zzattack

Didn't have too much time last night, but experimented with the idea of manually tenting via's within the BGA area. I applied a thin layer of UV curable solder mask over exposed copper that wasn't part of any BGA pads, cured it, and then mounted an interposer. This appears to have prevented any bridges from forming between solder ball and nearby via. Ideally this wouldn't be necessary and I do intend to figure out how this happened in the first place, and how to prevent it from happening, but this appears to be a simple workaround. I'll complete the install on this one tonight and hopefully book some success with it. It's about time.

No, CPU removed. I do intend to image them directly after soldering the interposer.

I've put the conversions on hold for a little while since my installs after the first successful one didn't work no matter what I tried. Cleaned up my soldering profiles, ensured my alignment was dead on, sought help from very knowledgeable people including N64 himself, to no avail. At some point I stopped counting, but I believe I registered 9 or 10 failures and decided to take a step back, since the time wasted for no results became a source of frustration. Uploaded some pictures here: https://imgur.com/a/z0YPuoN Ultimately I decided the sensible continuation would be to employ an x-ray device to verify correct soldering of the interposer. I found a dental device that I could accommodate with my budget and the device has since arrived. The sensor area is small (21x21mm) but resulting images are surprisingly usable. Can't tell for sure if this will allow detection of voids or head-in-pillow easily, but certainly I can identify spots where the balls floated slightly. I will adjust the interposer reflow profile to remain above liquidus slightly longer. More updates soon!

As I laid down in bed I couldn't contain my curiosity since I was still so perplexed that the 3 attempts after my initial successful one failed. I needed to check. The first board I used was a 1.4, the others were 1.0. Could there be a big difference in alignment between them? The top of the board sits very close to an SMD cap. You judge! On a 1.4, it's close but it fits: On a 1.0, oops, that'll elevate the interposer and certainly not benefit the soldering quality: I'm fairly assured this caused the poor BGA alignment. And so we learn. More later this week.

Time for a minor update. Not a very happy one just yet but new ideas for improvement did spring to mind, so I'll use this post to share some ideas I had. First, I believe @Bowlsnapperhad some concerns about aligning the interposer to the board "blindly". There's no need to go in blind, in fact, we can make it foolproof. I measured the silk screen print accuracy on the adapter and on the motherboard with a digital caliper. There's only about 0.1mm difference in outline width/height between the interposer and console boards. They're also accurately spaced around the solder pads, meaning this outline actually makes for a pretty decent calibration opportunity. What you can do it apply some kapton tape along the silkscreen outline on both the board and the interposer. Then, just make sure the tape aligns, and you're guaranteed a very reasonable alignment. And then you just overlay the crosses: When soldered down, it's probably still quite possible to remove the tape by pulling it down and wiggling it out. Or you can just cut it near the sides, no big deal either way. I'll use this method to verify my split vision system's accuracy on my next attempt. Since I'm not picking up the interposer in dead center, I'm no longer convinced the placement is as accurate as the visual indicates. This will either confirm my suspicions or rest my mind.

I'd recommend the ACHI for sure. Aligning the interposer isn't too hard. The top end of the board nearly touches a capacitor on the top. For the horizontal alignment, just turn the interposer over, align the balls by eye and rotate the board around while keeping the same lateral position, then shove it up to that capacitor. The good thing about BGAs is that you're still fine even if you're just under half a pitch off. That's 0.6mm of margin! While everything up until now has been smooth sailing, I've ran into the first snags today. An eBay package with Tualatins arrived, and even though I ordered only 2 from the seller, I actually received 3. Sadly, one doesn't work. Not a loss overall, but still, this kinda indicates there's bad CPUs out there. I'm now glad that I purchased a socket 370 motherboard solely for testing. The other problem I ran into today manifested on two boards: the interposer tilted when soldering it down. I didn't initially realize this on the first board and have already soldered on the CPU there, but obviously the system FRAGs. Inspecting very closely under the microscope reveals that only the very corner ball is floating over the motherboard. The balls just 1 pad closer to the center have soldered but they are obviously not ideally squashed. Reason for this must be that the weight of the interposer board rests only on the BGA balls, but the board protrudes over the BGA area. I'll have to remove the CPU and reball this one unfortunately. I set this board aside before making this realization, and soldered a 2nd interposer on another motherboard. Close inspection of that one showed the clearance between interposer and xbox wasn't perfectly even, and pressing in the corners flexes the board slightly. Will have to reball this one as well, but this gave me an idea for a fix. The 0.76mm balls when soldered down are roughly as thick as 0603 passives. The board is already resting on a set of 4x 10k 0603s on one side, so we can use such 0603's to support the board in the far corners. I soldered 2 resistors with one end to via's near the corners so they won't slip away, but they serve no electrical function of course. I then continued to attach another interposer to this 3rd board and it came out perfectly. There's zero flex, the balls are squashed very nicely, and the clearance is perfectly even around all edges. I'm confident that the issue's found and remedied, and I'll verify that tomorrow if my wife lets me on Mother's Day

