<GenTooMan>
It's hard to guess how many samples they made it depends on the process mostly generally engineering samples are the step before production so if using 8" the wafer scale package is 3.7x4.1 wafers 2100 die per wafer with a yield between 50 and 70 so hmm roughly 1300? I can see that happening, especially for what those things are, their is a big market for something like that.
<azonenberg>
GenTooMan: 50-70% yield sounds very low for small dies on a mass prod process
<azonenberg>
i'd expect 80% or more
<azonenberg>
probably 90s
<azonenberg>
Also, you don't know how many wafers they made. And don't forget to allow for scribe lines between dies
<azonenberg>
with small CSPs the width of the saw lines is a significant contributor to overall area :p
genii has quit [Quit: Hockey and beer time, not necessarily in that order..]
<rombik_su>
TSMC advertising >95% yield on 28nm nodes, but we don't know about any silicon errors. :3
<azonenberg>
You can't put a single yield on a process
<azonenberg>
It depends on a lot of things, like how much of each layer is using minimum feature sizes etc
<azonenberg>
how big your dies are is a big thing too, because defects are per unit area
<azonenberg>
the bigger your die is, the higher the probability of a random defect being on one
<azonenberg>
and the more silicon is wasted due to a single point defect
<rombik_su>
Absolutely, it was estimated for mid-range dual-core SoC project (can't remember the die size), so I guess it will be even better for smaller die.
<azonenberg>
Yeah. With a ~15mm^2 die i'd expect yield at probably 98% or better
<azonenberg>
on a mature process
rombik_su has quit [Quit: Leaving]
emeb has quit [Quit: Leaving.]
emeb_mac has joined ##openfpga
solo1 has joined ##openfpga
ewen has joined ##openfpga
ewen has quit [Read error: Connection reset by peer]
ewen has joined ##openfpga
ewen has quit [Read error: Connection reset by peer]
m_w has quit [Read error: Connection reset by peer]
Bike has quit [Quit: Lost terminal]
OmniMancer has joined ##openfpga
<GenTooMan>
azonenberg I hope you have considered that engineering samples are what’s left over of the first article batch in that thinking.
<azonenberg>
GenTooMan: yes, but fabs dont normally fine tune the process for each customer's wafers
<azonenberg>
unless you're an early adopter, your ES should have the same yield as your production
<azonenberg>
just might have some silicon errata
<GenTooMan>
It's unlikely they would use the full first release as engineering samples. :D
<GenTooMan>
That's really all I was thinking, it's hard to know anything more honestly. Well I wish you a good night, I am up too late.
solo1 has quit [Ping timeout: 258 seconds]
ewen has joined ##openfpga
<omnitechnomancer>
ZirconiumX: Yes you can use luts for add/sub but who knows why there is an invert on one input of the sum unit then, that could be done in LUTs too surely?
<azonenberg>
GenTooMan: the way it normally goes is, you order N wafers for the first batch
<azonenberg>
package a handful of them (untested) for initial testing and bringup work
<azonenberg>
once you're confident there's no fatal flaws, you write a factory test program, test and package the rest of those chips, then ship them off to customers as engineering samples
<azonenberg>
if you're lucky, that same mask set is used for mass prod
<azonenberg>
if early adopters / your in house testing finds bugs, you respin
<azonenberg>
maybe repeat the process sending out ES2 silicon
<azonenberg>
then once ready for prime time, you go to production
<ZipCPU>
A constant low is known as a "BREAK" condition
<ZipCPU>
Ahh ... now I see that gruetzkopf got that.
<ZipCPU>
My first serial ports were full featured, and as feature complete as I could make them. BREAK conditions were part of that support, as were 5-bit, 6-bit, 7-bit, and 8-bit characters, even, odd, mark, space, and no parity, and two stop bits. Then after using the core for a year or so, I realized I never used any of the other features.
<swetland>
yeah the world basically settled on 8N1 long ago. sometimes hardware handshaking.
Jybz has joined ##openfpga
oeuf has quit [Read error: Connection reset by peer]
oeuf has joined ##openfpga
dh73 has joined ##openfpga
genii has joined ##openfpga
dh73 has quit [Read error: Connection reset by peer]
Sinclair2 has quit [Quit: Bye Bye]
X-Scale` has joined ##openfpga
X-Scale has quit [Ping timeout: 268 seconds]
X-Scale` is now known as X-Scale
_franck_ has joined ##openfpga
Jybz has quit [Ping timeout: 248 seconds]
Jybz has joined ##openfpga
<azonenberg>
swetland: yeah my uart ip is 8n1 only
<swetland>
baudrate is pretty much the only tunable I bother including for uarts
ZombieChicken has joined ##openfpga
<GenTooMan>
for bit rate generation it's possible to make fractional generators for a wider range. It's uncertain however that will be ever used. It is amazing that UARTs have been around for more than or almost 70 years.
dh73 has joined ##openfpga
<TD-Linux>
break condition is mainly used on rs-485 and dmx512
<gruetzkopf>
these days, sure. on old sparcstations, break on console was equivalent to <stop>-a on the keyboard and dropped you into the openfirmware prompt (which is usable as a machine monitor)
<swetland>
I usually do 1-3mbps -- usually easier to generate than 115200, well supported by ftdi usb/serial adpaters, etc
Jybz has quit [Quit: Konversation terminated!]
emeb_mac has joined ##openfpga
<TD-Linux>
oh wait also I think on dmx512/rs-485 it's not actually break, but a delay with the signal *not* asserted (high)
emeb_mac has quit [Quit: Leaving.]
emeb_mac has joined ##openfpga
emeb_mac has quit [Quit: Leaving.]
_franck_ has quit [Ping timeout: 240 seconds]
* pie_[bnc]
wonders how much it takes to make a CAD with a business case...
<shapr>
I kinda miss openfirmware
dh73 has quit [Quit: Leaving.]
<gruetzkopf>
has anyone here built *sd*-sdi output support yet? i have stupid plans and old hardware
Bike has joined ##openfpga
Asu has quit [Remote host closed the connection]
rombik_su has quit [Quit: Leaving]
solo1 has joined ##openfpga
<mithro>
gruetzkopf: poke felix_ maybe?
renze has quit [Quit: Spaceserver reboot?!]
sgstair_ has joined ##openfpga
sgstair has quit [Ping timeout: 260 seconds]
<felix_>
gruetzkopf: nope, i'm only working on hd/3g/6g sdi
<felix_>
ok, tbh the project is a bit stalled at the moment, but it's still on my todo list and from time to time i do get around to continue to work on it a bit
renze has joined ##openfpga
<felix_>
oh and if you want to implement sd sdi, i'd highly recommend not to try to find the smpte specs, but use the itu recommendations
<felix_>
the itu stuff is much more useful and contains information the official spec somoehow lacks...
sgstair_ is now known as sgstair
<ZirconiumX>
You know, it never ceases to surprise me how well the ECP5 keeps up with the Cyclone V.
<ZirconiumX>
Despite the CV being LUT6 to the ECP5's LUT4