Another curiosity - SIDE3/U1MB/SRAM 64k

[woj]

Keeping in mind my last failure to properly diagnose my stereo problem I will try to be more careful in making judgments I just had a very enjoyable evening upgrading my 800XL with U1MB, VBXE, Spectre, and SRAM 64k module, all freshly arrived from Lotharek just today. It went quick because the computer was already taken apart and cabling was mostly prepped. The system ended up in a working state, U1MB firmware got updated (but through SIO, see below), and I was also checking it with every step on the way. But not with SIDE3. After all the mods were in, I plugged in SIDE3 and it just plain crashed, I got a garbage screen with some traces of the loader. Played a lot with U1MB settings, read this and that, updated U1MB firmware through SIO, still the same thing. Oh, and this is SIDE 3.0, not 3.1. What was crashing was just the loader, I could make the SDX boot off the SIDE3, and I could even boot SDX off U1MB and access APT drive from the SD card in the SIDE3. So all worked, besides the loader.

 

Now, I thought OK, I have one of those "sensitive" machines and I will have to see about that O2 fixer I also got as a backup, and if all else fails I will check with @flashjazzcat for further advice. But then it occurred to me, perhaps it is the 64K SRAM module, so I removed it, plugged in the RAMs back along with the delay line chip, and presto! SIDE3 got back to life! Since I started upgrading with U1MB, at no point I had a computer with just the SRAM module in, and I wonder if that is just the module breaking it, or the whole combination of all extensions / devices. At some point I also replaced the 74F08 with the O2fixer (just as it comes), but I forgot to check with 74LS08, perhaps that could be it.

 

So, anyone else had similar experience? 

 

(A side rant, the SRAM 64k module sitting in a socket makes the whole contraption way too high to clear the keyboard, I'd have to remove the socket and solder the module directly into the PCB, somehow not something I want or I am willing to do, especially not now seeing it produces more problems than it solves 😕).

Edited September 27, 2023 by woj
Spelling

[Peri Noid]

It's the matter of all the bricks. The problems are timings, line capacities and other very sensitive and sometime unpredictable stuff. The more hardware you put inside, the more problems you may encounter.

 

(SRAM, when properly mounted, clears with keyboard. Notice, that it's delivered IN a socket, which you can remove)

Edited September 28, 2023 by Peri Noid

[woj]

3 hours ago, Peri Noid said:

It's the matter of all the bricks. The problems are timings, line capacities and other very sensitive and sometime unpredictable stuff. The more hardware you put inside, the more problems you may encounter.

 

(SRAM, when properly mounted, clears with keyboard. Notice, that it's delivered IN a socket, which you can remove)

True about the socket, I missed that, in my defense it was getting late 😕 But that is another reason why I wrote it here, so a good soul like you would put me back on the right track, thanks!

 

Still did not get the module to work with SIDE3, tried now with 74LS08, still the same. I would like to know what is going on, but then I do not want to spend ages diagnosing this... Perhaps swapping the main chips from my other machine would give me some info. Will give it a could of more tries and then give it a rest.

 

In the meantime another problem occured, the stereo board hardly clears the keyboard, not with the folded keyboard cable in between the two 🤔 I can fold it one more time to move to the side of the board, but this is asking for trouble. To do it properly I'd need to solder up another board on low profile pin stripes, not even sure where to get those.

[Peri Noid]

From my experience, very subtle factors may influence stability. For instance the points in a computer, from where the signals for U1MB are taken. You don't have Rapidus in your machine, this is a challenge (can be done, though).

If you have one, use a 74F08 instead of an LS chip. It may also be good to try a CPU from a different producer, than what you currently have. Especially if it's a one made in Mexico - they are known to play tricks in expanded configs.

Edited September 28, 2023 by Peri Noid

[_The Doctor__]

try a memory upgrade that still uses the delay line. Or one that simulates it, handling signals/mmu/emmu/banking on it's own

Edited September 28, 2023 by _The Doctor__

[woj]

2 hours ago, _The Doctor__ said:

try a memory upgrade that still uses the delay line. Or one that simulates it, handling signals/mmu/emmu/banking on it's own

You see, the thing is, there is nothing wrong with my base memory or delay line, I got the SRAM module mostly out of curiosity (and the real prospective reason that would involve a lot of soldering is too complicated to picture here). 

 

I did try all 74x08 chips I have, I swapped GTIA/Sophia, ANTIC, and CPU from the other machine. I got the impression that it crashed a bit differently, but it still crashed. And yes, it is some interaction with U1MB, I removed that too for test, and SIDE3 was perfectly fine with the SRAM module then. I am giving up, I reinstalled "factory" RAM, sorted out and cleaned up the wiring, everything works perfect. I even managed to route the keyboard cable so that the stereo board clears the keyboard. I am not closing it up just yet, there is one more thing coming up, but for that some PCBs from JLC need to arrive, and I will show it off in my other topic  

 

And a picture to put some visual substance to my endeavors.

