<electronic_eel>
yes, I think Silanna is the manufacturer
<electronic_eel>
thanks
<electronic_eel>
I can't find any info on this ic on their website
<electronic_eel>
like when I first found info on this last year or so
<electronic_eel>
so I'm not so sure if floating the entire Glasgow is a reasonable approach
<electronic_eel>
you can get 100MHz isolators for less than a dollar per line
<electronic_eel>
and ti has proper ones who are rated for >5000V and reinforced isolation
<electronic_eel>
so you can use them for permanent connections to mains voltage levels
<whitequark>
a few options
<tnt>
Use a laptop not connected to mains and float your entire test setup :p
<tnt>
(a platic one preferrably :p)
<whitequark>
nooooooo
<tnt>
I was obviously jk.
<electronic_eel>
thats right next to cutting the earth line of your scope...
<electronic_eel>
I compared the concept of the Glasgow to the GreatFET
<electronic_eel>
They have a base board with just the microcontroller and about 80 "dumb" io lines
<electronic_eel>
you are expected not to directly connect to them, but put one of different addons on it
<electronic_eel>
Such a isolation addon would fit right in there and still have enough io after the dir lines
<electronic_eel>
The Glasgow has a basic usable set of functions built in
<electronic_eel>
but it is harder to add stuff like the isolation, analog mux board and so on onto it
<electronic_eel>
Not sure which one is the better concept io/addon wise
<electronic_eel>
Using an fpga + fx2 is the better solution than a fast micro I think though
<whitequark>
yep, different goals
<electronic_eel>
hmm, don't know if the end goals differ that much
<electronic_eel>
it is just the way of reaching them
<electronic_eel>
and how you value ready-out-of-the-box vs. expandability
<whitequark>
yes
<whitequark>
ready-out-of-the-box is a goal for glasgow
<whitequark>
well
<electronic_eel>
but if you sell the fx2&fpga-baseboard as a standard package together with the shifter board
<whitequark>
that's revE, basically
<electronic_eel>
then you still have ready-out-of-the-box but add the ability of easy addon boards
<whitequark>
addon boards are never easy.
<whitequark>
you get combinatorial explosion of complexity
<electronic_eel>
;)
<whitequark>
very few people seem to value managing long-term complexity in a project.
<whitequark>
it's always "oh we can just extend this".
<whitequark>
i'm not happy
<whitequark>
a large part of why i'm not happy is that most of my glasgow work involves looking at consequences of things that were "just extended"
<whitequark>
it really drives in the importance of careful evolution
<sorear>
managing long-term complexity always strikes me as closely related to "clearly understanding the eventual scope of a project, and being willing to say 'X is out of scope' "
<electronic_eel>
yes, a good long term management is what a lot of projects fail at
<electronic_eel>
especially commercial ones
carl0s has joined #glasgow
<electronic_eel>
they just want to sell new stuff to the existing customers
<whitequark>
glasgow has the *possibility* of managing addons like this
<whitequark>
but that possibility has to be carefully explored
<whitequark>
we don't even have the base case covered yet.
<electronic_eel>
but going out and expanding the user base soonish is a good idea IMO
<whitequark>
i'd actually rather have a version with individual isolated outputs than addon boards
<whitequark>
inputs/outputs
<whitequark>
sure, you want addon boards. you're also not the one stuck as the maintainer of the project.
<whitequark>
so you don't really have insight into the amount of work required to make this happen that will have to be expended forever
<whitequark>
if something i feel like the goals might be too ambitious as they are
<whitequark>
look. if that sounds as a downer, consider this
<whitequark>
you have 8 repositories on github. i have almost 200. a lot of these need regular attention. a lot of these i've been giving attention for the better part of ten years
<whitequark>
this is not to say that you're underqualified or anything. this is to say that open source maintainers all eventually end up getting stuck between a rock and a hard place
<whitequark>
honestly, i advise not being in my position if at all possible
<electronic_eel>
don't get me wrong, I'm all for very careful evaulation of any addon concepts
<whitequark>
i think the kind of addon board i would be most open to, is the addon board that could be reused with revE
<electronic_eel>
and I don't want all possible addons now
<whitequark>
that'd use samtec connectors and have dir/io lines on them
<whitequark>
since, well, the plan is to go to one of those eventually anyway
<electronic_eel>
and an isolation addon board is either very limiting (half ios) or intrusive
<electronic_eel>
so I think it is not a good candidate to start with
<electronic_eel>
but the +-30v io protection is a different thing
<whitequark>
i think the ±30V IO protection board is super exciting.
