<azonenberg>
i think this is everything but the psu
<azonenberg>
We have a whopping three unused pins on the MCU i can hook up to debug headers, LEDs, or some combination thereof
<monochroma>
sounds like it's begging for an RGB led ;)
<azonenberg>
Lol no its probably getting three green 0402s
<azonenberg>
I dont want to try and deconvolve RGB to figure out which status bits are set :p
<monochroma>
boring! ;)
_whitelogger has joined #scopehal
_whitelogger has joined #scopehal
<electronic_eel>
azonenberg: you can use 2 inverters in series with feedback from the output of the 2nd to the input of the first as latch, see the schematics I posted
<azonenberg>
what do you think of the version i linked?
<azonenberg>
using a dff+sr with clock tied low as a sr latch?
<electronic_eel>
it has the advantage that you can steer the setting at powerup with resistors
<electronic_eel>
but you have a mcu so that doesn't matter
<electronic_eel>
the '74 you used works too
<electronic_eel>
but there is a problem: the 393 has -5 as vss, so it pulls OVERVOLTAGE_N to -5 in case of overvoltage
<electronic_eel>
the '74 won't like that
<azonenberg>
Neither will the stm32 :p
<azonenberg>
Good catch
<electronic_eel>
you can either add a resistor and diode to shield the '74 from the negative voltage or use a fet and invert
<electronic_eel>
the pullup on RELAY_EN (R11) is now unnecessary as it is driven actively by the output of the '74
<azonenberg>
Deleted
<azonenberg>
ok so now my one action item on the input is fixing the -5V OVERVOLTAGE_N issue
<azonenberg>
As far as the power supply goes, my tentative plan is to have an inverting DC-DC that goes from +12 to -6.5V
<azonenberg>
then have two LT3093s, one for -5 and one for -6, off that. The 3093 has only 300 mV dropout so 500 should be a decent bit of headroom especially since the -6V rail is the lightest load
<azonenberg>
the gain stage is by far the most power hungry part of the system and that's a single +5V supply part
<azonenberg>
Then on the positive side, i have +6V, +5V which is the most power hungry, then +3.3 for the MCU, +0.9 and 2.5V common mode references
<azonenberg>
and +1.8 which is a clip rail that needs a source-sink regulator
<azonenberg>
plus +/- 0.5V reference gives me ten rails total
<azonenberg>
The +/- 0.5V references don't need to be super tight so i'm just going to use resistive dividers off the +/- 5V rail
<electronic_eel>
about the dropout: the features of the lt3093 will degrade with lower dropout
<electronic_eel>
better give it 1v of dropout to keep the low noise and good psrr
<azonenberg>
They say 190 mV dropout at 200 mA typical, 330 mV max in the datasheet
<azonenberg>
i thought 500 mV would be enough headroom
<electronic_eel>
yeah, the value they give as dropout is when it really becomes shitty
<azonenberg>
lol is this from experience?
<azonenberg>
i'm trying to find a graph of PSRR vs headroom or similar
<electronic_eel>
there is a graph in the datasheet, pssr vs. voltage diff
<electronic_eel>
page 12
<azonenberg>
aha. wow
<azonenberg>
yeah so we really do want 1V of headroom
<azonenberg>
ok so -7V is the intermediate rail then
<electronic_eel>
it works down to 330mv, but if you want the cool performance you pay for you need to give it 1v
<electronic_eel>
it is a bunch of rails you need...
<azonenberg>
Yeah
<azonenberg>
So, high level architecture i'm looking at at this point for the power subsystem
<azonenberg>
+12 -> +7V switching intermediate rail, then TBD LDOs for +6 and +5 off that
<electronic_eel>
how do you plan to create the positive voltages? where do you want to use switchers?
<azonenberg>
then tentatively a second SMPS off the +12 rail to produce 3.3 for the MCU. The MCU pulls enough current that I'd rather not LDO it all the way down from 7V, but i want to crunch power numbers before making a final decision there
<azonenberg>
a 2.5V reference fed by the 3.3 rail for the 2.5V common mode reference, then resistive divider off that for the 900 mV common mode reference (nobody makes 900 mV reference ICs)
<azonenberg>
Then TBD inverting switcher for the -7V rail, LT3093's for -6 and -5 from that
<electronic_eel>
the mcu will be using like 3mA or something when you use wfi, like 10mA without
<azonenberg>
hmm i swear i saw close to 100 mA in the datasheet with everything running but that does sound high. i must have been thinking of the gain stage or something
<electronic_eel>
that is at least what my 'f031 need at 48mhz
<azonenberg>
So the 3.3 rail will be LDO'd from the +7 then
<azonenberg>
anyway, then the +/- 0.5V references will use resistive dividers off +/-5
<azonenberg>
and finally, we need a positive source-sink regulator for 1.8v
<electronic_eel>
since you are using lt3093 for negative, I suggest to stay with lt and use lt3042 for positive
<azonenberg>
is that the positive counterpart? similar specs?
<electronic_eel>
yes
<azonenberg>
Ok
<azonenberg>
any ideas on the 1v8 rail then?
<azonenberg>
i was thinking maybe a ddr termination regulator but most of those seem meant for lower voltage. havent put any time into it yet
<azonenberg>
basically we just need something we can clip the final gain stage to that can sink current
<electronic_eel>
there aren't many sink/source regulators out there
<electronic_eel>
how precise do we need it to be?
<electronic_eel>
maybe just a shunt regulator like a tlv431?
<azonenberg>
absolute max of the hmc8012 input is +2.3V on either analog input
<electronic_eel>
if we plan 0.4v for the diodes that would be 1.9v, so doesn't need to be that precise
<azonenberg>
i'm getting there, sec
<azonenberg>
if we assume the lmh6552 is zero output impedance worst case, 5V into 100 ohms gives 50 mA with the amplifier saturated. The amp can easily supply this (+/- 141 mA short circuit current) so let's plan for that as the worst case shunt current
<azonenberg>
realistically it will be less but
<azonenberg>
and then we only will be doing this on one leg at a time as the output is differential, so we can never have both above or both below 0.9V at once
<azonenberg>
At 50 mA, yeah 400 mV across the schottkies sounds about right per figure 1
<azonenberg>
So basically +/- 50 mV or so of 1.8V is likely OK
<azonenberg>
absolute max of the tlv431 is 20 mA if i'm reading the datasheet right
<azonenberg>
recommended 15
<azonenberg>
we need something that can safely sink 50
<electronic_eel>
so you could use a tlv431 with a npn afterwards
<azonenberg>
yay discrete BJTs. my biggest weakness
<electronic_eel>
they don't have the example in the datasheet of the tlv431
<electronic_eel>
but it is in the datasheet of the tl431
<electronic_eel>
the tl431 works like the tlv431, just different values for the reference, current capability and so on
<azonenberg>
The tl431 can sink 100 mA by itself
<azonenberg>
no external transistor needed
<electronic_eel>
yes, but it has a 2.5v ref, can't go below
<electronic_eel>
need the tlv431 for 1.8
<azonenberg>
ah ok. so you're thinking tlv431 in the figure 31 topology?
<electronic_eel>
yes
<azonenberg>
any suggestions on the transistor / requirements for it?
