AY-3-8900-1 vs AY-3-8900-1A

[SmakyTaky]

Are there any functional differences between these two ICs? I for a couple of ICs labeled AY -3-8900-1A with the intent to repair an original faulty Intellivision. When I use these ICs on the console it boots and loads the cartridge but the background is missing and some graphical elements are missing two. Besides that, games are working... I can see bullets and crosshair moving around (Using Atlantis for testing) as well as enemy ships. 

 

I am wondering if these ICs with an A at the end are functionally equivalent to those without the A. When an AY-3-8900-1 working IC (from another Intellivision) is used, the console works just fine.

 

Thank you for any insight into this.

[mr_me]

https://forums.atariage.com/topic/350541-input-needed-on-tutor-pro-system-super-pro-tutorvision/?do=findComment&comment=5413169

 

A hint in the post linked above. 

"When Valeski ran out of the old graphics chip sets he had to go to Toshiba to see if they too could locate their old design and fabricate some STIC 1A parts. They could and did. The parts didn't quite work right—they really needed one more design iteration. But APh suggested workarounds that required a few extra parts. Valeski was spending his own money and didn’t want to pay for another iteration of chip design, and so elected to include the parts for the workaround on the boards."

 

According to that post, the Stic1a was a cost reduced design meant to be a compatible replacement. It didn't work right so they fixed it on the main board rather than fixing the chip. The post also suggests that the INTV1987 board revision was meant to work with the original Stic1 not Stic1A. I haven't seen an INTV1987 board but I have seen an INTV1987revA board. Maybe this chip works there.

 

What date code is on these chips and where did you get them?

 

 

Edited May 8 by mr_me

[SmakyTaky]

Both ICs have a datecode of 8315. and were sourced from UTSource.com, I was not aware of any difference so I ordered a couple, one to have as a spare and the other one to keep in my stash. I was thinking they could be fakes.. but since they do kind of work I was wondering if they might be real ICs but that should be used on other INTVs.

 

Edited May 8 by SmakyTaky

[Walter Ives]

On 5/7/2024 at 4:13 PM, SmakyTaky said:

AY-3-8900 vs AY-3-8900-1A

Are there any functional differences between these two ICs?

Thank you for any insight into this.

An end letter like the A here would have been used to indicate a revision that was nominally functionally identical to the original part but had some small difference that designers had to be aware of. For example, the first tens of thousands of units of Intel's original 8080 didn't meet its published output driving specification; Intel fixed the problem and all of the gazillions of parts it subsequently sold were marked 8080A.

 

GI's yields on Intellivision chip sets was never very good. The parts you have were manufactured toward the very end of GI's production run, when prices were getting super competitive and Mattel was putting a lot of pressure on GI to improve yields in order to lower overall manufacturing costs. Rev A was made to improve the yield, but recollections are too foggy to remember off-hand exactly what it entailed. The AY-3-9600 System RAM was reved too, and there are vague recollections that the two revs may only have worked together.

 

GI's only customer for the STIC was Mattel, and it only produced parts to order. All but the very last runs of parts it manufactured should have been delivered to Mattel and eventually made their way to Valeski to be used in building Super Pro Systems. The parts you have were likely discovered hiding under some rocks by bottom feeders when GI's and Radofin's manufacturing plants were later liquidated.

 

Insight enough?

 

WJI

 

[Walter Ives]

On 5/8/2024 at 10:15 AM, mr_me said:

According to that post, the Stic1a was a cost reduced design meant to be a compatible replacement.

You misunderstand. It was the logic board a whole that was meant to be a compatible replacement. The STIC 1A chip was not a part-for-part replacement for the General Instrument AY-3-8900: it was more the centerpiece of a system-level integration done by APh that took advantage of more modern processes and used off-the-shelf memories. As you can see from the following picture posted elsewhere on this site, it was packaged in a 68-pin package. The board uses a 28-pin version of the CP1610 and doesn't use the GI System RAM, GROM or Sound Chip.

