Another curiosity - SIDE3/U1MB/SRAM 64k

[Piotr D. Kaczorowski]

@Beeblebrox,

 

Quote

Instructions should be checked and correct before an item is listed for sale as you'll appreciate.

 

Well... Correct.  

 

The project is based on my idea and laboratory. Besides, my name is on the board. However, I decided not to take any remuneration (in Poland, the board costs 29 PLN, which is 6.6 USD / 6.3 EUR / 5.4 GBP) because it's a very small project, and I wanted all the other extensions to work together (Ultimate 1MB, VBXE, Rapidus, SIDE 3.1). For now, I personally earn from the SAVO boards. You can see how the instructions for them are written:

http://ftp.pigwa.net/stuff/projects/SAVO/

 

My colleague, the producer, and my colleague, the UK distributor, quickly wrote the instructions for O2/Fixer. I will correct it when I have some free time.

 

 

[Piotr D. Kaczorowski]

5 minutes ago, woj said:

I did that, see above, but not with the O2/Fixer or U1MB hooked up to buffered O2 otherwise, there was no change. I can try that with O2/Fixer, but not today, I am at the edge of breaking something, over the last two days I probably have removed and reinstalled the DRAMs about 10 times, sooner or later I will break something...

Because DRAMs are not for reinstallation... they should be removed just once

[Beeblebrox]

6 minutes ago, woj said:

I did that, see above, but not with the O2/Fixer or U1MB hooked up to buffered O2 otherwise, there was no change. I can try that with O2/Fixer, but not today, I am at the edge of breaking something, over the last two days I probably have removed and reinstalled the DRAMs about 10 times, sooner or later I will break something...

Sure. Totally understand.

[Beeblebrox]

10 minutes ago, Piotr D. Kaczorowski said:

@Beeblebrox,

 

 

Well... Correct.  

 

The project is based on my idea and laboratory. Besides, my name is on the board. However, I decided not to take any remuneration (in Poland, the board costs 29 PLN, which is 6.6 USD / 6.3 EUR / 5.4 GBP) because it's a very small project, and I wanted all the other extensions to work together (Ultimate 1MB, VBXE, Rapidus, SIDE 3.1). For now, I personally earn from the SAVO boards. You can see how the instructions for them are written:

http://ftp.pigwa.net/stuff/projects/SAVO/

 

My colleague, the producer, and my colleague, the UK distributor, quickly wrote the instructions for O2/Fixer. I will correct it when I have some free time.

 

 

Sure. So do I have to worry about anything in my friends 800xl given I've only just followed the current retrolemon.co.uk instructions for the o2 fixer installation?

 

As I say it was rock solid and had no issues so I am hoping not doing the leg bending step you alluded to isn't going to be an issue. 

 

My friend lives abroad and isn't technical, so it's not something he can do himself if something is needed doing. 

Edited September 29, 2023 by Beeblebrox

[Piotr D. Kaczorowski]

5 minutes ago, Beeblebrox said:

As I say it was rock solid and had no issues so I am hoping not doing the leg bending step you alluded to isn't going to be an issue. 

 

So what did you actually do? Did you connect the signals from O2/Fixer to the Atari ICs (Pokey/PIA/GTIA/Antic) without bending those O2/BO2 legs, right?

Edited September 29, 2023 by Piotr D. Kaczorowski

[Beeblebrox]

5 minutes ago, Piotr D. Kaczorowski said:

 

So what did you actually do? Did you connect the signals from O2/Fixer to the Atari ICs (Pokey/PIA/GTIA/Antic) without bending those O2/BO2 legs, right?

I did exactly as the instructions in retrolemon.co.uk site said. 

 

Pin 12 of the 02 fixer was hooked up by a 10mm wire to the o2 pad on the o2 fixer. Phi2 connection for u1mb soldered to pad 1 of the o2 fixer. That was it. 

 

I've not hooked up anything other than phi2 for the U1MB to the o2 fixer. 

 

Then the usual connections between the 800xl's pcb, U1MB and VBXE. These were in place before I installed the o2 fixer. 

 

The system was fully tested and rock solid and no issues running VBXE, U1MB, Side3.1. So hopefully OK as is. 

 

No leg bending at all at any stage as instruction make no mention of this.

