SpenceKonde
commented
2 years ago Tell me about it. (see attached image). I'm in line a little bit ahead of you for some of the first low pincount dd's. It sounds like the first chips came off the production line are severely broken such that even their 3/4ths assed QA declared them unfit. After a debacle where unusable DA32's made it into customer hands. However, since that release, everything seems to be coming out slower, but the initial errata lists have a table of contents that fits on a single page (and the 162x tinies first shipped as a REV E - a Rev D almost shipped and was rejected after the pre-release errata was posted, and they only shipped the E.. So maybe someone's cracking the whip at the QA folks now).
One might expect that at that point, since they're also calibrating the temperature sensor, that this would be where they binned the dies into industrial vs extended temp parts, but I find that hard to reconcile with the fact that there's no bit in sigrow nor marking on the package to tell you if you have an I-spec or E-spec on Dx-series (and the difference in temp range is the big one, 85C vs 125C. The E-spec also overclocks better at room temp, which isn't surprising). The tinies have the N or F after the part number indicating temp range, but the Dx parts don't, you have to ask microchip support about the week of mfg and lot code to find out, which is very strange IMO.
In any event, It's hard to see what the problem could be, since there is so little that is different on those parts vs the DD28/32, just less pins, Could they have had PC0 brought out to the pin instead of VDDIO2? (that should be easily fixed with no die change, just different bonding wire connections)? Is there a problem with the crystal oscillator driver, which has an added complication on the low pincount parts in that it can be either the 32k crystal or an HF crystal) on PA0/1? That's at least plausible..... If it's a problem with the new PORTMUX options that's present on all DD's and was only noticed when that was the only mux option they had on the small ones, they need to be slapped upside the head, everyone on the design and test teams, because they've been here before. They have managed to ship with at least one portmux feature broken on every family of parts released since the new architecture came out, and you'd really think verifying that would be on their checklist, especially being as it's simple to do....
If you consider how long it would take to get most other chips "order now, ships in July 2023" and how much those dates slip (my backordered 412s ordered from mouser originally scheduled for September of 23 are now May of '24. lol...) (I'm annoyed with myself, I ordered last September thinking the est. Ship date was the end of that month, not the end of that month in the following year, I have other parts stranded in that order, and because I paid via PayPal they can't modify the order, all they can do is cancel the order and I'd lose my place in line, and I want the 25 3226's in qfn that came with it - they show as being in stock (I think they were set aside from the rationed lot for my order). They stopped answering my emails about it a while ago), an "Order now and it ships in a few weeks" sounds downright speedy :-P I am very annoyed over how microchip and distributors have handled the QFN 3226's.Or as the case is, how they haven't handled them except in trays of 490 at a time. Making them basically unobtainable to individuals. I will be lucky to have either item in that order by this time next year :-(
At least in the case of my boards, I have to Rev the DD20 board anyway, because I omitted the v-groove and it's on a panel with 3 other designs (covered in photo by the partially assembled DD14 boards).
So far, 1 of those designs is demonstrated defective (I forgot to connect two critical pins in an area of the board too crowded to execute a fix. One of them is the power for the USB physical layer, and the other is power for the PLL. Turns out USB serial adapters need those things powered. (it's a quad serial adapter via FT4232, breaking out all serial and modem control pins on all 4 ports with pullups selectable to 3.3 vs 5 V, a switch to flip PORT1 from serial port to UPDI on PORT0's CTS line (all my recent AVR boards have a CTS to set of solder bridge jumpers defaulting to UPDI, but with options for ground or the default USART0 XDIR, There's also a jumper to switch DTR from being autoreset to going to the USART0 XCK pin, with the idea of making it easier to do combined UPDI+serial and RS485 using the same header), and with jumpers that let you permanently turn port 0 and 3 into UPDI programmers). It wouldn't even say "the last device you connected has malfunctioned") and needs respin too. I'd be lying if I said I was optimistic about the other two (can't test one until i can make it up town to grab a few parts I forgot) though I got a brilliant idea for something to replace the other one on the panel that'll be way better - except to fit it i need some AVR32DD20 in the QFN 20, and just like before, their ordering policies for individuals are "fuck off, we don't sell to insects who can't afford to order 490". I'm mighty tempted to buy a 490 tray, wait until they're all hopelessly backordered, and go into the chip scalping biz. Shoulda done that for the 3226's, still kicking myself over that, I have a close to 100 bare boards manufactured that need 3226's, but it's looking like I can't get them before the end of the year if at all, and might not be able to get the in less than quantities of 490 even then. >.<
Do they think that only big companies can solder QFN? T-962/962A aren't that expensive, and while they struggle with the temperatures needed for 217C SAC305 (and just forget about the stuff with less silver), they work fine for leaded and lower melting solders. With 158C ROHS solder (I have finally after trying close to 30 kinds of solder from a dozen brands, found a lead-free 158 one that seems very good, a "183" lead free that definitely melts below 183C, but makes beautiful joints, and have a "189 lead free" coming (after another brand of "189C lead free"solder than I got turned out to be silver bearing tin lead solder. Excellent tin lead solder, maybe the best I've used, but not lead free). So 1 good candidate and 2 that warrant further exploration, all with melting points above 150C. QFN goes on super easy with the good 158C stuff - much, much easier to rework when some inevitably come out crooked.
I now have a long document on solder that I'll be posting soon, discussing techniques, and solder + flux formulations and their pros and cons, and the scams and dishonesty I've encountered while buying it, and what my go-to fluxes are (I have one for hard tasks and another for less demanding ones that cleans easier) and what brands of solder are trash and what ones are good.
You might want to include any of this as a caveat to your 'DDs shipping' comments:
Don't trust microchipdirect's availability or ship dates on the avr32dd14 (and probably other DDs).
I had ordered 125 AVR32DD14-I/SL backk on 6/18, which they still say "Order now, up to 1,876 can ship on 09-Jul-2022". They sent me a confirmation saying they would ship on 6/25. on 6/27 they sent a delay notice and on 6/28 it said to contact them.
I discovered on my second call with them that they had placed this on a "planning hold" and don't know when they will be able to ship. I wish I knew this before I reved 3 designs to use this chip (based on the fact that I thought I could actually get it). PCBs on my bench, no eta on chips :-(. And due to their swapping of power and ground its not like I can substitute.