<azonenberg>
other than thermal
<azonenberg>
BJTs are *really* not my thing :p
<electronic_eel>
not too huge, because they tend to get slower then
<electronic_eel>
maybe something in sot89
<azonenberg>
ok well look around if you have a chance. I'm going to sleep, tomorrow i'll have my work cut out for me doing all of the other supplies
<electronic_eel>
I can take a look at some transistors
<azonenberg>
ok great
<electronic_eel>
the other alternative would be a power opamp
<azonenberg>
Yeah thats the route i was thinking of going before you suggested the tlv431 + bjt option
<azonenberg>
btw i used my probe to debug a customer's board earlier. gave pretty good results, waveform was nicer looking than the pico ta061 for sure
<electronic_eel>
cool
<azonenberg>
and the waveform tracked very well with the lecroy zs1500 i've been using as a "gold standard" probe
<electronic_eel>
I hope the probe tips and accessoires will work well
<electronic_eel>
I think that part is important for day-to-day use
<azonenberg>
This was my first time probing something other than a synthetic test signal with it
<azonenberg>
The only problem i had was the wobbliness of the long skinny Pico tips in the current tip socket
<azonenberg>
ergonomics were quite nice otherwise. And i'm respinning the board with shorter PMK tips in a tighter fitting socket so this won't be an issue
<electronic_eel>
yes, the wobbles, the ground leads, how to clip it to a soldered-on wire and so on
<azonenberg>
I have a minimal accessory kit for ten probes (one ground blade, one flexible ground lead for attaching to a square pin header, and one stiff wire ground for point probing, plus one tip) en route from PMK now
<azonenberg>
one of the things i will be evaluating in the beta is if these accessories are good, if i should add more, if any of them are redundant, etc
<electronic_eel>
do they offer small clips or similar to probe the signal off a small wire?
<azonenberg>
They can be obtained, i didn't buy them for this round
<electronic_eel>
would they fit into the same socket on the probe
<azonenberg>
Yes. This is the full list of accessories
<electronic_eel>
i swear i saw these grabbers somehwere else too, i think they source them from some other manufacturer
<azonenberg>
My minimal kit for the beta units is one ground blade, one solid tip, one Z-ground, one 7cm flex ground lead, and (bought from tequipment, not the PMK distributor, for nearly half the price) a bipod positioner
<electronic_eel>
i often solder small tinned wires to testpoints and grab the signal off them
<electronic_eel>
I find that is more stable than using a bipod positioner
<electronic_eel>
in a longer debug session I have to shuffle stuff around, like plug in something, solder/unsolder some resistors and so on. the bipod needs to be repositioned every time, a hooked wire not
<azonenberg>
if you do that, just buy a 0.75mm machine pin, e.g. ED1140-ND, and solder that directly to your test point wire
<azonenberg>
that will mate directly with the tip socket on the probe
<electronic_eel>
good idea
<electronic_eel>
gives probably a better signal than the grabber
<azonenberg>
anyway like i said, one of the things i want to do with the beta units is figure out what the best accessories to include is
<azonenberg>
i also want to see if i can get the whole tetris standard accessory kit, vs individual items, for a better price
<electronic_eel>
yes, selecting the right set of accessories is a good point
<electronic_eel>
also include preferences from beta testers, not just how you like to use it on your own
<azonenberg>
Exactly the idea
<azonenberg>
anyway, i'm off to bed. if you make any progress on the 1.8v rail, great
<electronic_eel>
have a good night
<Degi>
What is the 1.8 V for?
<Degi>
What does "Stencil Life: 4 hours at 20-25 °C" mean for solder paste?
<monochroma>
i believe that's the working life of the paste after you remove it from the container and do the stencil application on a PCB
<Degi>
Oh so I need to oven it 4 hours max. after applying?
<monochroma>
yeah, the solvent evaporates
<Degi>
Hm I guess I can just attach a temperature controller to a toaster oven to get a reflow oven?
<monochroma>
azonenberg's is a 100% stock convection toaster oven
<monochroma>
but yes, there are a lot of reflow oven controller projects out there
<Degi>
So eyeballing when the solder melts is good enough? Even for BGA? Well I guess then a toaster oven is on my shopping list if I can find that anywhere
<monochroma>
you tend to want to characterize the oven and can do timed profiles, but basically yes.
<Degi>
Hm I could stick a thermometer into it and record how it heats up and cools down
<monochroma>
thermocouple(s) on a dummy board are usually what is used. azonenberg has a reflow oven profiling board design
<electronic_eel>
Degi: the 1.8v are for overvoltage protection of the adc
<electronic_eel>
if you set a too high gain for your signal, the output of the final stage will go above the 1.8v the adc uses
<electronic_eel>
so the 1.8v is not a normal regulator designed to deliver 1.8v, but to sink anything that wants to lift it above 1.8v
<electronic_eel>
I'm actually right now working on the circuit for the 1.8v rail
<azonenberg>
degi: it's not "reflow 4 hours after applying"
<azonenberg>
it's "stencil 4 hours max after putting it on the stencil"
<azonenberg>
it stops spreading properly after that point
<azonenberg>
The lifetime from printing to reflow isn't specified, as that's normally fairly quick
<azonenberg>
but in a production environment you often have paste sitting on a stencil for a long time as you do many board prints off the same stencil
<azonenberg>
so that is where the lifetime issue comes in
<azonenberg>
Degi: Auber Instruments sells a very nice pt100 RTD that's super tiny and low thermal mass and has reflow compatible wiring, that's what i use - but almost anythign would work
<azonenberg>
I characterized my oven to *confirm* that the profile i was already using was good
<azonenberg>
But i had already been doing boards with good yield just by eyeball
<azonenberg>
i have a stock convection toaster oven that gets slightly higher than the temp on the thermostat
<azonenberg>
I set it to "cookies" and "450F" which is "232C", wait until solder melts, give it another 30 sec, then open the door and let it cool
<azonenberg>
you end up with a beautiful ramp-to-spike profile that peaks around 245-250C
<azonenberg>
I have a report to work on for $dayjob and will work on the psu after
<Degi>
Oh neat. 250 C? Do you use SnBi paste?
<Degi>
Oh 4 hours after applying to stencil before spreading. Interesting.
<Degi>
(I mean the solvent evaporating shouldn't be a big problem unless the solvent has flux-like activity)
<azonenberg>
yeah. after that, it's too sticky to print right
<azonenberg>
well, you want some stickiness when placing components too
<azonenberg>
so they dont fall off as you put it in the oven etc
<azonenberg>
just from handling the board
<Degi>
Hm yes
<azonenberg>
i would say 4 hours after opening the container any unused paste should be discarded and any boards you printed should be in the oven
<Degi>
Hm I think I'd get paste in a syringe, that it doesn't dry out as fast. Also what's with the shelf life? I found one paste with 6 monts at 0-10 °C, can I theoretically stick that in the fridge at -20 °C and make it last longer?
<azonenberg>
I use TS391SNL10-ND paste made by chipquik. It's type-4 spheres of standard SAC305 alloy with a no-clean synthetic ROL0 flux
<azonenberg>
Shelf life 12 months at room temperature
<azonenberg>
that's a *minimum*, they say it's likely good longer than that but if >1 year old to do a test run and discard if you see printing problems or it looks dried up or clumpy
<Degi>
Huh neat, Sn96.5 and works at 250 C
<Degi>
I tried making my own paste, a bit of a disaster.
<azonenberg>
"work at 250C" is not surprising
<Degi>
Hm your reflow profile is it just gets hot to peak temperature with no plateau before that?
<azonenberg>
235-265C is iirc the process window for sac305 reflow
<Degi>
Hm I thought about using that for doing double sided SMD with the other side being Sn alloy with a higher melting point
<azonenberg>
this is actual measured data off my oven. The rate of cooling is a little slower than i'd like, 3-4C/sec is better
<Degi>
Is 1 °C/s a good rate?
