YARRB [SOLD OUT]

discussion of games, software, hardware & emulators relating to the Acorn Atom and Acorn System machines.
User avatar
roland
Posts: 3395
Joined: Thu Aug 29, 2013 8:29 pm
Location: Born (NL)
Contact:

Re: YARRB [SOLD OUT]

Post by roland » Thu Jun 13, 2019 9:47 pm

OK, that is settled then :D
256K + 6502 Inside
MAN WOMAN :shock:

User avatar
etaoin
Posts: 28
Joined: Sat Oct 06, 2018 12:34 pm
Location: Vleuten, Netherlands
Contact:

Re: YARRB

Post by etaoin » Tue Jul 02, 2019 6:39 pm

Ok, almost working...

It is randomly switching graphic modes if I leave it alone. I'm sure I've seen that mentioned somewhere but can't find it right now. Something for tomorrow.
Attachments
IMG_7524.jpeg
IMG_7525.jpeg

User avatar
anightin
Posts: 329
Joined: Thu Aug 23, 2018 1:03 pm
Location: Cambridge UK
Contact:

Re: YARRB [SOLD OUT]

Post by anightin » Tue Jul 02, 2019 6:44 pm

Here are my settings:

JP3 on 1-2
JP4 on 1-2 (pin 1 is at the CPLD side).
JP5 and P5 closed
P4 and P6 open


I notice you have JP5 open?


Best

Andy
Last edited by anightin on Tue Jul 02, 2019 6:47 pm, edited 1 time in total.

User avatar
etaoin
Posts: 28
Joined: Sat Oct 06, 2018 12:34 pm
Location: Vleuten, Netherlands
Contact:

Re: YARRB [SOLD OUT]

Post by etaoin » Tue Jul 02, 2019 6:51 pm

anightin wrote:
Tue Jul 02, 2019 6:44 pm
I notice you have JP5 open?
No, JP5 is jumpered. P5 is open but that has been pulled up as suggested by Roland in his mods list.

User avatar
anightin
Posts: 329
Joined: Thu Aug 23, 2018 1:03 pm
Location: Cambridge UK
Contact:

Re: YARRB [SOLD OUT]

Post by anightin » Tue Jul 02, 2019 6:53 pm

Apologies, meant type P5 :oops:
To prevent writing to the EEPROM you can remove P5. But then you have to add a 4k7 resistor between the pins 31 and 32 of the EEPROM socket otherwise the EEPROM might not be readable (this is a design flaw in the board, this resistor should have been there).
Got it.
Last edited by anightin on Tue Jul 02, 2019 6:57 pm, edited 1 time in total.

User avatar
roland
Posts: 3395
Joined: Thu Aug 29, 2013 8:29 pm
Location: Born (NL)
Contact:

Re: YARRB

Post by roland » Tue Jul 02, 2019 11:29 pm

etaoin wrote:
Tue Jul 02, 2019 6:39 pm
Ok, almost working...

It is randomly switching graphic modes if I leave it alone. I'm sure I've seen that mentioned somewhere but can't find it right now. Something for tomorrow.
On the previous page, Sirmorris had solved this issue by replacing the 8255 and the chips in the clock circuit. If you have spares, this is worth a try. If you need a 8255 to test, just let me know.
256K + 6502 Inside
MAN WOMAN :shock:

User avatar
etaoin
Posts: 28
Joined: Sat Oct 06, 2018 12:34 pm
Location: Vleuten, Netherlands
Contact:

Re: YARRB [SOLD OUT]

Post by etaoin » Wed Jul 03, 2019 11:54 am

roland wrote:
Tue Jul 02, 2019 11:29 pm
On the previous page, Sirmorris had solved this issue by replacing the 8255 and the chips in the clock circuit. If you have spares, this is worth a try. If you need a 8255 to test, just let me know.
Yes, what he observed seems very similar. I'll first replace the sockets around the clock, as they're crap anyway. The 8255 already has a new socket and I think isn't the original 8255 anyway (it's a NEC chip). I've used lots of 8255's for DIY stuff ages ago so I probably pulled the original at some point and replaced it later. I probably have one or two (used) spares.

