https://open-source-silicon.dev logo
Join Slack
Powered by
<@U017H530WC9> No official specs yet. Just estimat...
# general
t
@User No official specs yet. Just estimates that were discussed here ATM, ~ 35 to 40 pins.
j
Wow, that's awesome
...and what kind of package? I got the impression from @Tim 'mithro' Ansell's talk it was WLCSP... or am I mistaken?
t
Package is a bit fuzzy ATM. WLCSP is/was the target but seems they need to find someone to do the RDL layer and packaging because skywater isn't doing RDL for that process. Alternative would be QFN maybe.
j
QFN would be easier to work with
t
There was talk of providing some of the WLCSP chips mounted on a carrier board that would break out user pins to castellated holes or something of that kind.
Also trying to pay attention to the pinout to make sure you can mount it on "cheap" pcb process (no need to crazy < 4/4 rules or micro-vias or blind vias or stuff like that)
j
i don't like the idea of a carrier board personally...
but a small QFN would be perfect
i have soldered WLCSP in many of my circuits but it's painful and you have a pretty high failure rate if you do it by hand with the kind of equipment a hobbyist has available
t
So far I'm 11/11 success soldering them with just hotair 😛 (and 0.4mm pitch and never soldered any before).
j
well done! many an experienced electronics engineer would have difficulty doing that
t
It'll come down to what gives the most pins, is the cheapest and allows for more applications (some concerns about bonded out packages is increased / indeterminate parasitics for stuff like multi-gigabit serdes, RF stuff, ....)
j
fair enough
for me all i want is a simple circuit with some random logic, a 512 bit shift register, a sin table rom, and a huge resistive DAC!
t
(Just to be clear, I'm not affiliated with google/efabless/skywater, I'm just trying to summerize what was discussed here the past couple of days)
j
that's fine, no need for a disclaimer
i should read through the whole thread
f
@tnt I think you made a good summary. The choice is between W
...WLCSP with good high frequency possibility but limited number of pins and some wire-bonded package like QFN, QFP, ... with much more pins but (serious) problems for high-speed signals.
From experience I think there should be solution for both so cost should not be deciding factor although @Tim 'mithro' Ansell seems to have different opinion.
@tnt @jrsharp There also seem to not be agreement which package is (most) maker friendly. Quite opposite views on it, often very outspoken.
t
I think there is just a wide range of people with different skills that fits in the "maker" term. Most of the one I know I'd qualify as "advanced" and wouldn't be afraid of WLCSP (especially if they have 100+ chips and could take a couple of tries). But at the opposite of the spectrum, there are people that just won't even solder any SMD, not even 1206 and just use through hole ... and then there is everything in between.
6 months ago, I hadn't soldered a single BGA (even after like 10 years of doing my own boards with SMD parts). Now I don't see them as much of an issue ... so it's sometimes also limited by "fear" rather than what you can actually manage.
👍 1
j
Personally I'm not afraid of BGAs... i just find them a bit annoying. I have access to a reflow oven which works well for soldering them. For the design I would like to do, it's not high speed, so a high speed package is not strictly necessary for me, but i'm sure some people will try to implement high speed IPs (like SERDES) and other crazy stuff. But what I will say about WLCSP packages is they can be very small which is good for my application.
Open in Slack
PreviousNext
Powered by