<electronic_eel>
this could be done on a per-port basis and wouldn't need software interaction
<whitequark>
i am completely and entirely in favor of prototyping, qualifying and (if it works well) producing them as soon as possible.
<sorear>
how loosely coupled can an addon board be?
<whitequark>
i would use the hell out of it.
<whitequark>
sorear: i have addons like a passive cable that's a floppy adapter
<whitequark>
or a resistor pack used for connecting to gameboy
<electronic_eel>
I got the depletion fets in the mail yesterday and tried them today, so happy that it worked right from the start
<whitequark>
electronic_eel: i admit, i don't entirely understand how it works
<whitequark>
as you may have realized, i am much better at software/gateware than analog design
<whitequark>
i think i see the general idea
<electronic_eel>
will need to think more about the exact connections
<electronic_eel>
may need to make sense to connect to the 5v rail
<electronic_eel>
may make sense..
<sorear>
unfortunately FETs are 4-terminal devices and it is very difficult for me to visualize an I/V curve in 4D space
<electronic_eel>
the trick are that these are not normal fets, but special depletion fets
<electronic_eel>
they are on by default and if the vgs voltage goes down, the current is cut off
<electronic_eel>
so they work essentially like constant current devices
<whitequark>
and you use a resistor to set the current, right
<electronic_eel>
you could leave the resistor in the middle out
<electronic_eel>
but then the current would be higher
<whitequark>
and only rely on FET resistance?
<electronic_eel>
means higher loss at higher voltages -> temp gets too high
<sorear>
is it not possible to do pinch-off with an enhancement mode?
<electronic_eel>
sorear: yes, but you need some active logic to do that
<electronic_eel>
and that gets complicated quickly if you also consider the negative voltage case
<whitequark>
it's a really clever trick, i think
<electronic_eel>
you'd have a bunch of comparators, voltage level shifters and so on for each pin
<electronic_eel>
these depletion fets are a bit of a rarity
<electronic_eel>
I stumbled over them a while ago when searching for something else
<electronic_eel>
microchip also makes some
<electronic_eel>
but they are higher resistance (like 1k or so)
<electronic_eel>
the point I'm currently not sure about yet is how to burn the excess voltage
<electronic_eel>
in the demo I used the TL431 for it
<electronic_eel>
and set it to 6.something volts
<electronic_eel>
so the current goes through the fets, gets limited there, but there are still 50ma going through
<electronic_eel>
they will go through the diode and into the power rail
<whitequark>
you might not need D1?
<electronic_eel>
the power rail will then rise and needs to be limited
<electronic_eel>
D1 is for redirecting the current from the logic gate into the power rail
<electronic_eel>
or for the negative case to provide a path to ground
<whitequark>
yeah, there is a protection diode to ground
<whitequark>
actually two
<electronic_eel>
you mean in the logic gate too?
<electronic_eel>
these are usually very sensitive and you probably don't want to push 50ma through them
<whitequark>
50 mA is a bit too much
<electronic_eel>
What I'm not sure about is how the shifters will react
<whitequark>
anyway, yeah, a protection addon really should feature its own protection diodes
<electronic_eel>
lets say they are outputting something to the pin, vio is at 3.3
<electronic_eel>
I gues the current will flow in reverse through the shifter into the 3v3 rail
<electronic_eel>
if you now have 8 ports with 50ma each, the current has to go somewhere
<whitequark>
no
<whitequark>
the shifter does not have a protection diode to vcc
<whitequark>
or vio in this case
<electronic_eel>
yes, it doesn't.
<electronic_eel>
but it has a pfet that is enabled
<electronic_eel>
it should be open in both directions
<whitequark>
hmmm
<electronic_eel>
the protection diode is just a thing for input mode
<whitequark>
yes
<whitequark>
you're right
<electronic_eel>
so there must be something that burns this current and makes sure it doesn't rise a bit above of what is set as vio
<whitequark>
yep
<whitequark>
tl431 works...
<electronic_eel>
the tl431 needs it's voltage to be set
<whitequark>
worse yet
<whitequark>
it wouldn't work at 1V8 Vio
<electronic_eel>
exactly
<whitequark>
AP432?
<whitequark>
uhh that's NRND
<electronic_eel>
isn't that just a 431 in different pinout?