Jig idea: generic stencil for 0.76mm balls at 1.27mm pitch, with unused pads masked off using kapton tape. Since the interposer boards are 1.6mm tick, I surrounded it with PCBs of identical thickness, secured onto an ABS sheet with (thin) double-sided tape. The interposer squeezes firmly in between the surrounding boards. A few overlaying strips of kapton briefly elevate the stencil. This gives the stencil some vertical clearance for the balls to fall through, and ensures the flux on the board doesn't touch the stencil. Regardless, I highly recommend cleaning off the stencil after every board for the best results. If balls don't fall through on the next one, you should've cleaned it. After pouring on the balls, I use a very soft ESD brush to spread the balls. They balls are not encaged so I used a very light touch, or else I'd be finding solder balls all over the place for months. With the balls placed, lift the stencil and manually repair any imperfections. You can see the bottom-left ball didn't make it. After having worked on some 0.2mm BGA stuff recently (not fun), manually placing 0.76mm balls is no longer any issue, even for pads that have 8 surrounding ones. On to the hotplate for attaching them. I highly prefer IR over hot air because the balls don't move. With hot air, if you have just the tiniest bit of excess flux, balls start drifting freely. Repeat... and after about 2 hours the work came to an end. Surely there's better ways to spend your Saturday but ultimately, it's good to have it done. Not perfect, but with a small (say airbrush) compressor and simple paste dispenser you can place drips very quickly, with reasonable accuracy, and reasonably similar amounts of paste. So I'd say it's fine.

Tomorrow I'll add some pictures. I've prepared a jig for reballing the 9 interposers I still have, since placing the balls by hand just once, however humbling the experience may be, just isn't for me. For that I used a regular stencil with the unused pads masked using kapton tape. Pics will clarify. I'll also add some info for the paste dispenser that I used, although I prefer having a stencil made instead.

Picked up one defective xbox for recapping but noticed it still did not turn on even with fresh caps. The clock cap had leaked and taken along some traces with it. Luckily superficial, so easy to spot and fix: 1) Assessment of the damage. Not too bad. 2) Scraping away all the crap to expose bare copper; treat it like a contaminant and get some clean margins. 3) Tin the exposed copper to prevent oxidization. Very minor trace interruptions can be bridged by dragging solder over them. 4) The larger interruptions require being bridged with some wire. I found AWG40 to match the existing trace width. 5) For rigidity and insulation, applied some conformal coating. 6) Result after curing. Now she powers up nicely Next!

It's taken about a decade of time and a year's income to equip it to the point where it is now, and it allows me to do a whole lot of stuff from home. A lot of it is luxury though, making things easier or providing additional room for fuckups before turning disastrous. I've been able to take a huge variety of repair jobs, which in turn funded acquisition of more gear to take on additional repair jobs. A nice circle indeed. That said, this particular kind of job in my opinion does require a BGA station, but an Achi/Honton/Scotle will do perfectly fine. I did probably about 200 BGA replacements before I exchanged my IR6000 for what I use now. Getting a stencil is probably better than using solder paste from a syringe, so if you don't have a paste dispenser, that's what I'd suggest. Aside from that, I don't think the requirements are too demanding for this job. Some good flux maybe The shipping fees will feel like a scam for sure but I think we can be creative on 20 year old stuff when it comes to VAT/customs

Yeah... somewhere in between those two numbers indeed. Purchased as payment for some freelance work in an income tax optimizing attempt Can do. Would be one console for me to hunt. Feel free to contact in PM to arrange that. There's no list I think, and I have enough 1.4GHz's to cover the 10 for which I have components, so I'm not sure this will work out.

Had less time than anticipated tonight, but I did manage to pickup the replacement electrolytic caps from Farnell at the pickup point. All the 1.6's are recapped, which as suspected resurrected the 2 dead ones as well. So, these have clean shells, boards + PSU's gone through ultrasonic cleaning, recapped, fresh thermal paste, I restored the LPC with a QSB (but did not install any modchip). I'm ready to part with them, and I'll be hunting to find non-1.6's to replace them in the coming days. Tomorrow I hope to get the remaining ones recapped and TSOP'd, plus maybe RAM upgrade on one or two of them. It's about a 50/50 chance I'll have 2 more Tualatins on Saturday, so fingers crossed we may get to complete two boxes entirely. I've dumped some pictures here for now on imgur: https://imgur.com/a/grMufPW. The forum itself feels very restrictive in regards to maximum and combined filesizes.