<Degi>
Well its the average rate of the datasheet ramp up, should be fine
<azonenberg>
1C/sec ramp is right in the middle. standard recommendation is 0.6 - 1.8 for ramp to spike
<azonenberg>
this is with no ramp rate control, it's the natural rate of heating of my oven with the elements on
<azonenberg>
plus the thermal mass of the board
<azonenberg>
Standard procedure for 2 sided reflow is to declare one side the "bottom" (does not need to be physically the bottom in the final assembly)
<azonenberg>
have it be mostly decoupling caps and other low mass components
<Degi>
Yes that was my idea, for decoupling
<azonenberg>
stencil paste, place components, reflow upside down (components on top)
<azonenberg>
If you must place large components, glue them. There's two schools of thought here
<azonenberg>
one is to use a thermoset epoxy that cures at reflow temps and have it cure automatically during the reflow profile
<Degi>
And the surface tension just keeps the SMD parts there when you turn it over, right?
<azonenberg>
the second, my preference, is to let them self align and solder at the lowest energy position, then glue them and do a second, much cooler oven cycle at around 125C for a minute or two to cure the glue
<azonenberg>
because if you place glue +paste at the same time, any placement issues become permanent
<azonenberg>
the glue cures before paste melts so the parts don't self align
<azonenberg>
anyway, then you just turn the board right side up and do the top side liek normal
<Degi>
Hm is the glue pre-mixed epoxy that cures at approx 100 C?
<azonenberg>
Yes. I use AD1-10S, also from chipquik
<Degi>
I think I can stay with small components for the bottom, 0402 decoupling caps and so on below BGA packages.
<azonenberg>
it's explicitly meant for this purpose. and sorry i meant 150C not 125. 90-120 sec cure at that temp
<Degi>
Oof that's pricy epoxy, but I guess you won't use more than a few µL per part anyways
<azonenberg>
pricey??? $10 for a syringe that will last you years?
<Degi>
Idk its almost 1 € per ml. I have some epoxy here that's like 1 € per 50 ml...
<azonenberg>
we're talking fractions of a cent per board
<Degi>
Yes I think I'll get it
<azonenberg>
Lol. meanwhile the aa-bond f110 i use in rework is $3.99 for a 2.5 gram premeasured pouch
<azonenberg>
although most of that is packaging, 250 grams in bottles is $65.99
<Degi>
Can you reseal the pouch? That's a bunch of glue heh
<azonenberg>
No. Unfortunately they don't make it smaller than that
<azonenberg>
so i try to batch rework and do a bunch of boards at once
<azonenberg>
but when i'm charging a customer $150/hr for my time, $4 of glue is not even an expense i worry about billing them for
<Degi>
Lol yeah
<azonenberg>
re surface tension there is a formula for relating the sum of SMT pad perimeters of a component and an alloy dependent constant for surface tension to calculate the max hold-down force of a component during the second reflow cycle
<azonenberg>
I don't usually bother as i have a gut feeling for parts that are "definitely ok with plenty of safety margin"
<azonenberg>
and anything even remotely questionable i glue
<Degi>
I mean idk I guess 0402 is fine, maybe up to 1206? if its a flat part
<azonenberg>
e.g. a samtec q-strip is getting glued, i wont even bother running the numbers
<azonenberg>
substantially all smt chip components are fine
<azonenberg>
a small QFN/DFN/WLCSP is likely fine, i'd glue a large TQFP (or better yet not use TQFP at all, i hate them, but for unrelated reasons)
<azonenberg>
a big tantalum or certainly a can electrolytic would probably be glued
<Degi>
Lol TQFP works sometimes?
<azonenberg>
a smaller tqfp, like a 32 pin or so, i think you could get away with especially if it had a thermal pad. but it's close enough i'd likely either glue it or run the math to be safe
<Degi>
Hm okay thermal pad...
<azonenberg>
if memory serves me right, i've done KSZ9031s on the underside without gluing
<Degi>
When you apply glue after reflowing it, do you just put it over the pins for TQFP / around the chip for the other packages?
<electronic_eel>
I wouldn't glue over pins
<azonenberg>
I'd just do a dot at each corner
<Degi>
Okay
<azonenberg>
you want to keep pins free for probing and/or rework if there's a solder defect you missed
<azonenberg>
for q-strips i do a C-shape of glue at each end then a little blob between each group of pins
<Degi>
And I guess SnBi solder runs the risk of the board failing faster when mechanically load cycled?
<Degi>
Hm is it possible to remove epoxied parts without damaging the PCB?
<azonenberg>
Yes. I would only recommend using low melting point solder on stuff that cannot survive a normal sac305 reflow profile
<azonenberg>
e.g. some led lighting
<azonenberg>
re removal of epoxied parts, you can machine it off? :p
<azonenberg>
i'd do it if i had to
<azonenberg>
but for the most part, the way i design boards i only need epoxy for large bottom-side connectors
<azonenberg>
and i've never had to take one off post reflow
<Degi>
Hm yes, I think for the scope PCIe board I only need a few decoupling caps on the underside. But that's good to know for bigger projects...
<azonenberg>
anyway, in order to start pipelining projects given the extended fab lead times in these times
<azonenberg>
as soon as the afe characterization board goes to fab, i'm going to start the respin of the probe. maybe won't order it quite until i've characterized the current version on the new test board, but i want gerbers ready to order with the final ground and probe tip designs
<azonenberg>
the other thing i want to start is design of the final 4x afe + adc board, including finalizing a connector pinout so we can start doing mating boards
<Degi>
Yes that sounds good
<Degi>
(Also are there any plans for probe power supplying? Like +- 6V? I'd like to design a fast high impedance probe sometime)
<azonenberg>
Active probes are definitely going to be happening. Since SMA is not a push-on connector i can't use something fancy like lecroy probus
<azonenberg>
i definitely want to have front panel power ports for active probes. you'll use a SMA cable of your choice and then some sort of multi-conductor cable for power
<electronic_eel>
I'm not sure if +- 6V is enough. as we've seen with the afe char board, you sometimes need 7v, other circuits may need more. Maybe +-15v like on some other scopes?
<azonenberg>
tentative thought is at least 3 conductors for +- power and ground
<azonenberg>
+/- 12 is what probus supplies, we can regulate from that
<monochroma>
azonenberg: BMA! :D
<Degi>
Heh BMA looks fun
<Degi>
Meh as long as the probes dont rip out lol
<azonenberg>
i'd likely want i2c as well for controlling any gain/offset or discovery parameters in active probes. So we'd be looking at something very similar to probus in terms of available signals (+/- 12V, ground, i2c, and maybe a 3.3V vccio rail for the i2c)
<azonenberg>
unsure on what the connector and cable would look like at this time
<Degi>
Hm maybe some alternative which only needs 1 data pin
<azonenberg>
once you have 3 power pins anyway, one vs two data lines is a non issue
<azonenberg>
just a question of what the physical connector will look like, i don't want something too massive
<azonenberg>
for ergonomic reasons
<azonenberg>
anyway, thats something to think about when we do the final system but doesn't have to happen at this phase of the design
<azonenberg>
my tentative plan is to design an active probe in parallel with the full system so i can sanity check everything. Most likely do the probe in isolation and make a standalone psu module for it
<azonenberg>
so i can then hook the probe to the afe test board, the output of the afe to my "real" scope, and do end to end system performance testing
<electronic_eel>
azonenberg: did you see the link I posted above for the 1.8v shunt regulator?