 

 

 

 

[Piotr D. Kaczorowski]

@Beeblebrox,

 

If it works well, there's no need to complicate things further. If it doesn't work, you need to successively bend the legs and connect them to OUT_2... OUT_4 until it works.

[woj]

37 minutes ago, Beeblebrox said:

No leg bending at all at any stage as instruction make no mention of this.

Yes, it does:

 

• BO2 signals for main chips (Pokey, GTIA, PIA) - make sure that connection is made to the pin angled and not connected to the motherboard

Reading this I realized I could also try things without my stereo board. And I can try bending some legs without soldering things, just use a connector.

Edited September 29, 2023 by woj

[+mytek]

1 hour ago, woj said:

BO2 signals for main chips (Pokey, GTIA, PIA) - make sure that connection is made to the pin angled and not connected to the motherboard

Why?  If the 02 Fixer is taking the place of the original 74LS08 then it already has the buffered 02 going to the appropriate places via the mother board traces. Any of the VSLI that isn't connected to buffered 02 never needed it. And as far as I've seen in my upgrade installations aside from changing out the 74LS08 for a 74F08, and perhaps the CPU if it's from Mexico (and only if it's problematic), all the original buffered 02 connections are fine.

 

Anyway I'd initially try it as a drop-in, and leave the stock Atari chips alone. Just make sure to have the U1MB getting its 02 from one of the extra buffered outputs on the 02 Fixer (there really should have been a header for those instead of solder pads - I never liked needing a soldering iron to uninstall or swap something).

[woj]

8 hours ago, mytek said:

Why?

Something about interferences that Piotr seems to know more about, and additional capacitors on these lines of the O2/Fixer.

 

As said, I might play a bit more with this, but I think in my particular case the problem lies elsewhere. The computer was sold to me untested last year, it had a history of playing up on me, in particular the factory RAM self test when the OS was ran off the original Atari rom was indicating bad RAM, not so with a self reburned eprom chip with the same exact OS, or even OS rom off my other 130XE machine. And that was as stock configuration as one gets, and already then this was an indication of some timing issues to me. 

[Piotr D. Kaczorowski]

@mytek,

 

Quote

If the 02 Fixer is taking the place of the original 74LS08 then it already has the buffered 02

 

74LS08 (or 74HCT08) has 4 AND gates.  O2/Fixer has 8 AND gates:   4 AND gates that works like 74xx08 and extra 4 AND gates, that are configured as "twin BO2 signals" in set of OUT_1 - CMOS like BO2 and OUT_2...OUT_4 - NMOS like BO2.

 

If you intend to use the twin BO2 signals, they should be connected to the top of the Ultimate 1MB system or to the bent legs of the Atari circuits so that they are not shorted with the main board.  

 

If someone asks about the latter, the answer is: that was my idea and it has been proven in the lab that it works. I had the opportunity to test the so-called reverse system degradation (or gradual system stability improvement) on a machine with SRAM, U1MB, VBXE, Rapidus + SIDE3.1.

 

The board that had the above components had a built-in 74LS08 circuit. SIDE 3.1 completely froze the computer upon startup.

1. After replacing the 74LS08 with 74F08, the computer started with SIDE 3.1 in 6502 mode, but did not start in 65C816 mode.
2. After replacing the 74F08 with the O2/Fixer prototype (containing 74F08 and a second 74F08 generating four twin BO2 signals) and connecting the PIA to the first twin BO2 signal, the computer started in 65C816 mode, but SIDE3.1 froze the computer upon entering the loader via Ultimate 1MB.
3. After switching the Pokey to the next twin BO2, the computer worked in the SIDE 3.1 loader, but the cartridge had trouble handling the SD card (it was still disturbed).
4. After connecting the GTIA to the next twin BO2, the computer worked correctly in 65C816 mode with SIDE3.1 reading SD cards.

A similar effect can be achieved by replacing the PIA, Pokey, and GTIA circuits. The PIA can be replaced with the WDC PIA NMOS (W65C21N), which has limited clock line capacity expansion due to newer technology. The same is true when changing Pokey to PokeyMax and GTIA to Sophia 2.

In the case of O2/Fixer, technology was also used that significantly reduces the capacity introduced on the clock line. Additionally, through additional capacitors, the NMOS characteristic has been preserved.

