exxos blog - random goings on

[exxos]

While "attempting"to tidy up a bit, I came across the Falcon memory boards.. Which go way back to this topic https://exxosforum.co.uk/forum/viewtopi ... 087#p93087

I connected the solder paste printer and got some DRAM soldered on.

IMG_2625.JPG (302.33 KiB) Viewed 676 times

I don't believe there was any conclusion as to why the Falcon memory was failing intermittently or totally. However, these memory boards have filters on all the IO lines and voltage suppressors. This is as realistically good as it can be in order to protect the RAM from being killed by whatever is happening on the bus.

Having a quick read back through the threads it does seem to suggest that there is also some stability issues leading to intermittent RAM faults. Thats another problem (but could be "bad bus" leading to DRAM failure I guess) .. However the RAM itself also seems to fail more on power up. It does not tend to fail when running for several hours.

Also looking back, I think some DRAM is inherently more resilient to "abuse". As some DRAM seems to work fine, and others fail in weird ways. Either by intermittent "bit faults" or a total DRAM IC failure.

The annoying thing is the problem is very intermittent, so trying to draw any conclusions as to if these "noise and spike immune" Falcon RAM boards are helping or could be somewhat tricky. However if anyone has Falcon's and is having RAM boards fail, then it could be evident that this is rather a widespread problem of sorts, which is only more recently been realised..

[exxos]

I've not been able to do much lately because of the usual hand pain and fighting server issues

Though I did manage to make a little bit of progress with the Falcon memory boards..

Previous waffle is here...
https://exxosforum.co.uk/forum/viewtopi ... 102#p93102

Basically some Falcon seem extremely bad in accessing RAM and can lead to sporadic bit failures such as this...

IMG_0144.JPG (93.48 KiB) Viewed 450 times

And can ultimately lead to total bit failure..

When I was selling the integrated Falcon RAM board, they all passed my tests at the time in my Falcon, however after I took them off sale due to reports of them "failing", I retested them again in my Falcon and they all had weird issues.

Initially errors were only showing up upon initial power up so it could be when the machine is cold things are worse. Though as illustrated in the above image, these random bit errors can happen sporadically at any time even after a few hours of use!

Was a bit of a faff getting my solder paste printer running after its last use will got there in the end..

IMG_2623.JPG (293.51 KiB) Viewed 450 times

And then had to adjust my template to hold position DRAM chips, as doing this manually by hand with tweezers is practically impossible

IMG_2631.JPG (227.75 KiB) Viewed 450 times

IMG_2632.JPG (235.58 KiB) Viewed 450 times

My new 3-D printer has so far been a dream to use thankfully! No messing about for several hours before I get a successful print, everything "just simply works".. Famous last words..

Then a bit of cooking and soldering..

IMG_2634.JPG (363.2 KiB) Viewed 450 times

And finally testing...

IMG_2656.JPG (112.03 KiB) Viewed 450 times

Unfortunately I had various problems with the previous few board revisions

While I had previously worked out the best filter values for the signals, it seems that the effect of "capacitive loading" seems accumulative across all signals

For example, if I had something like 100pF loading, on a couple of data lines it would be perfectly fine. However, when having that across all data lines, the signals ended up being skewed and slowed down on the rise and fall in the signals ended up more like a triangular wave. So it resulted in a non-booting machine or corruption

IMG_2626.JPG (213.71 KiB) Viewed 450 times

So I don't think the Videl has a set amount of output current capability per pin. I suspect its total output capability is across all datalines. So some arbitrary numbers of say 100mA on one pin, or 100mA across all pins. But not 100mA per pin kinda thing. So I basically just had to guess a few times on what values to us. It resulted in a rather a lot of boards being chucked away as it was basically trial and error

The current revision has been on test for a long time without any errors so far..

IMG_2658.JPG (117.1 KiB) Viewed 450 times

1.jpg (48.33 KiB) Viewed 450 times

So things are moving in the right direction at least now... Albeit very slowly...

[exxos]

Not had chance for much Atari stuff over the past few months unfortunately, and probably won't over the next few months either I hardly even visit the forum anymore never mind read anything or post anything. I just do not have the time in one respect or another any more

Mostly I have been testing out parts for my store when I actually have some free time. So the prototype noise immune Falcon memory board has had a right hammer lately. It has not tripped up once yet! No random bit fails or ICs randomly dying for no reason yet!

I've not been sleeping much lately. So I've been pottering around a new power supply design. I was not going to produce the power supplies again because they take up so much time effort and costs. So a few minutes here and there has resulted in a new design. I moved away from the Texas chips as just wasn't happy with them in the end. The new design is a lot more complicated, but I think it performs much better than the Texas chips. There's been all sorts of changes like adding in separate 5V and 12V trimmer pots. So when people's machines are very voltage sensitive, people can up or down the voltage as required.

IMG_2805.JPG (253.16 KiB) Viewed 268 times

Last time I checked some time ago, several others were selling power supplies using "off-the-shelf" modules and selling that PSU at the same price as mine, was frankly annoying considering sheer amount of time and effort mine take to assemble. So because of this, plus the cost of parts rising, a new design, and moving to UK fabrication and assembly, the cost of this new power supply will be a lot higher than before. For people who want to buy the cheaper power supplies using off-the-shelf modules, there's already a DIY kit in the store and has been for some time. With the lack of EU sales now and the price hike, I am not really expecting many of these new designs to sell, as such I will not be building up many of them up this time around.