<azonenberg>
skimming now, but have to work on a report for $dayjob before i spend too much time on this
<electronic_eel>
do you consider the transistor in TO-252 too large for the afe?
<electronic_eel>
I ordered this part for my io protection board, there I need a bit more current sinking capability than you
<electronic_eel>
so I had it on hand and could do actual prototype testing
<azonenberg>
at a quick skim that looks like a good design. re the physical size of the transistor, i'll look around after i finish the report
<electronic_eel>
another transistor will most probably work, but maybe you need to tweak the feedback cap a bit
<awygle>
I have 2u of rack space at the top of my rack, a 1u PDU and a 1u shelf
<awygle>
Should I put the PDU in slot 13 and the shelf in slot 14, vice versa, or do I really need to rearrange the whole rack so the PDU is in slot 1 at the bottom?
maartenBE has quit [Ping timeout: 260 seconds]
<azonenberg>
awygle: what is going on the shelf?
<azonenberg>
and what is being plugged into the pdu?
maartenBE has joined #scopehal
<awygle>
Lots of stuff (SA, scope, PSU at least) will go into the PDU
<awygle>
The shelf is just a shelf. Probably nothing too heavy.
<azonenberg>
On my 4-post racks i normally put the PDUs mid-height on the back of the rack
<azonenberg>
so i dont have super long cables going to the top/bottom
<azonenberg>
electronic_eel: so i think in order to maximize throughput given the lead time etc
<azonenberg>
i'm also going to try and build a HMCAD1520 board that mates one of my existing fpga devkits, probably the starshipraider x1 prototype
<azonenberg>
the idea being that i can pair this with the afe characterization board and get a single channel scope that i can actually acquire data from
<azonenberg>
and start with some limited firmware dev, software side driver support, etc
<azonenberg>
while the actual scope boards are in production
<azonenberg>
I think the entry level scope is probably going to have a digital trigger based on the ADC only. This is more jitter, but shouldn't be a huge issue at the sample rates we're working with
<azonenberg>
I might implement a simple interpolator that fits a curve or linear slope to the ADC data and calculates an ideal time of the edge with sub-sample precision
<azonenberg>
the midrange and high end ones will have a comparator based trigger with TDC so i can actually measure time of the edge crossing between samples rather than interpolating
<Degi>
What is TDC?
<Degi>
I had the idea of setting a SR latch with a comparator and then resetting on clock cycle and then measure the time interval by charging a capacitor and sampling that with a high resolution ADC to get precise timing. (Using two SR latches on I and Q of the clock, that at least one capacitor is somewhat charged)
<azonenberg>
Time to digital converter
<azonenberg>
What you're describing sounds like one possible way to build one
<azonenberg>
but basically the idea is to have a level trigger that measures the phase of the clock wrt the level crossing of the signal
<azonenberg>
and horizontally offsets the waveform by a fraction of one sample based on that
<Degi>
Ah yes
<electronic_eel>
aren't there TDC ics available off the shelf?
<Degi>
You could use that for ET sampling too
<azonenberg>
Haven't looked. It's somethign i consider not cost effective for the entry level scope
<azonenberg>
LeCroy scopes do this in hardware, and use it both for display and random interleaved sampling
<azonenberg>
not sure about other scope vendors. The LeCroy driver fully supports this, and the waveform data model has all the necessary fields
<Degi>
Lol 12 ns tho
<azonenberg>
Yes. Slow TDCs exist
<azonenberg>
But 55ps resolution for example, when a 10 Gsps scope samples at 100 ps per point
<azonenberg>
barely buys you half a sample of resolution. that would give you... not quite 20 Gsps equivalent time performance
<azonenberg>
absolutely not worth the cost / firmware overhead / board space
<Degi>
Hm 55 ps is not a lot
<electronic_eel>
can't you use a fast comparator and connect that to a fast input on the fpga instead of a dedicated tdc ic?
<Degi>
If you charge a capacitor in 10 ns and use a 16 bit ADC you could get like 1 ps
<azonenberg>
fast input and do what with that fast input?
<azonenberg>
Degi: That's the sort of thing i'm talking about. Basically you need to convert the clock into a triangle/sine wave, have a sample and hold triggered by the level trigger
<Degi>
You could implement a SR latch in the FPGA and then lead the pulse out to charge a capacitor and ADC it
<azonenberg>
then ADC that
<azonenberg>
The challenge is getting GHz bandwidth and fast enough trigger times
<azonenberg>
It's going to happen, but not in this scope
<azonenberg>
possibly the midrange, definitely the high end
<Degi>
Hm the SR latch may need to be outside of the ADC though, especially for 100 ps pulses, I think for example the ECP5 can't deal beyond 1 GHz signal input, so it would totally miss a comparator going high for 100 ps
<azonenberg>
Yes, it needs to be fully discrete logic. The FPGA won't be able to keep up
<azonenberg>
My waverunner 8000 series scope is 200 Gsps in random interleaved mode. Which means the TDC has 5ps resolution
<Degi>
Hm does the HMC675LP3 "latched" mean that it latches on once the threshold is reached?
<azonenberg>
and that's on a 10 Gsps srealtime rate
<azonenberg>
No. There's a D latch on the output
<electronic_eel>
yeah, I think it is better to get the entry level scope working without a fancy trigger than taking the time to develop it
<azonenberg>
that is either in transparent or latched mode
<Degi>
MY TDS520 says 100 GS/s in interleaved mode (it has 500 MS/s ADC) (but I didn't get it to work beyond 1 GS/s at all)
<azonenberg>
electronic_eel: My plan is to use the proceeds from selling the entry level scope to fund development of the higher end ones
<azonenberg>
So yes, shipping the first model as soon as we can get it working reliably, with nice user friendly software etc, is a must
<Degi>
Hm do we have a rough cost estimate?
<azonenberg>
BTW, what do you think of the rackmount enclosure having a small touch LCD (like 1x3") for initial IP config etc, as opposed to a uart console?
<Degi>
That'd be nice
<azonenberg>
it would be off by default, wake up when touched
<Degi>
I mean it could have both
<Degi>
Hmm does UART mean RS232 port?
<azonenberg>
and just have a few menus for setup
<azonenberg>
TBD. Possible options include cisco RJ45, DB9, and integrated ftdi
<Degi>
And maybe have the display show the sampled waveform
<azonenberg>
No. It will entirely be out of band setup
<Degi>
Hm okay
<azonenberg>
it's not intended to be a UI you use frequently
<azonenberg>
its sole purpose is to get the scope on the network
<azonenberg>
either for static ip config or to find out what dhcp address it got
<Degi>
I mean mostly to look cool. Like why else would it have a nice gigabit network connection lol
<azonenberg>
i plan to build something like that for the clock generation project and other stuff too
<azonenberg>
so its a piece of hardware / firmware i need to do anyway
<Degi>
I think showing the IP address for DHCP / allowing to enter one would be really useful. That way you don't even need to look into your DHCP provider device to find out the IP...