 

https://content.invisioncic.com/r322239/monthly_2022_04/20220404_140252.jpg.55ed9181ee1dbf5a5919ce66a03747d9.jpg

 

WJI

 

Edited June 8 by Walter Ives

[Walter Ives]

On 5/8/2024 at 10:15 AM, mr_me said:

It didn't work right so they fixed it on the main board rather than fixing the chip.

The problems with the first parts, which Toshiba delivered to Mattel in the spring of 1983, could have easily been fixed in a single rev, which was extremely good for a chip of that complexity. The reason Toshiba was able to deliver an almost-working part on the first try is that was that it had been given a fully working hardware emulator with the internal timings all carefully worked out and tested. Mattel put making the rev on hold only because sales had collapsed and it was drowning in inventory of the original chip sets. It was a good decision: at the time it was made, Mattel had more than enough inventory of the original chip sets to last it through the rest of the year, at which point the follow-up STIC 1B would be available. (STIC 1B wouldn't be mothballed until late July or early August 1983, after Morris replaced Denham.) As things worked out, it took Valeski five years to run through the remaining inventory of original chipsets. When he did, the very few problems STIC 1A did have could be fixed with just six external parts—that's truly amazing! And, with those fixes in place, it was very reliable. Way to go, Toshiba!

 

WJI

 

[Walter Ives]

On 5/8/2024 at 10:15 AM, mr_me said:

The post also suggests that the INTV1987 board revision was meant to work with the original Stic1 not Stic1A.

Check out the parts lists for the INTV 1987 and INTV 1988 boards at https://wiki.console5.com/wiki/INTV_System_3:

 

INTV 1987 REV A (2609-4379)
    U1: CP-1610A
    U2: RAM (System): RA-3-9600A
    U3: ROM (EXEC): R0-3-9504-021
    U4: STIC: AY-3-8900-1A
    U5: ROM (Graphics): RO-3-9503-003
    U6: Sound Generator, Controller Input: AY-3-8917
    U7: SCM2114ALE-4
    U8: SCM2114ALE-4
    U9: ROM (EXEC): RO-3-9502-011
    U10: Color Processor Chip: AY-3-8915
    U11: DM7407N
    U12A: SCM2114ALE-4
    U12B: SCM2114ALE-4
    U13: M74LS27P
    U14: T74LS125AB1
    U15: T74LS125AB1
    U16: 74LS00
    U17: SN74LS86J
    RF Modulator: Biwave, Inc. VM-445

 

Notes: I highlighted the interesting part number in red. U1 is the 28-pin, 5-volt only version of the CP1610 that GI was commissioned to lay out for use as part of the Intellivsion II cost reduction effort. The other parts are pretty much the same parts used in the original Intellivision, except the original U12 256x8 RAM was obsolete and was replaced with U12A and U12B, two industry standard 2114 1Kx4 RAMs, only 240 bytes of which were enabled.

 

 

INTV 1988
    U1: CPU: CP1610A
    U2: 3504-730A WBEXEC
    U4: KS74HCTLS174AN
    U5: SCH21014E
    U6: SCH21014E
    U7: SCH21014E
    U8: SCH21014E
    U9: STIC 1A BV2 8930 AY-3-8930
    U10: KS74HCTLS374AN
    U11: ROM (Graphics): 0098-0210 GROM
    U12: EL6116LP-10
    U14: KS74HCTLS257AN
    U15: KS74HCTLS257AN
    U16: 7407
    U17: KS74HCTLS08AN
    U18: KS74HCTLS04AN
    U19: KS74HCTLS73AN
    U20: KS74HCTLS30AN
    U21: KS74HCTLS163AN
    U22: KS74HCTLS163AN
    U23: +5V Regulator: 7805
    U34: KS74HCTLS174AN
    RF Modulator: Biwave, Inc. VM-445

 

Notes: I highlighted the interesting part numbers in red. U9 is not an AY-3-8930—the poster made an erroneous inference, the '8930' printed on the part is a date code, not a part number. U11 is an industry standard 24-pin ROM that contains the GROM data, not the special GI 28-pin GROM. I think you'll find that U17-U22 were used to implement the required fixes, but I don't have the time to verify that at the moment. Eliminating them would knock six IC packages off the parts count. All of the power supply circuitry is on this board—no second power supply board is required. This is basically the circuit that was slated to be introduced into Intellivision II as a running change.

