Nemesis final test issue

[catmando]

I've recently installed a Nemesis inside my Falcon.

I followed all the steps in the manual and DML's diagram http://www.atari-wiki.com/index.php/Nemesis.

At every test stage I had no problems until the very last test.

As soon as I run NEM_HI.TOS upon return to desktop the desktop freezes immediately.

NEM_LO.TOS works no problem and NEMBENCH reports 125% for all tests.

I am using coax in place of the brown and yellow wires, I haven't connected the red or orange wire yet.

Any ideas appreciated

Thanks

[exxos]

Wasn't there a 20mhz and 24mhz switch ? I assume thats the TOS programs you quote ?

FWIW, I never got 24mhz working on my falcon, at least not stable. In the end I just took it out totally as was hearing a lot of "falcon deaths" back in the day, so didn't want to risk killing mine

[catmando]

Wasn't there a 20mhz and 24mhz switch ? I assume thats the TOS programs you quote ?

FWIW, I never got 24mhz working on my falcon, at least not stable. In the end I just took it out totally as was hearing a lot of "falcon deaths" back in the day, so didn't want to risk killing mine

Yes the TOS programs, NEM_LO is for 40MHz (so 20MHz cpu and 20MHz bus) and NEM_HI is for 48MHz (so 24MHz cpu and 24MHz bus)

The 48MHz oscillator does have small dent in the top of the case, plus the legs were pretty bent before I rectified it (also made sure that PIN 1 had been removed). So I guess it could be damaged, if anyone knows where I can get a spare or borrow one just to eliminate the possibility.

The manual does mention that some Falcons are not happy running at 24MHz, so I guess there's the slim chance mine could be one of them.

[DarkLord]

Had that same exact problem with my first Falcon. Lower speed, CPU and bus, was
all it would run at in a stable manner. This was with the Nemesis.

[catmando]

Had that same exact problem with my first Falcon. Lower speed, CPU and bus, was
all it would run at in a stable manner. This was with the Nemesis.

Thanks for confirming

[dml]

An easy diagnostic is to take the 40MHz clock and put it in the 48MHz socket, and then try NEM_HI.

If it starts to work, then the machine isn't happy about a 48/24 master clock. If it still doesn't work, then something may be wrong with the board.

(All NEM_LO / NEM_HI does is select one or other clock socket, IIRC)

The software-based switching scheme can also cause problems. You can hardwire it to 48MHz (= NEM_HI) by connecting the 2nd ACIA switching line to +5v via a resistor. If that works, it's the switching that's the problem. [EDIT] I think the blitter also needs to be switched to 8mhz or turned off, or it can interfere with booting up at 24MHz.

Getting 48MHz to work is a bit tricky and has a lot to do with the location of the board, coupling of the board to the ground plane, wiring, RAM board speed and the machine itself. Because there are several clock lines involved, it just takes some excessive interference on one of them to upset things...

[catmando]

An easy diagnostic is to take the 40MHz clock and put it in the 48MHz socket, and then try NEM_HI.

If it starts to work, then the machine isn't happy about a 48/24 master clock. If it still doesn't work, then something may be wrong with the board.

(All NEM_LO / NEM_HI does is select one or other clock socket, IIRC)

The software-based switching scheme can also cause problems. You can hardwire it to 48MHz (= NEM_HI) by connecting the 2nd ACIA switching line to +5v via a resistor. If that works, it's the switching that's the problem. [EDIT] I think the blitter also needs to be switched to 8mhz or turned off, or it can interfere with booting up at 24MHz.

Getting 48MHz to work is a bit tricky and has a lot to do with the location of the board, coupling of the board to the ground plane, wiring, RAM board speed and the machine itself. Because there are several clock lines involved, it just takes some excessive interference on one of them to upset things...

Would adding the red wire make any difference do you think?

[dml]

If you mean the red one marked "CLK32/COMBEL #20/R234" - this is the wire that actually accelerates the bus & cpu. I'm not sure how you get 125% without that one connected? Unless the R234 resistor is still in place?

[catmando]

If you mean the red one marked "CLK32/COMBEL #20/R234" - this is the wire that actually accelerates the bus & cpu. I'm not sure how you get 125% without that one connected? Unless the R234 resistor is still in place?