<electronic_eel>
I'm not sure about touch. I prefer a rotary encoder (which you can also press to ack a value)
<Degi>
I'd prefer a keypad
<Degi>
But tbh to enter an IP touch should be fine
<electronic_eel>
a full keypad may be a bit large for a 1U thing
<Degi>
With small buttons like on a flipphone
<azonenberg>
electronic_eel: yes thats my thought. i dont like touch a ton either but this is literally only for ip config
<azonenberg>
its not something you will use often
<Degi>
Flip flops seem to cost a good buck for 10+ GHz
<electronic_eel>
there should also be leds next to each channel to show if the channel is enabled or if there is overload
<Degi>
Meh if your ADC is a few thousand, not like that matters
<azonenberg>
degi: exactly
<azonenberg>
that is going to be a premium, spare-no-expense scope
<azonenberg>
With a price tag probably on par with a "real" scope, especially in low volume
<Degi>
*browses rad hardened parts*
<azonenberg>
lol
<Degi>
Like meh parts like a few meg of ram get absurdly pricy in tens of k$
<azonenberg>
electronic_eel: i'm thinking RGB LED per channel
<azonenberg>
disabled = off
<azonenberg>
active = color code by channel ID
<azonenberg>
overload = blinking red
<electronic_eel>
wfm
<Degi>
Hm maybe restrict color code to G and B and have R on when its overloaded
<Degi>
(Idk I dont really like blinking LEDs, especially if they're too slow)
<azonenberg>
Degi: i want to have a standard color code for channels so you can ID probes etc
<electronic_eel>
I don't know how often you plan to overload your inputs degi...
<Degi>
hmm
* Degi
orders some 60 dB input attenuators
<azonenberg>
the front panel print overlay will most likely have a colored ring printed around each SMA
<azonenberg>
to make it more obvious which one is which
<azonenberg>
but if i'm going to have indicators, i'll light up the active channels
<azonenberg>
the overload indicator would probably be 50% duty cycle 1 Hz or similar, and keep blinking until the channel is turned off then re-enabled
<lain>
could have overload be rapid blinking of the channel color
<lain>
normal condition == solid on
<azonenberg>
that seems like it could be confused with activity indicators
<azonenberg>
maybe alternate channel color with red?
<lain>
hm yeah
<Degi>
How about making it customizable?
<azonenberg>
and a bit of a gap in between, for the pink channel?
<azonenberg>
Degi: i mean the firmware will be open source
<Degi>
I'd suggest having an extra overload LED though. Maybe a bit brighter.
<azonenberg>
and the leds will be rgb so you can do whatever you want :p
<azonenberg>
I'm not going to go as far as... broadcom, i think? went on some switch chipsets
<lain>
my problem with customization in instrumentation is that in a lab setting, one person might have things setup one way, another person on another bench might have theirs set a different way, and then confusion abounds when they try to use each others' instrument
<azonenberg>
and literally put in a separate MCU just for running indicator LEDs
<Degi>
How about making the whole front panel a mix of OLED and eInk
<azonenberg>
because each oem had their own idea of how they wanted to do leds
<lain>
hehe
<azonenberg>
they gave up on doing it in silicon and built a mcu that literally just runs port indicators
<azonenberg>
and made a blob for each oem :p
<Degi>
Lol
<lain>
LOL
<electronic_eel>
can we detect which kind of probe (or even individual probe with cal data) is connected to a channel? for example with nfc tags next to the sma?
<Degi>
Tbh a hardwired indicator LED for the overload relais would be good
<Degi>
Hm if we have power+data conenctors for the probes, we could?
<azonenberg>
electronic_eel: If it's an active probe with the i2c hookup, we can read a discovery eeprom
<azonenberg>
that's exactly what lecroy does with probus
<electronic_eel>
active probes yes, but passive ones
<azonenberg>
passive probes will just have a SMA hookup
<azonenberg>
The user will be expected to key in the probe serial number manually if they want to de-embed
<electronic_eel>
yes, that is why I was thinking nfc tags
<azonenberg>
i don't think that's unreasonable
<azonenberg>
nfc seems like a lot of complexity
<azonenberg>
especially keeping in mind that the SMA may be user supplied, not part of the probe
<Degi>
And I guess I can just stick in random SMA things without any configuration / press a "custom probe" button?
<Degi>
Yes
<azonenberg>
Yes. you'll have a selector in the software for specifying probe gain/attenuation
<azonenberg>
or you'll be able to load a s2p to de-embed
<electronic_eel>
custom stuff will always work, but I think to make using the default things as easy as possible
<Degi>
What does de-embed mean? (and s2p?)
<azonenberg>
again the problem with a passive probe is that the cable isnt likely to be part of the probe
<azonenberg>
so we cant really put any kind of discovery on it
<Degi>
Unless we solder the cable to the probe and take 200 € extra for a 50 cm longer cable lol
<electronic_eel>
but if the user keeps the cable with the probe he could use this and have working autodetect
<azonenberg>
Yes, if you have a custom cable that's attached to the probe somehow
<azonenberg>
but i'm trying to avoid going into this proprietary hell like other scope vendors
<electronic_eel>
I don't want to hinder users doing it differently, just offer a way to automate stuff
<azonenberg>
maybe having some kind of collar you can clip around the coax with a tag on it? but at that point you also risk reading the wrong channel's tag
<azonenberg>
since we'd have 8+ channels in close proximity
<Degi>
I mean you could have a calibration waveform generator and just stick your probe on that? I mean isn't the most interesting part about a probe the bandwidth and attenuation?
<electronic_eel>
the nfc tag would be something you can stick to the cable if you want to or not use it at all
<azonenberg>
i just dont see it being worth the effort
<azonenberg>
Degi: for full de-embedding you want 2-port s-parameters for the probe (basically attenuation and phase at regularly spaced frequencies)
<azonenberg>
so you can invert the transfer function of the probe
<Degi>
de-embedding means making it so like you're sampling the signal with an ideal ADC?
<electronic_eel>
and if you also keep the cable, you can do fancy stuff like phase matching different probes
<azonenberg>
Basically yes. removing effects of the probe, frontend, and even a test fixture the DUT is mounted to potentially
<Degi>
I mean you could stick the probe onto a calibration waveform generator on the fancier scope models
<azonenberg>
no this takes a full VNA
<Degi>
Hm can't you do VNA with a DAC and ADC?
<azonenberg>
No you need phase information too
<azonenberg>
plus you'd need to calibrate the dac and adc for frequency response, etc
<Degi>
I mean you could get that from that I think, at least if the ADC and DAC are phase locked to each other?
<azonenberg>
The high end price tier for the probes will include a factory VNA calibration and both hardcopy and machine readable s-parameter data
<Degi>
Hm yes, that too
<azonenberg>
as an optional value-add service
<azonenberg>
if you're not de-embedding, or plan to characterize them yourself and have the gear to do so, you can just by the basic edition and get a probe that's tested to "work" but not characterized beyond that
<azonenberg>
but i suspect a lot of labs would love to buy pre-characterized probes that come with cal data out of the box
<azonenberg>
90% of users will probably be fine just assuming it's an ideal, flat 10:1 probe
<Degi>
I mean for me as long as its roughly on the same level of a rigol or so of the same price, that's pretty good
<Degi>
But yes, for labs that'd be pretty neat
<azonenberg>
Exactly
<azonenberg>
For both probes and scopes my plan is to have four possible price/support tiers
<Degi>
You mean the two higher end scopes?
<azonenberg>
no i mean for all versions of the probe and scope
<electronic_eel>
if we find nice and easy to use connectors for the active probes, we could also offer them for the passive ones, just with the i2c eeprom for autodetection
<azonenberg>
option 1: Build it yourself off the published kicad designs: you're hardcore and dont mind sourcing parts yourself plus paying PCB NRE, will probably end up costing you more than buying an assembled unit, but you have the satisfaction of DIY
<azonenberg>
plus if i go out of business the project isn't dead
<Degi>
TBH I think making kits available would be nice too. Or bare PCB's, that's usually very inefficient to make one just for yourself.