 

 

[+mytek]

Presently on an 800XL Antic, GTIA, Pokey, PIA, and the Cart all derive their Phi2 as buffered by a single gate in a 74LS08 chip. This works, and only has trouble when adding additional devices that also require Buffered Phi2 (B02). So in my opinion and for what it's worth, if only the 'added' devices were to be connected to the extra B02 outputs on the 02 Fixer it would seem sufficient. And if one of those added devices is a Multi-Cart and problems are seen with it plugged into the system, then perhaps it would be beneficial and logical to provide it with it's own independent B02 output as well, although that would require cutting a trace on the mother board.

 

However with all that said having a single strong and fast B02 clock provided by the substitution of a 74F04 for the original 74LS08 has worked wonders in systems with a U1MB in combination with a Multi-Cart (Side 1-3, AVG, Ultimate, ect.). Although certainly there's nothing wrong with providing additional independent B02 outputs such as your 02 Fixer is doing, but if simple chip substitution works, that probably makes more sense than bending up pins and making separate connections, and it would also seem logical to try that simple substitution fix first.

 

Obviously there have been separate issues with Rapidus being added to the mix, so probably anything that would provide the cleanest clock signal to it would be called for, and that might indeed be the 02 Fixer.

[Piotr D. Kaczorowski]

@mytek

 

There are many solutions, just like in IT. It turned out the way it did and it works. It provides additional possibilities. I think the project is so simple that there's no need to consider additional improvements.

 

My earlier concept was to create a "Sally-to-Sally" adapter. It would be similar to the "6502-to-Sally" adapter, but the inserted processor would still be the 6502C Sally. The main task of the board would be to disconnect pin #39 and introduce its own O2 signal there, generated similarly to the Atari 400/800 (or the Sweet16 65C816 project) externally from the Phi0 signal using a 1-gate. This would result in a strong O2 signal being fed to the board. Of course, tests would need to be conducted again. From what I've seen, Lotharek plans to release a board that will contain such a circuit along with the WDC65C02S processor in the QFP version. It's possible that he will also release such boards with the 65C816 processor, as well as versions with SRAM included."

 

Quote

However with all that said having a single strong and fast B02 clock provided by the substitution of a 74F04 for the original 74LS08 has worked wonders in systems with a U1MB

 

I think the "F" series didn't change the conditions due to speed or signal strength, but rather due to the lack of introducing additional parasitic capacitance - which is why it could also operate quickly. The "F" series operates up to 125MHz according to the specifications.

[+mytek]

6 hours ago, Piotr D. Kaczorowski said:

I think the "F" series didn't change the conditions due to speed or signal strength, but rather due to the lack of introducing additional parasitic capacitance - which is why it could also operate quickly. The "F" series operates up to 125MHz according to the specifications.

You keep mentioning parasitic capacitance which is primarily an external effect due to conductors such as wires, PCB tracks, and inductors. It can also be an issue in MOS transistors. However TTL by it's very nature has extremely low parasitic capacitance. Perhaps what you are really thinking of is often referred to as interdevice capacitance, which is lower in the 74F08 vs. the 74LS08.

 

The biggest difference between 74LS vs. 74F is that the switching propagation delay of the 74F has been reduced by a factor of 3 (it's much faster), and its fan-out is nearing twice the load capability (making it a stronger buffer). These are the points I make of the advantages that come with switching from the original 74LS08 over to the 74F08. The 74AS08 would be even stronger in both regards, although sometimes that comes with a disadvantage - over and under shoot (e.g., switching noise). I think the 74F08 is a good compromise.

 

Here's some additional information...

 

 

EDIT: In most scenarios simply swapping out the 74LS08 for a 74F08 will fix operational reliability of the Atari 8-Bit Computers. And in certain cases when this isn't enough (e.g. Rapidus), then using the 02 Fixer board instead is the reasonable path to take. Either approach can be applied in a routine basis without system detriment.

