exxos blog - random goings on

[exxos]

Still printing white posts :lol: Been experimenting with temperatures and I had to bin a lot of the post because top tabs were too fragile. :roll: But I was printing at 180c during those. Was trying to reduce the stinginess of the top section more than anything. But ironically even printing at 230c does not seem to make any visual difference but the part is are a lot stronger now. I may just abandon the slotted approach and just do a solid circle which just simply fits properly.

A side-quest relating to this post viewtopic.php?f=15&t=5914 With the failing RAM chips...

I have connected my Falcon up on the floor and just left it running. I even heated up all the chips to "burn your finger hot" and none of them have died. So I don't think it is just a simple matter of chips failing. I suspect the Falcon is literally killing them somehow. RAM to fail the most in these machines, even on ST's. Really this needs a huge investigation going into it all.. But it would be a lot of time effort and money for something nobody really cares about anyway. But indeed if something does fail on this board on a particular bit then I can focus on that bit compared to the other ones and see if there is any significant difference the signals.

Problem really is there is no timescale to end this board potentially fail. It could be hours days months or years. Obviously with the price of electric I'm not going to leave this on more than a few days anyway. Either way I've taken the remaining 10 RAM boards off sale until more is learned about the problem. If anything.

I could design a new RAM board with the signal "snubbers". Though as said in the other thread, convincing the "general public" that there is a problem and getting them to pay 3x for a board where "nobody else" has problems would just be a pointless endeavour anyway. Nobody would buy it unless somebody repeatedly tried different boards From various sellers and they all failed. Even so those odd couple of people just isn't worth investing time and money into it.

IMG_0139.JPG

I did look at the voltages for the particular DRAM I am using. It basically says VCC + 0.5V positive and negative maximum input voltage. So with this spiking to negative 1.5V and positive 7volts, then no wonder the DRAM fails. I also wonder if this contributes to the Videl falling in Falcons as well. Which is a little worrying to say the least.

[exxos]

<Linus> Ohhh interesting </Linus>

It only did this once in a almost 3 hour run... Same as in the other thread..

IMG_0144.JPG

EDIT:

It is a little difficult to prove that there is more noise on D13 than others. D15 also looks the same, but then D11 does seem to overall have 200mV less spikes high and lower.

IMG_0148.JPG

IMG_0147.JPG

IMG_0146.JPG

IMG_0145.JPG

EDIT2:
Been running about another hour and no fails again.. All kinda worrying really :hide:

[exxos]

Over five hours now and still no more fails :roll:

I cannot do much but I did the schematic part of the hypothetical upgrade to this board.. So if anyone wants a weekend project.. :lol: I somehow wonder if I should have gone for smaller than 0603 parts :roll:

Capture.PNG

[exxos]

:WTF: :lolbig:

Capture.PNG

Probably a virus or a glitch in the matrix :lol:

I'm a bit behind anyway...

Capture.PNG

[exxos]

oops.

Capture.PNG

[exxos]

The Falcon RAM board was going over 11 hours yesterday and aside from that 1 error (on bit 13) , it passed fine.

IMG_0151.JPG

This morning I turned it on and..

IMG_0152.JPG

Powered back off and on again..

IMG_0153.JPG

Interestingly the guy who had problems last year also had bit 11 issue. But its probably just coincidence.

eagle060.jpg

So either powering down the falcon last night, or turning on this morning killed something. These boards do have a 5V voltage suppressor on them and the PSU is recapped. The 5V rail doesn't spike or go negative during power up/down so I don't think it's the 5V rail itself causing it.

As the remaining RAM boards are at my girlfriend's, I can't do anymore testing until Monday when I can bring a board back. Then I will see if I can replicate the issue and see if bit 11 fails again.

[exxos]

Thought I would note this while I think about it. That this is both the same Bit11 but literally taken a few seconds apart. So would seem dependent on whatever bus activity is happening at that particular time.

IMG_0155.JPG

IMG_0156.JPG

Pretty bad spiking up to 6.80V and negative 2.40V!

[stephen_usher]

Install diodes to clamp any negative voltages? They'd need to be fast diodes though.

Or possibly small decoupling caps to filter the high frequency spikes?

[Badwolf]

This morning I turned it on and..

:pullhair:

BW

[exxos]

Install diodes to clamp any negative voltages? They'd need to be fast diodes though.

It's a good thought but I've been down that road before. The fastest diodes with lowest voltage drop tend to be schottky diodes. But they have a huge capacitance which is hard to drive. Normal diodes are either not fast enough or voltage drop is higher than you think. Also a "hard clamp" can cause even more oscillations which can aggravate the problem. I can't remember what I was testing at the time, but I know I gave up in the end. I think you could clamp to maybe half a volt at best but thats still on the limits of the DRAM anyway :(

Or possibly small decoupling caps to filter the high frequency spikes?

viewtopic.php?p=93026#p93026

Even 100pF wasn't enough :( Various tests of resistors and caps were done. I guess spikes is not really the correct term, it is actually a oscillation which needs to get squatted out.

IMG_0127.JPG

So a combination of resistors and caps are needed. Resistors to limit the oscillations and the cap to help smooth the signal and also helps clamp oscillations.