<azonenberg>
option 2: student/hobbyist edition assembled unit. Soldered, tested, not characterized beyond that. Must sign a 1-year noncommercial use contract or provide proof of nonprofit/student status
<miek>
kitting is a huge pain
<azonenberg>
You get no support, but i sell you the hardware for basically what it cost me to make it plus shipping, little to no profit
<azonenberg>
i.e. you get to take advantage of my bulk discount for parts, but you're on your own after that
<Degi>
Hm what's so hard about kitting vs soldering it? That you need to label every component?
<azonenberg>
degi: i don't think you realize quite how many parts will be in these things :)
<azonenberg>
option 3: small business/consultant assembled unit. Assembled, tested, no restrictions on commercial use, I charge a bit more, and will provide basic assistance if you have problems
<azonenberg>
option 4: pro. Assembled, tested, and individual calibration certificate and s-parameters supplied for each subsystem/channel/probe
<Degi>
Hm with 100 different parts it could be quite a PITA, yeah. I still think that making the PCBs available for a small cost would be good (since they get reaaallly pricy if you only want one, and very cheap for more)
<azonenberg>
optional 1-year support contracts available for an extra charge
<azonenberg>
i suspect the bom will be >100 different parts but we'll see
<electronic_eel>
for the scope yes, but not for the probes
<azonenberg>
Oh, bare probe PCBs could totally be a thing
<electronic_eel>
if you already have a nice scope but just want a good probe
<azonenberg>
i could sell the enclosure, the bare PCB, the tip/ground connectors, and you'd supply your own resistors for whatever attenuation you want
<azonenberg>
we'd suggest a few combinations that make common gain values
<azonenberg>
and are tested to have good freq response
<azonenberg>
so probe kits could totally work
<azonenberg>
scope kits i don't see being viable
<electronic_eel>
ack
<azonenberg>
it's open source and if a third party is sufficiently motivated they can make it
<azonenberg>
But i don't think this will be popular. like, "build your own raspberry pi" just isn't a thing
<azonenberg>
the pool of people with the skill set to do it is smallish, and most of them would rather just buy an assembled unit
<Degi>
Hm would it be a bad thing if some company took the design and made it for cheap and sold it on ebay or so? (like nanoVNA etc., I'd kinda like it because that means that it'd probably get made for cheaper etc.)
<azonenberg>
If somebody forks the design and competes with me i'm fine with that. i'm not trying to make a business out of it per se, i'm trying to recoup my R&D expenses so i can move on to the next project
<electronic_eel>
see the icebreaker for example, there is now a seller on aliexpress
<azonenberg>
i consider my lab a toy that subsidizes more toys
<azonenberg>
the actual paid work is billed hourly time+materials to clients
<electronic_eel>
unfortunately said seller also copies the full logo and everything
<Degi>
I mean if the scope is popular enough, you'd probably make the money back from sale model 3 and 4
<azonenberg>
but the open hardware stuff, basically, i just don't want to take a net loss on
<azonenberg>
and yes, that's the idea. I want to offer the student version as cheaply as i can without LOSING money
<azonenberg>
and then milk companies with budget for as much as they'll let me :p
<azonenberg>
It's a very communist pricing model :p
<Degi>
Heh yeah I find that good
<Degi>
Especially since you could get bulk discounts
<azonenberg>
and especially with the high end stuff sold significantly above cost, the value-add is the calibration and support
<azonenberg>
if some aliexpress seller wants to sell nist-traceable s-parameters with each clone of my probe they make, that service is actually worth money
<Degi>
I mean yeah, don't think that somebody is gonna copy that
<azonenberg>
and if they just sell boards with no support, i still have an edge over them for high end
<azonenberg>
even if we're making the exact same open design
<electronic_eel>
I think the buyer should have an easy way to find out who actually made the stuff he bought, so some kind of trace tag or similar
<electronic_eel>
the plan for glasgow was to add an digital signature ic for this
<Degi>
Maybe when you make the PCB's, add a tag onto the open source design
<Degi>
Lol
<azonenberg>
electronic_eel: All of my units will have a Molex track-it traceability tag on them
<azonenberg>
it's a little tinned copper pad with a laser engraved 2d barcode serial number on it
<azonenberg>
as well as an external printed label
<electronic_eel>
can't you just fake that?
<Degi>
And you just solder that on?
<azonenberg>
you'll be able to pop it under a microscope and scan the barcode, or type in the serial on the label, and get s-parameters for your unit etc
<Degi>
It's 1.8x2.8 mm, doubt you need a microscope...
<azonenberg>
that's 1.8 x 2.8 mm with a fairly large 2D barcode laser etched in it
<azonenberg>
rough counting of the digikey photo suggests something like 11x30 dots
<azonenberg>
you want magnification, trust me :p
<electronic_eel>
so to fake that you just take one valid serial number / barcode and a cheap engraving laser
<azonenberg>
This is not intended for anti-counterfeiting
<azonenberg>
it's for traceability if you have a unit and need to look up the cal
<azonenberg>
generally the user will know if they bought it from a sketchy reseller or an authorized dealer
<electronic_eel>
why not just stick a regular printed 2d barcode on?
<azonenberg>
Because these things are premade and can be just soldered onto the board
<Degi>
Hm my phone camera is more than enough for that and my eyes is probably workable too to be honest (my phone can zoom to approx. 3 µm per pixel at the nearest distance)
<azonenberg>
super low effort
<azonenberg>
there will also be a printed serial somewhere on the enclosure, likely the same S/N as the barcode on the motherboard
<electronic_eel>
not so sure about the effort: you need to match the soldered on and printed serials
<azonenberg>
well my thought was during the test/systems integration phase
<electronic_eel>
if you just assign the serials by sticking preprinted labels on it is easier
<azonenberg>
i'd have to scan the barcode on the board in order to upload the s2p from that AFE to my database, etc
<Degi>
Something like a label printer with USB connector could print barcodes too
<azonenberg>
Cal data all has to be easily downloadable in order to let end users de-embed stuff
<Degi>
Or an EEPROM with the s2p data
<Degi>
Not sure if keeping the cal data separate from the scope is that good
<electronic_eel>
at work we do it like this: 1. the hardware is made without serials
<electronic_eel>
2. we stick preprinted labels on the units
<electronic_eel>
3. we put them on the test desk
<electronic_eel>
4. we scan the label with a barcode scanner
<electronic_eel>
5. the automated test runs and sends the serial from the barcode into the device
<electronic_eel>
6. all test results are stored into the database
<Degi>
Why not embed the data on onboard flash/EEPROM/OTP EPROM?
<azonenberg>
Degi: cal can drift over time. having the data external makes it easier to update
<Degi>
Hm to calibrate it, don't you have to take the entire unit to the calibration place anyhow?
<azonenberg>
since the de-embedding will be happening externally anyway
<azonenberg>
Depends on how you do things. If you have enough gear on-site cal is an option
<Degi>
Hmm that's true... There's a diskette for a MS DOS program for the TDS520 which can do that
<azonenberg>
the company i was just talking to has a $2000 minimum, but they'll send a tech out to my lab and cal all of my scopes, vnas, probes, psus, meters, etc in one go while i wait
<azonenberg>
Or i can ship hardware to their lab and pay per unit instead
<Degi>
Is it a company external to the manufacturers of the devices?
<miek>
interesting, i would've thought the minimum was much higher
<azonenberg>
Yes this is a third party cal lab
<azonenberg>
miek: that's plus travel charges i think, but they have customers in my area and i could probably arrange for them to do me the same day as one of them
<miek>
those molex tags seem to be going obsolete by the way
<azonenberg>
oh?