Yes R234 is still in place, I followed the manual as it only mentions removing R234 and connecting the RED wire (on page 25) in the event that problems relating to appendix 1.1.4 are found. Since my specific problem was not mentioned in that appendix I didn't connect it up.

[catmando]

Does anyone have photo of the back of the nemesis board out of interest

[dml]

Yes R234 is still in place, I followed the manual as it only mentions removing R234 and connecting the RED wire (on page 25) in the event that problems relating to appendix 1.1.4 are found. Since my specific problem was not mentioned in that appendix I didn't connect it up.

Ok it looks like R234 removal + red wire is optional then. The units I saw all had R234 removed and both wires fitted.

With R234 in place, one clock line feeds both VIDEL + COMBEL, using the resistor as a splitter. With it removed, you can feed both chips with its own buffered clock. It may be that 48mhz is more achievable by removing the resistor but I can't say for sure.

[catmando]

Yes R234 is still in place, I followed the manual as it only mentions removing R234 and connecting the RED wire (on page 25) in the event that problems relating to appendix 1.1.4 are found. Since my specific problem was not mentioned in that appendix I didn't connect it up.

Ok it looks like R234 removal + red wire is optional then. The units I saw all had R234 removed and both wires fitted.

With R234 in place, one clock line feeds both VIDEL + COMBEL, using the resistor as a splitter. With it removed, you can feed both chips with its own buffered clock. It may be that 48mhz is more achievable by removing the resistor but I can't say for sure.

Okay, looks like I might give the red wire and remove R234 a try, maybe there's not enough grunt without it.

[catmando]

Well I removed R234 and hooked up the red wire with mini coax to the upper pad of R234, brown still connected to lower pad.

Appears to have the same issue when I run NEM_HI.PRG, NEM_LO.PRG still works fine though.

Any other suggestions?

[dml]

Did you try swapping the 40/48mhz clocks over as mentioned a few posts above? i.e. test 48mhz using NEM_LO and 40MHz using NEM_HI?.

This would at least help narrow down if it's the high speed that is causing issues or the integration of the NEM_HI switching control (either the wiring to the mainboard, or the Nemesis itself).

Another variation on this test is to use a mid-speed clock e.g. 42MHz, 45MHz in the NEM_HI slot. I remember using these odd speeds to do various tests with my own.

It's most likely some issues relating to the speed - but I'd do this test to be sure. Even if it doesn't work *properly*, you can at least see if the result looks the same or not. What you might do next will depend on that outcome.

[catmando]

Did you try swapping the 40/48mhz clocks over as mentioned a few posts above? i.e. test 48mhz using NEM_LO and 40MHz using NEM_HI?.

This would at least help narrow down if it's the high speed that is causing issues or the integration of the NEM_HI switching control (either the wiring to the mainboard, or the Nemesis itself).

Another variation on this test is to use a mid-speed clock e.g. 42MHz, 45MHz in the NEM_HI slot. I remember using these odd speeds to do various tests with my own.

It's most likely some issues relating to the speed - but I'd do this test to be sure. Even if it doesn't work *properly*, you can at least see if the result looks the same or not. What you might do next will depend on that outcome.

Okay, I decided to replace the grey wire with coax and now NEM_HI no longer freezes the machine, NEMBENCH reports 150% on all tests.

So I think we can now rule out any issue with the 48MHz oscillator and have proven that the machine can run at 24MHz.

The strange thing is that now the machine does the same freezing problem when using NEM_LO.

So am a bit baffled as to what step to take next?

[exxos]

A scope of the clock lines would help. I can't remember if it has resistors in the clock lines, but something like 33-68R might help.

[catmando]

A scope of the clock lines would help. I can't remember if it has resistors in the clock lines, but something like 33-68R might help.

Unfortunately a scope is something I don't have sadly.

[catmando]

Hmmm, wondering if someone else could confirm something I've just spotted, one of the chips below the 48MHz oscillator appears to have two of the pins bridged with solder, I've checked a few other pics of the nemesis and can't see another that matches what I am seeing on mine

nemesis.jpg

[catmando]

Here's a better picture of the same board before I started the install

nemesis2.jpg

[dml]

TBH if it works properly at 24mhz then it sounds like its mostly ok. Usually when there are bad grounding/wiring problems it works at 20 but not at 24 - either excessively flaky when switching to the higher speed or lockups after running some minutes.