[Piotr D. Kaczorowski]

@mytek,

 

Ok. I wouldn't want anyone to think that I'm advertising O2/Fixer because I don't financially benefit from it. I understand the issue with 74F08 and that it's a first aid solution. That's indeed the case. However, for more complex issues, O2/Fixer helps. Another topic is that most people connect the Ultimate 1MB to the O2 signal - for example, to the SRAM 64K V2/V3/V4 board. So, this is a simplification, because U1MB, like everything else, should be connected to the BO2 signal, and here O2/Fixer provides the most appropriate signal with CMOS characteristics (more rectangular) than NMOS. Essentially, my intention was for the price of O2/Fixer to be set in such a way that it would essentially be an addition to other extensions, which wouldn't significantly increase costs.

 

74F = FAST => F.airchild A.dvanced S.chottky T.TL  and is up to 125Mhz

 

I believe that not increasing the parasitic capacity is a technological element that makes the "F" work so fast.

 

An additional story is that when starting experiments, I accidentally connected an oscilloscope with a probe set to a 1x multiplier and connected it to pin #39 of Sally. SIDE 3.1 stopped working. Thus, the capacity of O2/BO2 and its impact on other devices can be easily checked. I know that a few years ago, other fixers were developed in place of 74LS08, which were designed by Simius. From what I know, they were based on PLL and signal reproduction. However, I also know that this system did not help with all the problems. I hope that the ability to separate between systems and feed them cloned twin signals will improve stability for most users. I conducted tests on about 15 Atari boards at home. Lotharek did on a few at his place.

 

The project, in its conceptual phase, was consulted with Simius, XAngel (Polish Atari hardware hacker), and AcidMaker (SIO2SD / Medusa Upscaler). Ultimately, it was finalized and is now being produced by Lotharek, which gives it a scale of availability.

[woj]

Another update from my side, still no go. What I tried (with the SRAM board and O2/Fixer):

 

- replacing Sophia with factory GTIA,

- removing the stereo board and sticking in just one Pokey, trying both of the them separately

- connecting the one Pokey O2 leg to BO2 output of O2/Fixer

- connecting the O2 leg of PIA to BO2 output of O2/Fixer

- connecting U1MB O2 to one of the capped outputs of O2/Fixer

 

Not only it still does not work, there is no indication whatsoever that something changes / gets better / gets worse.

 

Putting aside that it might be something unrelated to O2 buffering, one last possibility to try would be to feed the BO2 from O2/Fixer to the cartridge slot, I am, however, not going to cut traces on the motherboard, it suffered enough from my inadvertent incompetence already. 

 

Perhaps, I have the SysCheck2 and it has the cartridge slot, I could connect SIDE3 through there, I only do not remember if one can configure SysCheck to totally neutral setting, as in not being in the diagnostic mode and not doing any memory extensions not to interfere with U1MB.

[_The Doctor__]

U1M does not want the capped o2 output.

Pulling a fixer cap and then feed your S Ram from that pad and make sure you are not back feed o2 through the sram (which if the machine already worked without it and you have ram on the u1m (one of the biggest reason to have it)) I have no idea why you want sram...

keep all the others the way they were.

or use it the way that worked and was suggested where none of the leg bending was done.

 

You are effectively trying to use three different RAM change/replacements at once, SRAM/U1M/SYSCHECK. That's just not making sense to me. Effectively you will be turning the memory off on the SYSCHECK, the SRAM is just spinning it's wheels, and U1M is providing all your memory. Any overlap, accidental or otherwise will lead to contention issues.

 

Edited September 30, 2023 by _The Doctor__

[Peri Noid]

@woj Yes, you can make Sys-Check not do anything. If I'm correct, it will then buffer cartridge slot lines. Maybe it will help.

 

@mytek What @Piotr D. Kaczorowski nowhere mentions treating it as obvious for some unknown reason, is that O2 fixer gives 4 independent BO2 outputs so that each device connected to it gets it's own copy of this signal. The more devices are connected to thie same line, the higher overall capacity and thus more problems with signal propagation, high level can be delayed and so on. When all the outputs are separated, this effect is reduced to the minimum.

 

There's one thing about O2 fixer that I don't understand - if the input pin is selected by connecting one of the input pins to the pad in the middle, how is the MAIN output pin selected? I mean, the one that outputs the signal to the mainboard?