<azonenberg>
electronic_eel: re serialization the other option is to use the fpga boot flash serial number as a machine readable serial and just print that out on the side of the unit at the test station
<azonenberg>
anyway this year i have just my dmms and psus up for cal since one scope is new and the other was just serviced
<azonenberg>
so once the apocalypse is over i'll be shipping them out
<azonenberg>
next year i'll have the scopes up for cal, plus probably a decent VNA
<azonenberg>
Interesting
<azonenberg>
i wonder why that is
<Degi>
Hm some kinda way to store the cal onboard (or in the computer connected to it?) would be neat if your lab doesnt have internet
<azonenberg>
the cal wouldnt be downloaded each time
<azonenberg>
the point is that i want to allow you to download the s2p given only the hardware
<azonenberg>
it's assumed you'd save it locally and use the local copy after that
<azonenberg>
glscopeclient will look for the files locally
<Degi>
Hm okay
<azonenberg>
keep in mind characterization reports for a lot of commercial gear are just a printout of the report
<azonenberg>
no machine readable data at all
<Degi>
(I think at uni a bunch of lab computers don't have internet. Not allowed to hook up Win7 and older to the internet)
<Degi>
Oof
<Degi>
I think the TDS520 stores cal data internally on a battery backed SRAM
<electronic_eel>
it is just a s2p text file, you can easily download it with a browser, put it onto an usb stick and bring it to the lab pc
<electronic_eel>
no fancy "activation" thing, no spyware
<Degi>
Yes that sounds okay I think
<electronic_eel>
no license manager!
<azonenberg>
well it will probably be a bunch because i'd want a s2p per channel
<azonenberg>
for the afe etc
<azonenberg>
then one for each probe in the batch, and so on
<azonenberg>
Still some work to do to figure out logistics on that side
<Degi>
Huh I just designed a time to digital converter in spice and it seems to be able to make 500 µV output difference for a 1 ps trigger time difference. Max range is I think 300 mV or so
<electronic_eel>
Degi: how large is the cap you were using?
<Degi>
10 pF
<Degi>
Charged over 1 kOhm and a diode
<Degi>
The diode makes some nonlinearity though... Gotta find a better solution
<electronic_eel>
did you factor in the charge the adc or opamp is going to take off that?
<Degi>
Hmmm good point
<Degi>
Oh wait 1 pA on 10 pF is 0.1 V/s
<Degi>
On the other hand it could be bigger. 1 ps is really small for a 100 MHz scope
<electronic_eel>
1 pA? that is easily flowing over the soldermask or over the air
<Degi>
Hm we could put a guard ring around it. And it doesn't really flow through air at these voltages unless you have radioactive probes on the go
<Degi>
And keep out solder mask nearby
<azonenberg>
The TDC for a fast scope is going to be an *extremely* nontrivial design
<lain>
also beware dielectric absorption.
<azonenberg>
we're talking multiple iterations of spice sim and field solver models of the actual gerbers
<Degi>
Dielectric absorption? THat thing where a cap fills back up after discharging it?
<lain>
which is the effect that a real capacitor will not discharge completely, even when discharged through a short
<electronic_eel>
taking stuff down into the pA region is a science of it's own
<lain>
yeah
<azonenberg>
and probably a couple of board spins
<Degi>
The last time I did something with picoamps I used PTFE for insulation and didn't let the picoamp side touch anything but PTFE at all (and handled everything with gloves), it seemed to work
<azonenberg>
It's not going to happen on the first round scope for certain
<Degi>
I mean a nA of leakage would make 100 V/s on the cap. If we sample within 10 µs that is 1 mV of error
<Degi>
I could use a nF too
<Degi>
On the other hand, my picoamp meter was running in a high (10^-4 to 10^-6 mBar) vacuum, maybe that helped a little with reducing leakage. (Did you know you can put entire cheap PCB's without removing flux into a vacuum chamber and still get 10^-4 mBar?)
<electronic_eel>
nF and nF sounds much more reasonable to me. you just have to stick to NP0
<electronic_eel>
nF and nA
<Degi>
Yes a nA is a bunch of current. Just don't smear all over it with sweaty fingers...
<azonenberg>
meanwhile i'm reminded of the crystal oscillator i was debugging for awygle the other day
<azonenberg>
that would sometimes work and sometimes not after i reworked some capacitors to swap them with different values
<azonenberg>
turned out it needed a 1M feedback resistor
<azonenberg>
and it would work/not work depending on how well i had cleaned off flux residue after soldering
<Degi>
Lol
<awygle>
load-bearing flux ftw
<Degi>
Hahaha
<Degi>
Hm on the other hand I didn't measure how much current my detector actually detects (how do you make precisely 1 pA (with like 50% error) anyways?), considering that I used a 100 GOhm resistor from the soviet union for 2.5 € (IIRC I cleaned it though)
<electronic_eel>
was it just a quartz with the osc inside the ic or a a full oscillator?
<azonenberg>
quarts + osc in ic
<azonenberg>
ic needed an external 1M resistor across the crystal and that was forgotten in the design
<electronic_eel>
was the 1M in the datasheet?
<azonenberg>
Yes, but he forgot to put it on the board :p
<electronic_eel>
ah, ok
<electronic_eel>
unusual requirement, but ok
<azonenberg>
we thought it was bad load caps so i was binary searching my way through the range of possible cap values to see what worked and what didnt
<azonenberg>
it was acting random and had no observable correlation
<azonenberg>
sometimes worked, sometimes very slow startup
<azonenberg>
touching it with a probe would make it work or not work, etc
<electronic_eel>
yeah, the usual fun
<electronic_eel>
after some time I tend to resign to reread the datasheet
<Degi>
Wtf there are NP0 caps 7.62x10.16x2.5 mm rated for pulse discharge on mouser lol
<Degi>
SMD caps with builtin discharge resistor
<azonenberg>
meanwhile i was just showing somebody the other day a pic of a LARGE pulse discharge cap
<azonenberg>
350 uF 4 kV, not sure about the internal construction bt it came from some kind of pulsed laser system
<azonenberg>
had giant ceramic ringed insulators on top
<Degi>
Hm I have smaller versions of that... 2 kV 30 µF. And you can overcharge them and they're still usable after that, they make a weird noise heh
<azonenberg>
these were the size of... probably about a large pillow or small couch cushion?
<azonenberg>
the bank of six was the size of a small refrigerator
<Degi>
Mine are like a 300 ml metal can
<Degi>
Hm I have some 18 mF 400 V DC link caps too... Found them for cheap on ebay. At 100 V 4 of them make quite an impression
<azonenberg>
we discharged them into random fruit, burgers, etc as part of the electronics club end-of-semester party, shortly before EH&S came in and shut the club down :p
<Degi>
EH&S?
<azonenberg>
environmental health and safety
<azonenberg>
the running joke was, the club's lab didn't have safety hazards
<Degi>
Wait a bank of six, that's almost 17 kJ lol
<azonenberg>
it WAS a safety hazard :p
<azonenberg>
More than that
<azonenberg>
25 kJ fully charged if memory serves me right? i might be misremembering the capacitance
<Degi>
Hm did you overcharge them?