User avatar
etaoin
Posts: 28
Joined: Sat Oct 06, 2018 12:34 pm
Location: Vleuten, Netherlands
Contact:

Re: YARRB [SOLD OUT]

Post by etaoin » Thu Jul 04, 2019 7:15 am

Quick question: is the duty cycle of the 4 MHz clock expected to be way off 50%? Does the CPLD care anyway?
Attachments
Schermafbeelding 2019-07-04 om 09.11.12.png

User avatar
etaoin
Posts: 28
Joined: Sat Oct 06, 2018 12:34 pm
Location: Vleuten, Netherlands
Contact:

Re: YARRB [SOLD OUT]

Post by etaoin » Thu Jul 04, 2019 7:21 am

etaoin wrote:
Thu Jul 04, 2019 7:15 am
Quick question: is the duty cycle of the 4 MHz clock expected to be way off 50%? Does the CPLD care anyway?
Sorry, I'm being stupid. My 12 MHz sample rate will skew the duty cycle.

User avatar
roland
Posts: 3395
Joined: Thu Aug 29, 2013 8:29 pm
Location: Born (NL)
Contact:

Re: YARRB [switching to CLEAR 3]

Post by roland » Sat Oct 12, 2019 10:01 pm

I did some measurements on my Yarrb'd Atom that continuously goes into CLEAR 3 graphical mode:
yarrb-setup.jpg
This was the first result:
Capture1.png
Please note the small spike in NRDS a few microseconds before the trigger (vertical white line) ....
A closer look:
Capture-zoom.png
This short NRDS pulse is a cpu read but the 8255 seems not to be ready. It's data lines are still in three state and the CPU "reads" B0. At the next write cycle which is part of the keyboard scan routine (the cpu writes a value to #B000 for selecting a row in the keyboard matrix) the cpu writes this erroneous value to port A which activates CLEAR 3.

As you can see in the trace, the cause of this error is in a short Phi2 pulse. With my (new \:D/ ) scope I checked Phi2 on the Yarrb board. It shows of course the same small pulses. Next I checked what is going into the 6502 (Phi0 which is the divided clock out of the cpld). That also shows those pulses. So the cpld outputs an incorrect clock signal. After seeing that, I was curious what was going into the cpld (ClkIn). That is the 4 MHz from the Atom main board.

When I set my probe on that track (pin 4 of the header right next to the 6502 on the Yarrb board, then I see a steady 4 MHz wave with some undershoot. But from that moment, the Atom won't go to CLEAR 3 any more. It stays in normal mode. As soon as I remove the probe, it goes into CLEAR 3 again. I can do the same trick with a small screw driver: touching that point makes the Atom run stable.

The next thing that I will try (tomorrow) is adding a > 10k resistor at ClkIn and GND or +5V to see what the effect will be. I have no real explanation for what is going on in the cpld. I cannot imagine that the divider is not correct because than I would expect every Yarrb board be faulty and that does also not explain why touching ClkIn (the 4 MHz signal) also solves the problem.

To be continued.....
256K + 6502 Inside
MAN WOMAN :shock:

bprosman
Posts: 341
Joined: Sun Mar 29, 2015 10:27 pm
Contact:

Re: YARRB [SOLD OUT]

Post by bprosman » Sat Oct 12, 2019 11:01 pm

What new toy.. ehh scope do you have Roland ?

User avatar
1024MAK
Posts: 9236
Joined: Mon Apr 18, 2011 4:46 pm
Location: Looking forward to summer in Somerset, UK...
Contact:

Re: YARRB [switching to CLEAR 3]

Post by 1024MAK » Sat Oct 12, 2019 11:20 pm

roland wrote:
Sat Oct 12, 2019 10:01 pm
When I set my probe on that track (pin 4 of the header right next to the 6502 on the Yarrb board, then I see a steady 4 MHz wave with some undershoot. But from that moment, the Atom won't go to CLEAR 3 any more. It stays in normal mode. As soon as I remove the probe, it goes into CLEAR 3 again. I can do the same trick with a small screw driver: touching that point makes the Atom run stable.
The small amount of capacitance from either the ‘scope probe (including lead and ‘scope input capacitance), or from you and the screwdriver could be just enough to alter the signal slightly...

Mark

User avatar
etaoin
Posts: 28
Joined: Sat Oct 06, 2018 12:34 pm
Location: Vleuten, Netherlands
Contact:

Re: YARRB [switching to CLEAR 3]

Post by etaoin » Sun Oct 13, 2019 11:50 am

roland wrote:
Sat Oct 12, 2019 10:01 pm
As soon as I remove the probe, it goes into CLEAR 3 again.
That sounds suspiciously like what I had, the "It is randomly switching graphic modes if I leave it alone."