[Peri Noid]

My guess is, the CPU adapter you're using to get U1MB signals is what causes problems. I'd suggest taking the signals from the board as adviced by Candle. Especially the O2 (Phi2) signal.

[woj]

With SRAM module I use the pads provided there, but indeed this would be fully equivalent to using the adapter (which works just fine with the factory RAM BTW) Yes, I also know that the Phi2 on the picture is buffered? or whatever it's called, it's not the same as at the CPU leg.

 

Now don't laugh, for the very first test I had the U1MB hooked up like on the picture, in the process of going forward and tidying up I decided to hook up to the SRAM module directly (assuming the designers knew what they were doing). But all that was long before I got to trying SIDE3. For the sake of science, I will try it again with Candle advised hook up points, tomorrow I hope.

[Beeblebrox]

Also try an o2 fixer for the phi2

 

https://retrolemon.co.uk/atari-8bit-upgrades/219-o2-fixer.html

 

 

[Peri Noid]

Just remember that with this O2 fixer, the signals must still be acquired either from it or from the board but not from the CPU directly. 

[Beeblebrox]

Yup. It takes it from cpu pin 39 via the original 74ls08 socket. So on an 800xl for example it's pin 12 of the fixer that is a connection to pin 39 of the cpu. 

 

For XE machines it's pin 1, where the o2 pad on the fixer is connected to the fixers pin 1 (for XE's)  as illustrated below, and then the phi2 connection for u1mb is connected to one of the 4 solder pads on the fixer. 

 

 

Edited September 28, 2023 by Beeblebrox

[+mytek]

8 hours ago, woj said:

I did try all 74x08 chips I have, I swapped GTIA/Sophia, ANTIC, and CPU from the other machine.

Well the whole point of substituting a 74F08 for the original 74LS08 is to make the Phi2 buffer (one of the '08' gates) faster. So with you taking Phi2 directly from the CPU you circumvented the buffer for the U1MB, and no matter how many variations of the '08' chip you try that will always remain the same. I believe all the problems you are having stem from that.

 

Edit: BTW pin-29 of the Antic chip is a reliable place to pick up the buffered Phi2 signal in any Atari 8-bit computer.

[woj]

18 minutes ago, mytek said:

Well the whole point of substituting a 74F08 for the original 74LS08 is to make the Phi2 buffer (one of the '08' gates) faster. So with you taking Phi2 directly from the CPU you circumvented the buffer for the U1MB, and no matter how many variations of the '08' chip you try that will always remain the same. I believe all the problems you are having stem from that.

 

Edit: BTW pin-29 of the Antic chip is a reliable place to pick up the buffered Phi2 signal in any Atari 8-bit computer.

Thanks! Yes, I see the complete picture now. Since O2 fixer came in the box with the goodies I may as well try to make proper use of that, will report back.

[Beeblebrox]

1 hour ago, woj said:

Thanks! Yes, I see the complete picture now. Since O2 fixer came in the box with the goodies I may as well try to make proper use of that, will report back.

It's definitely worth a try if you have an o2 fixer already. 

 

I popped one in an 800xl with an U1MB and VBXE recently. Fingers crossed but rock solid stability as is and with my side3.1 cart.

[Piotr D. Kaczorowski]

@woj,

 

1. O2/Fixer should fix that problem. O2/Fixer is made of 1-gates and capacitors that simulate the NMOS characteristic with the smallest possible increase in parasitic capacitance on the BO2 clock line.

 

2. The answer to the question of whether SRAM affects the O2/BO2 characteristic is YES. This is due to several factors. Firstly, on the SRAM board, the O2 clock signal is used to control the SRAM memory, where it passes through a 1-gate and is supported by a pull-up resistor (on different boards these values range from 2.2k to 10k). The 6502 processor specification speaks of an O2 load with a maximum of one TTL connection and a maximum capacitance of 135pF (information from a book I have). Therefore, it should be remembered that in the case of the SRAM board, the O2 signal reaches the 1-gate on the SRAM board and to the 74LS08 or 74HCT08 gate, which are factory-installed in Atari XL/XE. In the case of NMOS technology, disturbances at low signal throw are subject to interference, and further on, this disturbance can translate into powering another system that receives the signal and transferring the disturbance further. This disturbance can also be frequency filtering or frequency filter generation (this can interfere with the operation of SIDE3/3.1). Using O2/Fixer makes it resistant to disturbance transfer and capacitance loading of the BO2 line. Additionally, you have at your disposal four more twin BO2 signals, one of which has a characteristic well seen by Ultimate 1MB.