<azonenberg>
i know it was 4 kV each and we charged the bank to somewhere around 20 kV
<azonenberg>
cant recall the capacitance
<Degi>
Maybe 500 µF. That is 24 kJ
<Degi>
Alibaba has some really weird caps. 100 kV and 10 nH of inductance or so
<azonenberg>
Based on high speed photography we dumped the whole bank in <1 ms
<azonenberg>
we were filming several hundred FPS with a rolling shutter and the flash didn't even light up the entire frame
<Degi>
"How to get that one wireless agency really pissed at you"
<Degi>
Yes that's usual. A ms is pretty logn
<Degi>
*long
<azonenberg>
So peak pulse power was in the tens of megawatts easily
<Degi>
Yes haha
<Degi>
My 400 V electrolytic cap bank should be able to hit 4 MW or so into a matched load if the caps are good, that's not so much
<Degi>
At those voltages you can get lots of power
<azonenberg>
I remember one time we blew up a slice of cake with it after squeezing orange juice over the top to make it more conductive
<awygle>
it was in the applications section of the datasheet but nowhere else
<azonenberg>
One spectator reported being hit by a chunk of charred frosting. We never found the rest of it
<Degi>
Hm wish I had a fast enough scope to measure HV pulses, but once I made a somewhat transmission line based HV pulse generator (with a very mismatched spark gap) that should be able to output ~100 MW
<azonenberg>
it was completely obliterated
<azonenberg>
on the high speed there was cake, next frame blinding white light, next frame just a chocolate colored blur
<Degi>
Hm did you know that you can fill liquid nitrogen into a thick plastic bottle (like those bottles you buy water in at the supermarket) and hold it with your jackets arm lol
<Degi>
Huh I can't find PTFE caps on mouser. weird
<Degi>
Oh wow this time to digital with 330 pF performs somewhat linearly with 737 µV/ps at 1 ps difference from an arbitraty 0 position and 763 µV/ps at 3000 ps difference
<Degi>
I might just construct this circuit and test it later, the FPGA has builtin delays which have 40 ps tuneability
<azonenberg>
Fine for a prototype, but not nearly the precision we need
<azonenberg>
for the 2 GHz scope, we're talking about needing 5ps resolution on delays from 0 to 100ps
<azonenberg>
basically start on posedge trigger clock and stop on trigger
<Degi>
I mean I only have a 40 ps delay, do you have any idea how to construct a delay that can be adjusted in 5 ps steps?
<Degi>
I mean this circuit seems to work fine with 1 ps
<azonenberg>
you mean for testing the TDC?
<Degi>
But for testing I only have something with 40 ps yes
<azonenberg>
just get two slightly unequal oscillators and measure the beats
<azonenberg>
start on one and stop on the other
<azonenberg>
you should be able to get arbitrarily long/short delays
<azonenberg>
of course you won't know what the delay is, but knowing the frequencies you can bound the range and expected distribution etc
<Degi>
Hm I think I'll start with the 40 ps delays at first though
<Degi>
(This design doesn't work beyond really max. 1 GHz with the FPGA I'd like to use, so I think its more suited for a 100 / 400 MHz scope)
<Degi>
Like the whole thing is maybe 10 €
<azonenberg>
well we have a few more board iterations before the full scope happens
<azonenberg>
so if you want to prototype it, we can see how it performs and decide if we want to include it
<Degi>
Hm yeah
<Degi>
Hmm it'd need three FPGA pins though, one for comparator input and two for the capacitor charging. We could put a FPGA onboard or maybe use logic gates. The FPGA costs like 7 € for the max speed grade.
<azonenberg>
I think you are going to need to do as much as possible in discrete logic. FPGA has too much jitter etc for this level of timing precision
<Degi>
SR latches seem to be a bit hard to find for some reason
<Degi>
Its all D latches for some reason, at least the cheaper ones
<Degi>
I think for the lower end 100/400 MHz scope, the jitter of an FPGA could be okay ish with an external clock gen
<azonenberg>
You can use D latches with external S/R
<azonenberg>
like i did for the overvoltage lockout
<Degi>
Hm did you upload a newer version of the AFE characterization pdf?
<azonenberg>
i've been pushing new versions every few hours as i work on it
<azonenberg>
just uploaded latest
<Degi>
I see, just needed to press F5. Thanks
<Degi>
MC10EP29 seems to be around 400 ps propagation delay. Neat
<electronic_eel>
but how do they intend it to be used?
<electronic_eel>
is it flexible while the ground blade they also have isn't?
<electronic_eel>
or is it especially designed to connect to the adhesive copper strips they also sell?
<electronic_eel>
or does the groove it has go onto some common pin type?
<Degi>
Maybe the inductance between the tip and the end is lower compared to a round wire
<electronic_eel>
hmm, maybe. but why don't they tell how to use their stuff or what the advantages are?
<electronic_eel>
it isn't in their manual of their probes either, they just have a graph showing the accessories and their order number there
maartenBE has quit [Ping timeout: 256 seconds]
<electronic_eel>
not how they intend them to be used
<miek>
the blade is meant to be used with the adhesive copper (stuck on top of an IC)
<miek>
i think the groove is meant to be a socket for different tips
maartenBE has joined #scopehal
<electronic_eel>
ok, but then why the big leaf thing and not some flexible, insulated wire? they also have those
<Degi>
Something like desoldering wick could work too
<electronic_eel>
Degi: desoldering wick works well as ground point. I use my old and bad desoldering wick sometimes like that
<miek>
the big wide leaf thing has lower inductance than a wire
<electronic_eel>
how does it lower the inductance? isn't the inductance at rf freqs mostly defined by the surface area? can't you get the inductance in a wire down by using very fine braid?
<azonenberg>
electronic_eel: litz wire isnt good past a few MHz because you cant insulate the individual strands effectively and still terminate it at the end
<electronic_eel>
ah, makes sense
<azonenberg>
the leaf is lower inductance than the blade. I haven't used the leaf much but the blade has worked well for me. Both are flexible
<azonenberg>
the z-ground is the one i actually use most, with the flexible ground as a close second
<electronic_eel>
"flexible ground" being the blade?
<azonenberg>
No its a short wire
<azonenberg>
like 5cm or so, with a machine pin at one end and a socket for pin headers at the other
<electronic_eel>
so you connect it to a pin header on the pcb, right?
<azonenberg>
yeah
<azonenberg>
it's the replacement for the big alligator clip grounds on old school passive probes
<electronic_eel>
I hate those
<azonenberg>
for browsing around when you dont need super high performance and there isnt necessarily a ground conveniently close by
<azonenberg>
then i use the z-ground often because you can adjust spacing of tip to ground and just hit a nearby top mounted decoupling cap or similar
<electronic_eel>
can you use the z-ground together with the positioner or is it not stable enough for that?
<azonenberg>
i do so routinely
<miek>
i really need to get/make all these for my active probe, it didn't come with any of the accessories
<azonenberg>
however, it's a little bit stiff and can be tricky to align
<electronic_eel>
damn, I whish I didn't cancel going to the embedded world fair this year, could have tested all these there
<azonenberg>
maybe next year when the apocalypse is over :p
<miek>
i don't remember seeing any probing stuff on show at electronica, maybe it's different at embedded world?
<electronic_eel>
they don't have a picture yet
<azonenberg>
I'll send you a pic of my ZS1500 accessories when i get out to the lab in a bit (at my desk having a late lunch, then need to finish a report for a client)
<electronic_eel>
miek: they had a embedded world logo on their website, so I guess they were there
<azonenberg>
the ZS1500 is a 1.5 GHz tetris OEM'd for lecroy with a slightly longer connection box that has the probus connector and discovery eeprom in it
<electronic_eel>
all the other scope vendors usually are there too (if there are no viruses around)
<electronic_eel>
do you have some kind of grabber for the ZS1500?
<azonenberg>
i have two little grabbers that i havent used much, i'll show you the pic