[exxos]

When is a fuse holder not a fuse holder......

I've always got the fuse holders from China, good value and nothing really wrong with them at all..

However I forgot to order some, so ordered them from a UK company thinking that they would basically be the same, just cost a little bit more...

How wrong I was!

The lighter coloured one is the UK fuse holder, and a darker coloured one is the original ones I got from China.

The lighter coloured ones simply do not hold the fuse or grip it in any reasonable way. The fuse is a very sloppy fit. I tried to squash the metalwork inwards to grip better and while it did help a bit it was still loose.

I think the "( )" parts are to "long" and do not match the contours of the fuse properly either. I think the metal is just so cheap and badly made that the only thing I could do with a box of the things is just simply bin them all

My China fuse holders grips the fuse really well and it is not loose and a tight fit. It is actually rather difficult to get the fuse out of the fuse holder. But the ones I got from UK company, you could almost shake the fuse holder and the fuse would move that is how bad they are.

I purchased the China sockets for literally years and always use them without problem, so really have no idea where these UK fuse holders came from... I mean you can pretty much assume China are probably selling factory rejects, but these UK ones like the rejects of the rejects they are that bad...

IMG_2822.JPG (229.66 KiB) Viewed 210 times

IMG_2821.JPG (558.84 KiB) Viewed 210 times

IMG_2820.JPG (492.1 KiB) Viewed 210 times

[exxos]

A few weeks ago I got some more simm PCBs made. Finding RAM at a good price was incredibly difficult but I was ultimately triumphant. The problem was, I had to adapt the DRAM package on the PCB which somehow led to a short on the edge connector

This was incredibly frustrating because I had to do a special order for some slightly thicker than before PCBs. It was somewhere around the £400 mark I basically had to do another order

So then I started wondering if I could adapt one of my crappy 3-D printers to "score" the short automatically. There are around 900 PCBs, so trying to do them by hand and accurately enough is just not viable.

I contacted the metalwork guy who made me my blog for my chair and he seemed happy to do me a nozzle of sorts which could be used as a scoring tool. However he seemed to vanish about two weeks ago and I have not heard from him since

I had to start getting creative and thought I could simply mash a needle into the nozzle .. One needle had was 0.6mm, so I bought a 0.6mm nozzle and jammed it in there...

Then came the almighty ordeal of converting Gerber data into Gcode and with the help of GBT4, turning the code into various "loops".

Capture.PNG (15.84 KiB) Viewed 189 times

So each line is drawn about 30 times, then the head moves down like 0.01mm, and does another 30 times, and the head moves down around 10 times in total. And it also has to move ahead upwards about 10 mm for the travel movements as well.

The problem was all compounded by coming up with a solution to clamp the PCB onto the print bed, then the print bed itself was not actually straight.. The PCBs themselves don't lie flat either. So I basically had to start the cutting movements slightly above the PCB to take into account the uneven surface.. Then working out the X and Y offsets.. Needless to say took a while to figure out...

So this was the first of many test runs...





I had to order a 1mm nozzle and the 0.6mm needle was bending So I need to figure out what my Z offset is again and repeat the test and see if things progress better this time....

I also think the bed isn't level, I'm not sure if ABL is causing that or not, but I cant ABL anymore anyway.. It seems the right side of the bed is about 0.2mm lower than the left side of the bed. I can't manually adjust this printer bed which is annoying

Also just figured out there was a bad connection on the X stepper motor causing some "random drift" for no reason

[Icky]

I am surprised you haven't replaced the print head with a high powered laser @exxos

lasers.jpg (16.97 KiB) Viewed 172 times

[exxos]

I am surprised you haven't replaced the print head with a high powered laser @exxos

Ironically laser was my first point of call, but someone else with a laser.. The guy basically said he had the same problem with a short on PCBs and even a relatively high power laser was not powerful enough to cut through the copper

[Icky]

In the electronics firm I worked at in the early 90s we used to have one of these to cut tracks on substrates and PCBs

chicago_cls-960_1039597-prev.jpg (29.92 KiB) Viewed 167 times

A Chicago Laser System no longer around but you can pick these things up.

[exxos]

In the electronics firm I worked at in the early 90s we used to have one of these to cut tracks on substrates and PCBs
A Chicago Laser System no longer around but you can pick these things up.

Nice! Bit of an overkill for a one-off project though probably be cheaper to get the PCBs redone.

[exxos]

These cheap printers are so crap!! I'm running into all the usual issues that I normally do with them

Managed to turn off ABL as it seemed to be all over the place. Things seem more consistent now. However the bed seems to be leaning to the right about 0.3mm which is rather a lot! This has basically caused a "head crash" and bent my needle slightly, so had to adjust everything again!

Then for some reason totally unknown, it decided to "forget" to move the Y-axis, so the needle crashed into my metal clamping bars screwing up alignment yet again

I just don't have all this crap with my QIDI printer.. It "just works".