wanted to design sound chips for childrens toys or somethin? But, first I want to look at it as a product in it's whole. It has...

* 3 Square Wave channels
* 1 Noise/White Noise channel
* 1024 different frequencies
* 16 steps of volume

Like this, it seems like a descent product. You can have 4 different sounds playing at once at lots of octave ranges. Not to mention that it has 16 steps of volume; which is enough for velocity fades and such. Seems like something you could use for an ice-cream truck or a children's toy. It poses usefulness for some things. Mabye if you had two or three of those running at once it would create a nice chorus of sounds. But, here's some things that I see wrong wtih it already...

1) Why not allow it to play other types of sounds besides square waves? Sawtooth and Sine Waves are nice :o)
2) Why give it only 3 channels? It takes at least 3 to play a chord, so why not have 5 or 6 channels?
3) Why would you need 1024 different frequencies? There's 12 different frequencies in an octave and your human ear can only hear about 7 or 8 octaves. 1024 is damn there 43 octaves!

But now, let's look at it from a programming perspective. There is one data port, so all sound manipulation is done via encoded bytes to that port. If there's 1024 different frequencies, that requires exactly 10 bits. But, there's only 8 bits in a byte. So in order to change the frequency for even a single channel, it would require at least 2 writes. Not to mention volume changes and noise changes are done by the very same port. Yet, somehow programmers mannaged to get everything done and the chip remained throughout the 80s. Mabye even you all don't agree with me but I think it stinks. Right off of the bat, I would have designed the chip to have either more ports or more registers. I'd say the best would to be to add an extra port and have internal registers; similar to the way Yamaha did things. But hey, times have changed and we have bigger, better, and more effecient programming chips today. Why do I even bother to reflect on previous sound chips? Just to compare, contrast, and learn a lot more on the chip itself.

Chris :o)

The answer is oh, so very simple:

That family of chips existed to be cheap, small, and (in some applications) consume little power.

The features are simple because the circuit needed to be simple. More complicated features mean more transistors and more complicated design, and that means more expense. As well, it would probably require a large chip, and use more power. It might also need more heat dissapation.

These days, technology has advanced to the point where complicated circuits are cheap to produce. I could probably program a $2 PIC to do anything the TI chip can do, and more.

However, in the SMS days, a SN chip probably cost $10-$15 when purchased in bulk (I'm speculating the cost), vs. maybe $30-$40 or more for a good FM chip. Considering that the sound chip is only one component of an SMS, and Sega really needed to get it in stores well under $199, ideally around $99. And of course sega sells them to stores for less than that, so certain sacrifices needed to be made.

And previous to the SMS, the costs were even higher, probably over $100 for an FM chip, much less for a TI.

Of course, over time prices go down and performance goes up, to where a $99 console can do 24 channels of compressed sound with DSP effects plus streaming CD audio, in addition to 3-d graphics, 3.5 megs of onboard ram, and a 33mhz processor, and be considered -terribly outdated-

such is technology.

Quote

> The answer is oh, so very simple:

> That family of chips existed to be cheap, small, and (in some applications) consume little power.

> The features are simple because the circuit needed to be simple. More complicated features mean more transistors and more complicated design, and that means more expense. As well, it would probably require a large chip, and use more power. It might also need more heat dissapation.

> These days, technology has advanced to the point where complicated circuits are cheap to produce. I could probably program a $2 PIC to do anything the TI chip can do, and more.

> However, in the SMS days, a SN chip probably cost $10-$15 when purchased in bulk (I'm speculating the cost), vs. maybe $30-$40 or more for a good FM chip. Considering that the sound chip is only one component of an SMS, and Sega really needed to get it in stores well under $199, ideally around $99. And of course sega sells them to stores for less than that, so certain sacrifices needed to be made.

> And previous to the SMS, the costs were even higher, probably over $100 for an FM chip, much less for a TI.

> Of course, over time prices go down and performance goes up, to where a $99 console can do 24 channels of compressed sound with DSP effects plus streaming CD audio, in addition to 3-d graphics, 3.5 megs of onboard ram, and a 33mhz processor, and be considered -terribly outdated-

> such is technology.

Quote

> 1) Why not allow it to play other types of sounds besides square waves? Sawtooth and Sine Waves are nice :o)
> 2) Why give it only 3 channels? It takes at least 3 to play a chord, so why not have 5 or 6 channels?
> 3) Why would you need 1024 different frequencies? There's 12 different frequencies in an octave and your human ear can only hear about 7 or 8 octaves. 1024 is damn there 43 octaves!

Why doesn't your Soundblaster card support 3783781 channels at once?
Why doesn't it play MP3 natively and can play 3783781 on the same time, on each channel ?

Think about it.

Quote

> 3) Why would you need 1024 different frequencies? There's 12 different frequencies in an octave and your human ear can only hear about 7 or 8 octaves. 1024 is damn there 43 octaves!

im *assuming* those 1024 different frequencies increase linearly, where as octaves are exponetial increases (increasing by one octave is doubling the frequency). thats about 10 octaves in total the chip can cover.
12 different frequencies in an octave? maybe on a sheet of music, definatly not in the real world were the number of frequencies an octave covers tends toward infinity