User avatar
hoglet
Posts: 8551
Joined: Sat Oct 13, 2012 6:21 pm
Location: Bristol
Contact:

Re: YARRB [SOLD OUT]

Post by hoglet » Sun Oct 13, 2019 1:22 pm

Roland,

I have also had a problem very like this before - on the new Electron AP5. In that design there is a 4-bit counter, and the symptom was every so often the counter would seem to increment by two rather than by one. And just like you are seeing Roland, as soon as you apply a scope probe to the CPLD clock input, it stops happening.

What I concluded is that the XC9500XL series CPLD are very sensitive to noise/ringing on the clock input.

In YARRB you are taking the 4MHz clock input from IC44 pin 13:
ic44.PNG
This is driven directly by the 4MHz oscillator, and the feedback resistor (R26, 470R) my well be affecting the levels.

I would look at the 4MHz clock on your new scope. I'd look at:
- what's the nominal low level
- what's the nominal high level
- how much noise is there (use infinite persistance mode and leave it running for 5 minutes)
- can you catch any runt pulses (use the runt pulse triggering mode)

The fix might be as simple as adding a small capacitor (50-100pF), or a small series resistor (30-100R), or both.

So post some screen shots from your nice new scoop!

Dave

User avatar
roland
Posts: 3395
Joined: Thu Aug 29, 2013 8:29 pm
Location: Born (NL)
Contact:

Re: YARRB [SOLD OUT]

Post by roland » Sun Oct 13, 2019 3:02 pm

hoglet wrote:
Sun Oct 13, 2019 1:22 pm
The fix might be as simple as adding a small capacitor (50-100pF), or a small series resistor (30-100R), or both.
Dave,

Short version:
We both came to the same solution and for people in a hurry: this works \:D/

Long version:I remembered this article in Atom News (sorry, only in Dutch) about reflections in long cables that were causing under- and overshoot. The solutions that I did once bitd was installing a series resistor of about 150 Ohm. I tried to make some screen dumps with my new scope but somehow I managed only to store a bin file of the wave form before the resistor was installed:
before.png
Blue = channel 1 (4 MHz)
You can clearly see the undershoot peak in this signal. After installing the resistor the signal looks like this:
after.png
Yellow = channel 1 (4 MHz)
Blue = channel 2 (Phi 2)
After installing this resistor:
resistor.jpg
Series resistor 180 Ohm just before pin 37 on the main board.
Now the Atom is running stable for more than half an hour at 1 and 2 MHz. And noise free :lol:

I will add this resistor by default (maybe 100 Ohm or so) in the Yarrb2 design, as close as possible at the ClkIn input of the CPLD.
256K + 6502 Inside
MAN WOMAN :shock:

User avatar
roland
Posts: 3395
Joined: Thu Aug 29, 2013 8:29 pm
Location: Born (NL)
Contact:

Re: YARRB [SOLD OUT]

Post by roland » Sun Oct 13, 2019 3:07 pm

bprosman wrote:
Sat Oct 12, 2019 11:01 pm
What new toy.. ehh scope do you have Roland ?
An Owon XDS2102A 12-bit High Resolution 100 MHz 1Ga/s with USB and LAN and VGA with many buttons and knobs :lol:
A picture is here.
256K + 6502 Inside
MAN WOMAN :shock:

User avatar
anightin
Posts: 329
Joined: Thu Aug 23, 2018 1:03 pm
Location: Cambridge UK
Contact:

Re: YARRB [SOLD OUT]

Post by anightin » Sun Oct 13, 2019 6:55 pm

roland wrote:
Sun Oct 13, 2019 3:02 pm
hoglet wrote:
Sun Oct 13, 2019 1:22 pm
The fix might be as simple as adding a small capacitor (50-100pF), or a small series resistor (30-100R), or both.
Dave,

Short version:
We both came to the same solution and for people in a hurry: this works \:D/

Long version:I remembered this article in Atom News (sorry, only in Dutch) about reflections in long cables that were causing under- and overshoot. The solutions that I did once bitd was installing a series resistor of about 150 Ohm. I tried to make some screen dumps with my new scope but somehow I managed only to store a bin file of the wave form before the resistor was installed:

before.png

You can clearly see the undershoot peak in this signal. After installing the resistor the signal looks like this:

after.png

After installing this resistor:

resistor.jpg

Now the Atom is running stable for more than half an hour at 1 and 2 MHz. And noise free :lol:

I will add this resistor by default (maybe 100 Ohm or so) in the Yarrb2 design, as close as possible at the ClkIn input of the CPLD.
Thank you Roland, tried that mod on my randomly misbehaving ATOM and so far so good! (1 used 100R) =D>

Post Reply