bgamari has quit [Remote host closed the connection]
bgamari has joined #scopehal
<azonenberg>
So I had a few minutes before bed and started work on the OpenCL rendering stuff
<azonenberg>
laid a bunch of groundwork but nothing usable yet
<_whitenotifier-1>
[scopehal-apps] azonenberg pushed 1 commit to master [+1/-0/±4] https://git.io/J1XJ3
<_whitenotifier-1>
[scopehal-apps] azonenberg 5fcec32 - Initial skeleton of OpenCL rendering path for analog waveforms. Doesn;t do anything useful yet.
<azonenberg>
There's a preference to enable it, if enabled it will call an empty kernel when you try to render a waveform
<azonenberg>
then fall through to the existing OpenGL renderer to actually give you something to look at
Degi has quit [Ping timeout: 268 seconds]
Degi has joined #scopehal
bvernoux has joined #scopehal
bvernoux has quit [Quit: Leaving]
someone-else has joined #scopehal
<Degi>
someone-else: I mean you could totally have an analog equalizer composed of varactors or something
<someone-else>
would work fine at 2-10ghz frequencies required
<Degi>
Hmm, the step size is a bit large but that sounds neat
<Degi>
Ooh I see, it uses switched capacitors, so it probably has better linearity than a varactor maybe
<someone-else>
yep, they are excellent
<someone-else>
even have models in the datasheet
<someone-else>
you have to translate them into SPICE yourself, but it's quite easy
<someone-else>
don't let min/max frequencies from the datasheet fool you, they actually work fine from dc to light
<someone-else>
as long as RLC values do something useful for you
<Degi>
Hmm, you could put the 600 fF one in series with like a 200 fF capacitor and an inductor such that it resonates around where the amp has its peak and then use that in parallel to the transmission line to have a variable frequency damper. Ideally some variable resistor could adjust the Q factor
<someone-else>
yep
<someone-else>
I have some around so I might bodge something on the existing lmh3401 flex
<Degi>
I think as a variable resistance we should use a piece of tungsten sheet which gets inductively heated to change its resistance
<Degi>
Neat
<someone-else>
Degi: simulated pe64906 with a 10ohm series resistor, does pretty much exactly what is needed
<Degi>
Nice
<Degi>
How is the response otherwise below 6 GHz?
<Degi>
Hm, maybe we can have two with different series resistances (though that might be a bit overkill)
<someone-else>
serial pins are named SCL and SDA but the interface is actually SPI
<Degi>
So usually they would be SCLK, SDI and CS? lol
<Degi>
Figure 10 even shows its SRF vs the state
<someone-else>
they almost trapped me into connecting pe64906 to the i2c bus on another pcb in progress
<someone-else>
"Each bit of the 8-bit telegram is clocked" they call it a telegram, what is going on there..
<Degi>
Lmao
<someone-else>
I'm going to use 8-bit telegrams everywhere from now on
<Degi>
widely supported 3-wire (SPI compatible) interface" connected with SCL and SDA and a maximum frequency of 26 MHz
<Degi>
I2C and SPI kind of give the impression of something standardized, but then one manufacturer has a bidirectional SDIO pin and another mixes it with I2C naming. And isn't the CS line usually inverted?
<someone-else>
obviously they didn't care lol
<someone-else>
on the other hand they have pretty alien SOI technology, so why not have a special serial interface, too
<someone-else>
*SOS (silicon on sapphire)
<someone-else>
pretty interesting lecture from the founder: v
<someone-else>
but otherwise a seriously amazing device
<someone-else>
i'm using it to trim 1meg scope frontend in the works
* GenTooMan
laughs at SPI, "well considering it's the combination of 5 different companies standards no surprise. TI had their protocol Motorola National semi and a few others. As a consequence SPI has is and will be a mess.
<GenTooMan>
so we are stuck with a bunch of proprietary "standard" protocols from the 1980's and 1990's :D
<d1b2>
<Uwe> Peregrine is now PSemi, a daughter of Murata...
<azonenberg>
uwe: First of its name, king of the DC-DCs and the ceramic capacitors?