 

WJI

 

[+Lathe26]

5 hours ago, Walter Ives said:

You misunderstand. It was the logic board a whole that was meant to be a compatible replacement. The STIC 1A chip was not a part-for-part replacement for the General Instrument AY-3-8900: it was more the centerpiece of a system-level integration done by APh that took advantage of more modern processes and used off-the-shelf memories. As you can see from the following picture posted elsewhere on this site, it was packaged in a 68-pin package. The board uses a 28-pin version of the CP1610 and doesn't use the GI System RAM, GROM or Sound Chip.

 

https://content.invisioncic.com/r322239/monthly_2022_04/20220404_140252.jpg.55ed9181ee1dbf5a5919ce66a03747d9.jpg

 

WJI

 

One bit in one STIC register originally was hardwired to be read as a 1.  In the STIC-1A, it reads as a 0.  Specifically, this was located at address 0x0022.  The original STIC returns 0x3FFF while STIC-1A returns 0x3FF7.  Do you know the purpose of this bit?  My 2 guesses are to indicate:

• "Hey, I'm a STIC-1A and should be compatible but be aware that timing might be slightly different"
• "Hey, I'm a STIC-1A and there is an enhanced mode if you write the correct magic value(s) to a register" (this would be similar to what GI did with the AY-3-8910 vs AY-3-8930 sound chips)

[Walter Ives]

6 hours ago, Lathe26 said:

One bit in one STIC register originally was hardwired to be read as a 1.  In the STIC-1A, it reads as a 0.  Specifically, this was located at address 0x0022.  The original STIC returns 0x3FFF while STIC-1A returns 0x3FF7.  Do you know the purpose of this bit?

"Hey, I'm a STIC-1A."

 

STIC 1A systems were spec'd to be functionally identical to STIC systems, with no enhancements. This was because no cartridge could take advantage of any improvements, for any such cartridge that relied on an improvement would be incompatible with the installed base. Enhancements were reserved for STIC 1B upgrade, which was being developed as a modification to STIC 1A and following about 4 months behind. So there was no motivation to sneak improvements into STIC 1A. Schedule was everything.

 

There was one exception: APh deliberately fixed the 159 pixel problem.

 

[Lathe26 obviously knows this, but for other readers: although the STIC nominally supports an active picture area 160 pixels wide by 192 pixels tall, the rightmost column of pixels is not displayed, so the actual displayed picture area is 159 x 192.]

 

When GI produced the first test run of original STIC chips, they found that their timing didn't work for some combination of scroll registers. STIC has time to make 28 GRAM/GROM fetches each scan line: 20 for the background cards, 8 for the moving objects. The moving object images are fetched during horizontal retrace. The exact timing changes according to the values in the horizontal scroll register. Memories are vague here, but the best recollection is that for some values of the scroll register the rightmost column of pixels was displayed as a vertical black line. The pressure to meet schedule was enormous and fixing the problem would have required completely reworking timing circuitry that had already been laid out, so Steve Maine implemented a quick-and-dirty work-around by not displaying the 160th pixel. If you want the authoritative story, Maine still lives, so buy him a beer and ask him what he remembers about it. Since APh designed the timing from scratch it worked as originally intended--they would have had to insert a kludge to break it. That offended their sense of elegancy, so they instead verified that the fix did not cause trouble with any existing cartridge.