In the case of a large number of internal extensions and stability issues, it is also worth connecting the BO2 signals of subsequent Atari systems from the top - not through the mainboard, but directly to the twin signals. This will cause the separation of BO2 between systems, as well as the reduction of parasitic capacitance.

 

In the end, it's worth noting that the wires between subsequent extensions should be as short as possible. Long and twisted wires act as antennas and generate as well as receive interference.

As for SRAM, I always use it in my installations because, after all, it is a single memory in new technology that also does not require refreshing, so it generates significantly fewer disturbances. On XE boards with VBXE, you can completely get rid of the Freddie system, which also reduces interference. So there are more pros than cons.

[woj]

47 minutes ago, Piotr D. Kaczorowski said:

In the case of a large number of internal extensions and stability issues, it is also worth connecting the BO2 signals of subsequent Atari systems from the top - not through the mainboard, but directly to the twin signals.

Is my understanding correct that in this case one needs to cut he motherboard traces?

Edited September 29, 2023 by woj

[Piotr D. Kaczorowski]

2 minutes ago, woj said:

Is my understanding correct that in this case one needs to cut he motherboard traces?

Absolutely not. I simply insert the chip into the socket. Then I bend one leg, which corresponds to the O2(BO2) signal, and connect that leg with a Kynar wire to the O2/Fixer.

[woj]

32 minutes ago, Piotr D. Kaczorowski said:

Absolutely not. I simply insert the chip into the socket. Then I bend one leg, which corresponds to the O2(BO2) signal, and connect that leg with a Kynar wire to the O2/Fixer.

Indeed, I even read about this yesterday on the page with O2/Fixer instructions, just totally forgot, too many things in my head ATM.

[Beeblebrox]

@Piotr D. Kaczorowski

 

Just to check, for the o2 installation there is no mention on the retrolemon.co.uk page instructions about any requirements to bend chip legs. 

 

See here below. Can you comment? I didn't do any chip leg bending with my o2 chip install in the 800xl with u1mb and VBXE I mentioned and all is working. 

 

• Description
•  
• Product Details

O2 fixer is a replacement for 74LS08 chip in all Atari XL/XE machines. It provides four additional BO2 signals that provide option for separation Atari chips on the BO2 (Buffered O2) clock bus. It also lowers the parasitic capacitance on the BO2 path on the motherboard. Idea by Piotr D. Kaczorowski, design/production by Lotharek.

 

 

 

Installation Manual

Photos below are made with Atari 65XE motherboard. For other models and motherboard versions steps are the same, but the location and appearance might be dfferent.

 

1. Locate 74LS08 chip on the motherboard. If it is socketed, skip point 2 below.

   

    

2. Desolder the chip and solder back a 14 pin DIL socket (not supplied).

   

    

3. Insert O2 Fixer into the socket. Make sure that it is oriented properly.  4 soldering pads at the PCB should be next to pin 1. 

   

    

4. Atari should start normally at this stage. Test it!

5. Locate O2 signal from the CPU on O2 Fixer: with the multimeter check for the continuity from pin 39 on the CPU to subsequent pins of the fixer. For most XE machines and 1200XL it is pin #1, for Atari 600XL it is pin #5, for 800XL - pin #12, but for 800XLF, it is pin #2.

6. Connect the located pin with a wire or a bit of solder to the middle pad called O2_IN.

   

 

From now on, the soldering pads marked OUT_1, OUT_2, OUT_3 and OUT_4 will provide 4 buffered and noise free BO2 signal outputs. You can use them for:

• BO2 signals for main chips (Pokey, GTIA, PIA) - make sure that connection is made to the pin angled and not connected to the motherboard
• Ultimate 1MB connection
• cart slot O2 signal replacement - this is most suggested use, resolves problems with fast cartridges like SIDE 3.1

 

Additional Info

• 3 of 4 outputs have been connected to GND with 680pF capacitors - look at the bottom side of the PCB. Ultimate 1MB needs no capacitor - please use soldering pad marked OUT_1 for U1MB.
• if you use "capped" output (OUT_2, OUT_3 or OUT_4) and Atari behaves odd or crashes, please remove corresponding capacitor for the given OUT soldering pad. Installing O2 Fixer in the socket allows quick removal of the capacitor when required.

[Piotr D. Kaczorowski]

@Beeblebrox,

 

 

The original O2/Fixer prototype was built using two 74F08N chips. As you can see, the Pokey, PIA, and GTIA are connected via twin BO2 signals provided by the second 74F08. The current O2/Fixer is constructed with eight 1-gate chips and capacitors that emulate NMOS characteristics.