I just made some offers on a few CPUs on eBay yeah. One of the sellers said "I have more" and I purchased 8 off of him. In total I will have 11, assuming at least one's a dud. I've got a WDS-620 station.

Regarding progress, yesterday I started on the pile of 10 by softmodding them/nulling HDD keys, primarily so that I can throw all disks on a pile without having to worry about keeping pairs. All motherboards and PSU's have received ultrasonic cleaning, while the enclosures took a bath to get rid of dust and potential smells. As far as I could tell, none of them came from nicotine using owners, but 3 or 4 likely had pets, so I'm happy those smells are replaced with a fresh scent. While the enclosures were getting bathed, I set out to remove some clock capacitors, clean up potential leaks from said supercaps, and overwrite the stock BIOS on the TSOP memory. Sadly, 4 out of 10 are 1.6's so they won't have a future with me. I don't think it makes sense to go all-in with a CPU swap if you're inevitably going to remain stuck on 64MB RAM with an Xcalibur encoder. I'll be patching these up with new caps, new thermal paste, and then I'll probably dump them at a loss. From the remaining 6, one isn't powering on (likely due to bulged caps) and 2 more do work despite having bulged caps. New caps were ordered yesterday, so today I will be replacing the electrolytics on each of the 10 boards. Time permitting, the 6 non 1.6 revision boards will get a TSOP flash, and then I can start with the RAM upgrades over the weekend. When those are done, I can take a (long) break until the Tualatin CPUs arrive from China. Meanwhile I also arranged for an old Socket 370 system so that I can test the CPUs before (semi-)permanently attaching them to the xbox motherboards.

Batch sales would work for me but this isn't something I'll likely pursue beyond the 10 that I've set my mind to. I typically pick up a project, go nuts with it, and move on. My plan is to build a total of 10 xboxes with upgrades 'done right'. I've also decided that I should focus on the CPU + RAM upgrade only, and leave further options open. Costs shouldn't be inflated by adding a €35+ openxenium if the end-user is perfectly happy with a zero-cost TSOP solution. Or someone already has a modchip lying around which they'd like to (re-)use, or maybe they prefer a dead simple Aladdin/Jafar. I feel that it is not up to me to force choices upon anyone, but I do not want to offer every possible combination either, so the logical approach is to offer the bare package at an affordable price where further tinkering remains entirely optional. By 'done right', I mean the quality of the performed upgrades should be excellent and without compromise. The resulting systems should be reliable and brought into good shape with (preventative) maintenance carried out. That, in my opinion, entails at least the following: leaky clock supercap removal + thorough cleaning of any leakage that has already occurred; all residue must be scraped away, resulting bare copper re-tinned or solder mask reapplied all electrolytic capacitors replaced, regardless of whether or not they've bulged, with new Rubycon/Nichicon/Panasonic, from a reputable vendor (so mouser/digikey/farnell, not some dried up stock from the bottom of some abandoned warehouse in China) after upgrades are applied: ultrasonic cleaning to remove all remnants of flux, dust and other crap that's been gathering inside for 20 year. Those might sounds obvious but at the same time I've read of an upgraded console that supposedly lived for some 20 hours before giving in to poor capacitors. That's upsetting but entirely avoidable. Another point I want to focus on is the quality of the BGA rework. For starters that means we will not be using heat guns or paint strippers... In fact I'd prefer not to use manual hot-air at all. Rather, I will employ my reasonably sophisticated BGA rework station with temperature controlled profiles, so as to obtain a repeatable process giving consistent results for each install. I'll be using leaded solder balls underneath the interposer, and a solder paste with a Bismuth composition, i.e. a paste that melts at temperature lower than regular leaded solder. For my first device I used the same soldering profile for both passes, but ideally, soldering down the Tualatin should use a profile where less heat is applied from the bottom (since this might reflow the interposer itself) and it should be tailored to the lower melting point of the solder paste too. The BGA station is equipped with split vision so aligning the interposer, despite lack of visual markers, is easily done by overlaying the top/bottom balls/pads. Finally, I've been thinking about heatsink solutions. Nothing definitive here just yet, but I've already realized that the Tualatins' internal heat spreader and lower operating voltage of just 1.45V contribute massively to good temperatures. Reusing the original heat sink with fins slightly trimmed, held onto the processor using a fairly simple polypropylene (heat resistant) 3-d printed jig combined with a reasonably effective thermal paste should yield excellent results. So that's the plan. I'll try to keep a bit of a logbook in this thread and add some progress pictures every now and then.