Sometimes the switching lines themselves could be flaky i.e. the machine would run reliably at 20 and 24 if the wires are soldered permanently or placed on a physical switch, but would crash if using NEM_?? to software-switch. This might be what you're seeing.

So that could be a debouncing thing with the hi/lo switching wires going to the ACIAs - maybe noise or unwanted capacitance there. Inspect the connections for those carefully (and IIRC it involved lifting legs on the ACIAs?). You could try adding resistors there as well, but with a higher value resistor - 1k or more (refer to schematic to confirm if pullup or pulldown is required - but trial & error should be safe enough with 1k+).



You could also look into grounding and the location of the board relative to the mainboard. However since the machine seems to work at 24mhz its probably not worth disturbing all the high speed wiring again before ruling out the low-speed switching lines mentioned above....

I can't check my own installation just now but when I get my machine unboxed again (in a few weeks!) I'll open it and take a photo.

[EDIT]

The solder blob you mention doesn't look planned - but it could easily be like that on the PCB too (I see a track at that place on another photo).

But yes I'd still be tempted to clean that off until it can be confirmed. I had some other devices arrive recently from ebay with that problem.

[catmando]

[EDIT]

The solder blob you mention doesn't look planned - but it could easily be like that on the PCB too (I see a track at that place on another photo).

But yes I'd still be tempted to clean that off until it can be confirmed. I had some other devices arrive recently from ebay with that problem.

Removed the solder blob to find there was no trace beneath and tested the machine to find that the keyboard wouldn't respond. So re-added the solder blob and now the keyboard works again, so it looks like it was there for reason after all.

[catmando]

TBH if it works properly at 24mhz then it sounds like its mostly ok. Usually when there are bad grounding/wiring problems it works at 20 but not at 24 - either excessively flaky when switching to the higher speed or lockups after running some minutes.

Sometimes the switching lines themselves could be flaky i.e. the machine would run reliably at 20 and 24 if the wires are soldered permanently or placed on a physical switch, but would crash if using NEM_?? to software-switch. This might be what you're seeing.

So that could be a debouncing thing with the hi/lo switching wires going to the ACIAs - maybe noise or unwanted capacitance there. Inspect the connections for those carefully (and IIRC it involved lifting legs on the ACIAs?). You could try adding resistors there as well, but with a higher value resistor - 1k or more (refer to schematic to confirm if pullup or pulldown is required - but trial & error should be safe enough with 1k+).



You could also look into grounding and the location of the board relative to the mainboard. However since the machine seems to work at 24mhz its probably not worth disturbing all the high speed wiring again before ruling out the low-speed switching lines mentioned above....

I can't check my own installation just now but when I get my machine unboxed again (in a few weeks!) I'll open it and take a photo.

I can confirm that the legs 3&4 are both lifted on both the NEM_LO & NEM_HI switch return lines (the purple wires).

Can you explain more simply regarding the resistor implementation (how are the resistors added to the switch lines)?

Thanks for your help so far

[catmando]

Okay all solved by replacing both the purple lines with COAX as well.

Now I can switch from NEM_LO to NEM_HI and vice-versa, all NEMBENCH tests work perfectly too.

I found an old document in a thread contributed by Simbo that clearly lists the wires you need to use COAX on

Code: Select all

----------------------------
J1 == USE THE COAX FOR THIS
J2 == USE THE COAX FOR THIS
J3 == USE THE COAX FOR THIS
J4 == USE THE COAX FOR THIS
J5 == USE THE COAX FOR THIS
J6 == USE THE COAX FOR THIS 
J7 == USE THE COAX FOR THIS
------------------------------------------------------
J8 == NO NEED FOR COAX / NEMISIS BOARD SWITCHING
J9 == NO NEED FOR COAX \ NEMISIS BOARD SWITCHING
------------------------------------------------------
J10 == USE THE COAX FOR THIS
------------------------------------------------------
J11 == NO COAX NEEDED < NEMISIS BOARD SWITCHING
------------------------------------------------------