 

The first step is to replace the 74LS08/74HCT08 with the O2/Fixer. If you have the Ultimate 1MB, the second step is to connect the U1MB O2 pin to the O2/Fixer U1MB twin BO2 (OUT_1 - you can check that there is no capacitor on the other side of the board). The third step is to rewire the Atari circuits to the twin BO2 signals (OUT_2 ... OUT_3).

 

When I have a moment, I'll check the instructions on the RetroLemon website. I consulted on this description, but it was done very quickly, so there might be something that still needs to be added. To be honest, it's still better than the first version of this description

 

 

 

 

 

[woj]

I hate it to break it to you guys, but it still does not work when the SRAM module is connected, what I tried so far with connecting O2 to U1MB was with 74F08 and hooking up to BO2, and with O2/Fixer and O2 line hooked up to its first output. Both configurations crash the same way (see the picture) as with it being connected directly to the unbuffered O2 from the CPU. I tried both with the delay line in and out of the motherboard (that shouldn't matter, but I just had to check). It is of course possible that I do something totally wrong elsewhere, it happens more often than I'd like to admit, but the matter of fact is that putting DRAMs back in and eliminating SRAM module makes the whole thing work just fine.

 

I am throwing in the towel, starting to lift legs of chips and solder wires to them to connect to O2/Fixer is beyond what I am willing to do ATM, I would do that if saw any kind of improvement with O2/Fixer and some hope, but I see none. Especially that I do have a working system without the SRAM module. It will be safely stored for perhaps some future machine yet to be owned (I am on a lookout for a chipish to save PAL 600XL). 

[Piotr D. Kaczorowski]

Interesting... This weekend I will be disassembling and reassembling the Atari 800XL with U1MB and Rapidus. Inside, there's a PokeyMax and Sophia 2. Unfortunately, SIDE 3 doesn't work. We'll see how it goes this time.

[Beeblebrox]

5 hours ago, Piotr D. Kaczorowski said:

@Beeblebrox,

 

 

The original O2/Fixer prototype was built using two 74F08N chips. As you can see, the Pokey, PIA, and GTIA are connected via twin BO2 signals provided by the second 74F08. The current O2/Fixer is constructed with eight 1-gate chips and capacitors that emulate NMOS characteristics.

 

The first step is to replace the 74LS08/74HCT08 with the O2/Fixer. If you have the Ultimate 1MB, the second step is to connect the U1MB O2 pin to the O2/Fixer U1MB twin BO2 (OUT_1 - you can check that there is no capacitor on the other side of the board). The third step is to rewire the Atari circuits to the twin BO2 signals (OUT_2 ... OUT_3).

 

When I have a moment, I'll check the instructions on the RetroLemon website. I consulted on this description, but it was done very quickly, so there might be something that still needs to be added. To be honest, it's still better than the first version of this description

 

 

 

 

 

If you could check that would be great. I am slightly concerned as I followed the retrolemon.co.uk instructions and no mention of bending any legs. 

 

The u1mb/VBXE machine I installed this o2 fixer in for a friend has been posted to him. It was fully working including tested with my side3.1 so not reason to think therw will be any issue, but if you are saying the retrolemon.co.uk instructions are missing a step I am slightly concerned. Instructions should be checked and correct before an item is listed for sale as you'll appreciate.

[Beeblebrox]

1 hour ago, woj said:

I hate it to break it to you guys, but it still does not work when the SRAM module is connected, what I tried so far with connecting O2 to U1MB was with 74F08 and hooking up to BO2, and with O2/Fixer and O2 line hooked up to its first output. Both configurations crash the same way (see the picture) as with it being connected directly to the unbuffered O2 from the CPU. I tried both with the delay line in and out of the motherboard (that shouldn't matter, but I just had to check). It is of course possible that I do something totally wrong elsewhere, it happens more often than I'd like to admit, but the matter of fact is that putting DRAMs back in and eliminating SRAM module makes the whole thing work just fine.

 

I am throwing in the towel, starting to lift legs of chips and solder wires to them to connect to O2/Fixer is beyond what I am willing to do ATM, I would do that if saw any kind of improvement with O2/Fixer and some hope, but I see none. Especially that I do have a working system without the SRAM module. It will be safely stored for perhaps some future machine yet to be owned (I am on a lookout for a chipish to save PAL 600XL). 

@woj just out of interest try putting another gtia chip from another machine in there with the sram installed in the sys. Indulge me.

 

[woj]

6 minutes ago, Beeblebrox said:

@woj just out of interest try putting another gtia chip from another machine in there with the sram installed in the sys. Indulge me.

 

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...