 

The fact that the STIC 1A sound registers work differently is just a case of chips not working perfectly the first time. The problem would have been fixed in the next iteration, which was provided for in the schedule (the same one that would have fixed the DMA timing signals), but Mattel was awash in inventory of original parts and put development on ice. When Valeski ran out of his inventory of original chips a few years later he elected to forgo making the fixes and just used the part as is.

 

The fact that some of Valeski's STIC 1A-based INTV III systems allowed for 256 GRAM cards is due to the way the memories are wired. There's nothing in the original STIC chip that prevented it from accessing 256 GRAM cards—the 64 card limit is due entirely the fact that Mattel was only willing to pay for 512x8 GRAM on the original system, so the GROM chip only provided GRAM with 9 address lines. The chips used for STIC 1A's GRAM should have been wired to achieve the same result, but Valeski wanted extra GRAM for a private-label system he was making for a specific customer (TutorVision for World Book). When that project didn't work out, he used some of his board inventory to make Master Components. Because he was being cheap, institutional knowledge was spotty, and no one thought to disable the extra GRAM. He got lucky: the oversight only caused a problem with one cartridge then in circulation, Worm Whomper.

 

WJI

[BSRSteve]

5 hours ago, Walter Ives said:

Because he was being cheap, institutional knowledge was spotty, and no one thought to disable the extra GRAM. He got lucky: the oversight only caused a problem with one cartridge then in circulation, Worm Whomper.

Indeed, my emulator originally didn't bother to mask out the extra bits that weren't supported when addressing GRAM, and everything worked fine in Intellivision Lives!, but when we got the rights to Activision games for Intellivision Rocks, sure enough, Worm Whomper misbehaved. It was not immediately obvious what had caused the problem, but I don't recall it taking too long to figure it out and mask the bits off. I didn't realize until now that it had manifested itself way back in the eighties.

[+DZ-Jay]

On 6/9/2024 at 1:12 AM, BSRSteve said:

Indeed, my emulator originally didn't bother to mask out the extra bits that weren't supported when addressing GRAM, and everything worked fine in Intellivision Lives!, but when we got the rights to Activision games for Intellivision Rocks, sure enough, Worm Whomper misbehaved. It was not immediately obvious what had caused the problem, but I don't recall it taking too long to figure it out and mask the bits off. I didn't realize until now that it had manifested itself way back in the eighties.

How interesting!

 

Fun Fact:  Not masking the GRAM bits would break at least one home-brew game, although I believe there are others.  During the development of Pac-Man Christmas Carol, Joe Zbiciak suggested that there are four "extra bits" in the 13-bit BACKTAB word, including the two upper bits of the picture number when addressing GRAM, which were "ignored by the STIC."  I used these to track the state of consumable items within the logical game world map.

 

(I think this has been the primary source of graphical glitches in some recently created emulators when testing my game.)

I thought this was a very clever solution, but I did not think of it myself.  It can't recall right now if Joe discovered this technique on his own when creating his games, or if he actually observed it in practice when analyzing existing games, such as Worm a Whomper.

 

I believe Arnauld Chevallier and Carl Mueller Jr. were also aware of this technique, and probably used it as well.  It could have been discovered or observed by any of them, and then passed on to me as learned experience.

 

      dZ.

[SmakyTaky]

Thank you all for all the feedback... I would then understand I cannot use these ICs on a "regular" Intellevision motherboard but only on the System III ones? The slight differences may be the ones causing the "issues" I noticed when using these ICs on the motherboard I am trying to repair?

[mr_me]

2 hours ago, SmakyTaky said:

Thank you all for all the feedback... I would then understand I cannot use these ICs on a "regular" Intellevision motherboard but only on the System III ones? The slight differences may be the ones causing the "issues" I noticed when using these ICs on the motherboard I am trying to repair?

Looks like INTV System III with the 1987revA board revision.