Edited September 30, 2023 by Peri Noid

[_The Doctor__]

There is always the chance the o2 fixer has a component that isn't up to snuff in your case (it happens) in which case a different o2 fixer might be in order

[woj]

14 minutes ago, _The Doctor__ said:

I have no idea why you want sram...

I don't really, I got with all the other goodies just in case, as the system did show some problems in the past. Just in case I'd have problems like I do, but with the factory DRAM, not the SRAM. I just wanted to check it would work as expected, and it doesn't. You know, I just had to... 

[woj]

15 minutes ago, Peri Noid said:

There's one thing about O2 fixer that I don't understand - if the input pin is selected by connecting one of the input pins to the pad in the middle, how is the MAIN output pin selected? I mean, the one that outputs the signal to the mainboard?

It has the four gates, just independent, like the factory 74LS08, on the same pins, whatever it gets on each of the input pins, it buffers it on the corresponding output pin. But these 4 gates are connected differently on different motherboards, so O2 might be on gate 1, 2, 3, or 4, depending which machine you put it in. To get the O2 out of the board on additional connections you need put the wire to the input pin your current machine uses for O2.

 

Even to me it is so obvious that I am afraid this is not what you are asking about  Put it differently - you get 5 BO2 outputs, 4 on the side of the board, and 5th on the pin next to the input one. 

[Peri Noid]

4 minutes ago, woj said:

Even to me it is so obvious that I am afraid this is not what you are asking about  Put it differently - you get 5 BO2 outputs, 4 on the side of the board, and 5th on the pin next to the input one. 

Yes, sounds reasonable.

Edited September 30, 2023 by Peri Noid

[+mytek]

1 hour ago, Peri Noid said:

@mytek What @Piotr D. Kaczorowski nowhere mentions treating it as obvious for some unknown reason, is that O2 fixer gives 4 independent BO2 outputs so that each device connected to it gets it's own copy of this signal.

Yes I understood that. And I can see how it can off load some of what the stock B02 feeds presently, that's all good. However in my way of thinking the "Keep it Simple' principle comes into play. And the 74F08 swap almost doubles the buffering capacity all on its own.

1. Get stock Computer running reliably without any upgrades.
2. Then add upgrades one at a time (SRAM, U1MB, ?), until something breaks.
3. Then swap out a 74F08 for the original 74LS08 (or install the 02 Fixer if you don't have a 74F08 on hand).
4. If that fixes things, continue adding any additional upgrades.
5. If it breaks again, then swap out 74F08 for 02 Fixer (if you haven't already done so) and feed the last upgrade that caused a failure with one of the four independent B02 outputs.
6. If that doesn't fix it, then one by one use the remaining B02 outputs to feed any of the other upgrades that require it.

The idea is to systematically go through the steps and you should end up at a successful outcome. I doubt that it will require lifting pins on Antic and GTIA and then feeding them with independent B02 from the 02 Fixer. But I've learned to never say never.

[Piotr D. Kaczorowski]

@Peri Noid,

 

Quote

There's one thing about O2 fixer that I don't understand - if the input pin is selected by connecting one of the input pins to the pad in the middle, how is the MAIN output pin selected? I mean, the one that outputs the signal to the mainboard?

 

The first four AND gates are the same as in the 74xx08 circuit - according to the schematic. So it depends on the Atari computer where BO2 enters the mainboard. For example, in Atari XE, the original BO2 enters the mainboard from pin #3. I would like to point out that I deliberately call the additional BO2 signals "twin" and not "clone" or "copy". After connecting the O2 pin to the soldering point in the center, all BO2 signals are generated simultaneously, with the same propagation time. The only difference is the so-called twin BO2 OUT_1, which does not have an additional capacitor (has CMOS characteristics) and is dedicated to Ultimate 1MB.

 

[Piotr D. Kaczorowski]

@mytek,

 

Quote

And the 74F08 swap almost doubles the buffering capacity all on its own.

 

And according to my experiment, there was still a shortage of 4xTTL. As I said, after replacing PIA, Pokey, GTIA, and removing Freddie (installation with VBXE 3.5Mhz + SRAM, without Freddie), everything started to work. The laboratory version of O2/Fixer was built on two 74F08 ICs.

 

It should be noted that this computer had all the extensions, including Rapidus.

Edited September 30, 2023 by Piotr D. Kaczorowski