Author	Message
Consolemu Guest	PSG Stuff Posted: Sat Jul 15, 2000 4:21 am
Consolemu Guest	For the past couple of hours I've been doing some research on the SN76489 PSG chip and there's some things that I truely don't understand. Mostly out of it's design. What were the targed electronics that Texas Instruments were designing for when they created these chips? Were these processors made for engineers who wanted to design sound chips for childrens toys or somethin? But, first I want to look at it as a product in it's whole. It has... * 3 Square Wave channels * 1 Noise/White Noise channel * 1024 different frequencies * 16 steps of volume Like this, it seems like a descent product. You can have 4 different sounds playing at once at lots of octave ranges. Not to mention that it has 16 steps of volume; which is enough for velocity fades and such. Seems like something you could use for an ice-cream truck or a children's toy. It poses usefulness for some things. Mabye if you had two or three of those running at once it would create a nice chorus of sounds. But, here's some things that I see wrong wtih it already... 1) Why not allow it to play other types of sounds besides square waves? Sawtooth and Sine Waves are nice :o) 2) Why give it only 3 channels? It takes at least 3 to play a chord, so why not have 5 or 6 channels? 3) Why would you need 1024 different frequencies? There's 12 different frequencies in an octave and your human ear can only hear about 7 or 8 octaves. 1024 is damn there 43 octaves! But now, let's look at it from a programming perspective. There is one data port, so all sound manipulation is done via encoded bytes to that port. If there's 1024 different frequencies, that requires exactly 10 bits. But, there's only 8 bits in a byte. So in order to change the frequency for even a single channel, it would require at least 2 writes. Not to mention volume changes and noise changes are done by the very same port. Yet, somehow programmers mannaged to get everything done and the chip remained throughout the 80s. Mabye even you all don't agree with me but I think it stinks. Right off of the bat, I would have designed the chip to have either more ports or more registers. I'd say the best would to be to add an extra port and have internal registers; similar to the way Yamaha did things. But hey, times have changed and we have bigger, better, and more effecient programming chips today. Why do I even bother to reflect on previous sound chips? Just to compare, contrast, and learn a lot more on the chip itself. Chris :o)

Heliophobe Site Admin Joined: 25 Oct 1999 Posts: 2029 Location: Monterey, California	Posted: Sat Jul 15, 2000 4:53 am
	The answer is oh, so very simple: That family of chips existed to be cheap, small, and (in some applications) consume little power. The features are simple because the circuit needed to be simple. More complicated features mean more transistors and more complicated design, and that means more expense. As well, it would probably require a large chip, and use more power. It might also need more heat dissapation. These days, technology has advanced to the point where complicated circuits are cheap to produce. I could probably program a $2 PIC to do anything the TI chip can do, and more. However, in the SMS days, a SN chip probably cost $10-$15 when purchased in bulk (I'm speculating the cost), vs. maybe $30-$40 or more for a good FM chip. Considering that the sound chip is only one component of an SMS, and Sega really needed to get it in stores well under $199, ideally around $99. And of course sega sells them to stores for less than that, so certain sacrifices needed to be made. And previous to the SMS, the costs were even higher, probably over $100 for an FM chip, much less for a TI. Of course, over time prices go down and performance goes up, to where a $99 console can do 24 channels of compressed sound with DSP effects plus streaming CD audio, in addition to 3-d graphics, 3.5 megs of onboard ram, and a 33mhz processor, and be considered -terribly outdated- such is technology.

Consolemu Guest	Can we say, Playstation? (nt) Posted: Sat Jul 15, 2000 8:01 am
Consolemu Guest	Quote > The answer is oh, so very simple: > That family of chips existed to be cheap, small, and (in some applications) consume little power. > The features are simple because the circuit needed to be simple. More complicated features mean more transistors and more complicated design, and that means more expense. As well, it would probably require a large chip, and use more power. It might also need more heat dissapation. > These days, technology has advanced to the point where complicated circuits are cheap to produce. I could probably program a $2 PIC to do anything the TI chip can do, and more. > However, in the SMS days, a SN chip probably cost $10-$15 when purchased in bulk (I'm speculating the cost), vs. maybe $30-$40 or more for a good FM chip. Considering that the sound chip is only one component of an SMS, and Sega really needed to get it in stores well under $199, ideally around $99. And of course sega sells them to stores for less than that, so certain sacrifices needed to be made. > And previous to the SMS, the costs were even higher, probably over $100 for an FM chip, much less for a TI. > Of course, over time prices go down and performance goes up, to where a $99 console can do 24 channels of compressed sound with DSP effects plus streaming CD audio, in addition to 3-d graphics, 3.5 megs of onboard ram, and a 33mhz processor, and be considered -terribly outdated- > such is technology.

Zoop Joined: 24 Jun 1999 Posts: 1732 Location: Paris, France	Posted: Sat Jul 15, 2000 9:20 am
	Quote > 1) Why not allow it to play other types of sounds besides square waves? Sawtooth and Sine Waves are nice :o) > 2) Why give it only 3 channels? It takes at least 3 to play a chord, so why not have 5 or 6 channels? > 3) Why would you need 1024 different frequencies? There's 12 different frequencies in an octave and your human ear can only hear about 7 or 8 octaves. 1024 is damn there 43 octaves! Why doesn't your Soundblaster card support 3783781 channels at once? Why doesn't it play MP3 natively and can play 3783781 on the same time, on each channel ? Think about it.

Limbs a Flyin' Guest	Posted: Sun Jul 16, 2000 3:23 pm
Limbs a Flyin' Guest	Quote > 3) Why would you need 1024 different frequencies? There's 12 different frequencies in an octave and your human ear can only hear about 7 or 8 octaves. 1024 is damn there 43 octaves! im assuming those 1024 different frequencies increase linearly, where as octaves are exponetial increases (increasing by one octave is doubling the frequency). thats about 10 octaves in total the chip can cover. 12 different frequencies in an octave? maybe on a sheet of music, definatly not in the real world were the number of frequencies an octave covers tends toward infinity

Forums

View topic - PSG Stuff