<someone-else>
azonenberg: they discontinued my favorite LXDC2HL dc-dc converters from under me
<someone-else>
now intel discontinued my second favorite enpirion dc-dc converters
<azonenberg>
My favorite DC-DC these days for general purpose applications is the LTC3374A
<azonenberg>
so flexible
<someone-else>
now looking for a third-favorite dc-dc converter type hopefully in stock and not discontinued anytime soon. Murata is tempting me again with MYRGP/MYRGM
<someone-else>
azonenberg: pretty nice, though I was looking for something with inductors in package
<azonenberg>
I'm using a MYTNA1R84RELA2RA for VCCINT on an upcoming design
<azonenberg>
then an LTC3374 for everything else
<azonenberg>
fed by an OKL-T/3-W12P-C
<azonenberg>
That's my one complaint about the 3374, it won't do 12V in
<azonenberg>
so for a 12V application i have to make a 5V intermediate rail
<someone-else>
cool, I need something for the kintex too
<someone-else>
4A might not be enough though
Famine- has joined #scopehal
<azonenberg>
Yeah this was for a 7k70t
<azonenberg>
For MAXWELL I had an 1/8 brick module picked out, let me see what it was
<azonenberg>
IAF12020A007V-003-R
<azonenberg>
Two of them, actually, since the board has two 1.0V rails
<azonenberg>
1V0_1 drives Kintex VCCBRAM and VCCINT, 1V0_2 drives MGTAVCC as well as VCCINT/VCCBRAM for the Spartan-7 auxiliary FPGA
Famine_ has quit [Ping timeout: 256 seconds]
<someone-else>
I was eyeing 2x EN63A0QI but then again Pat Gelsinger decided no more Enpirion for me
<azonenberg>
(MAXWELL has a 7k160t)
<someone-else>
why, Pat?? such fine dc-dc converters they were
<someone-else>
could've sold them to somebody instead of killing off the entire product line
<someone-else>
7k160t/7k325t in my case
<someone-else>
where I've been shopping recently, 7k325t is cheaper than 7k160t lol
<azonenberg>
:o
<azonenberg>
i've avoided the >160s because they've got 4 digit price tags at digikey and don't work with free vivado
<azonenberg>
if I can't fit in a 7k160t i'd look at a ku025
<someone-else>
vivado license requirement is a downside, yep
<azonenberg>
So basically my list of potential higher end parts is 7k70t, 7k160t, ku025, ku035, ku3p, ku5p
<azonenberg>
but how many of those do you have to buy before you get one that works? :p
<azonenberg>
i've had... not great luck buying chips from unofficial sources
<someone-else>
one or two? i have a handful, seems ok
<someone-else>
my theory is that some chinese contract manufacturer dumped a few reels of these around the time 7-series bitstream security got broken
<someone-else>
and they have been gathering dust on some shelves there since
<someone-else>
works for me since bitstream security is something I don't care about
<someone-else>
btw, lattice announced some new kintex-y parts with fast transceivers which would probably have much saner prices compared to xilinx
<someone-else>
the plan is to move to them in a year or two in case something happens to chinese kintex supply
<GyrosGeier>
ooh
<Degi>
The CertusPro?
<Degi>
They sound very nice
<Degi>
Haha, wasn't there one case where some Arduino alike manufacturer bought ATMega32u4 or so and they were fake and just a solid piece of copper inside
<Degi>
Damn, 5 of then for 200 $
<someone-else>
yep
<someone-else>
same 28nm as kintex-7 but fd-soi
<someone-else>
hopefully reasonably priced
<someone-else>
level of capability to fake kintex-7 parts is so much higher compared to needed for fake arduinos so everybody possessing such tech has, hopefully, better things to do
<azonenberg>
well i'm more worried about speed grade rebins, damaged parts resold as new, etc
<someone-else>
could be but the price differential is so gigantic so it's probably worth the added hassle
<someone-else>
they are also pretty transparent about reballs and new parts with prices being ~2x different
<someone-else>
there's no evidence of re-marking anyway and doing it properly seems difficult with the metal lids
<Degi>
Some sandpaper and an inkjet printer...
someone-else has quit [Quit: Connection closed]
someone-else has joined #scopehal
<someone-else>
well, markings are lasered on
<someone-else>
no ink
<Degi>
I think sometimes on fakes its done with ink (at least on the plastic ICs)
<someone-else>
sure, but these kintexes just look too pretty
<someone-else>
to be fake
<Degi>
Huh, you can get a XC7K325T PCIe card for 350 €