<azonenberg>
best simulation to date of the ordered probe pcb, compared to vna measurements of the real thing
<azonenberg>
sim has all resistor footpritns shorted, vna data is with resistors so it's scaled
<azonenberg>
the shape of the curves seems similar so i think it's a start
<azonenberg>
it seems like reflection off the sma footprint is the main source of this loss
<azonenberg>
i think the shift is caused by the resistors changing the length of the resonant stub, but could be wrong
<Degi>
Hm red is sim?
<lain>
huh.
Degi has quit [Ping timeout: 272 seconds]
Degi has joined #scopehal
_whitelogger has joined #scopehal
futarisIRCcloud has joined #scopehal
<azonenberg>
yeah
<azonenberg>
Just ordered a N torque wrench. Had some problems with them acting funny when not tight enough
<azonenberg>
93R resistors should be here thurs or fri
<azonenberg>
and the minicircuits splitter is coming tomorrow
<azonenberg>
So should be good to proceed with more testing of the AFE soon
<azonenberg>
Still have to do lots of TCP code, but i don't think that is the highest priority right now
<azonenberg>
I think a reasonable next step is to build test boards for the active probe subsystem
<azonenberg>
As we can do that without any dependencies on the passive probes, the AFE, or the FGPA subsystem
<azonenberg>
but we need it to be hardware proven before we can spin the final input stage board
bluezinc has quit [Ping timeout: 246 seconds]
bluezinc has joined #scopehal
yourfate has joined #scopehal
<azonenberg>
Degi, electronic_eel, lain, monochroma: so i've been thinking more about the active probes
<azonenberg>
i'm thinking of using a STM32G071 which is available in a 4x4mm 28QFN package
<azonenberg>
it has a usb-pd PHY so we could implement an actually standard alt mode
<azonenberg>
the thing is, it does *not* have a usb 2.0 data phy
<azonenberg>
so my thought is we could perhaps use the sideband channels for uart or i2c
<azonenberg>
and not use usb at all
<electronic_eel>
why not a STM32F042 in QFN 28? it has a usb fs phy
<azonenberg>
i dont think it has a PD phy
<electronic_eel>
no, it does not have pd
<electronic_eel>
but implementing proper pd is a lot of work
<azonenberg>
and implementing usb is not?
<electronic_eel>
no, there are lots of libs and examples you can use
<electronic_eel>
also regular usb is common for years, so lot's of experience out in the world
<electronic_eel>
pd is new and known to be complicated
<azonenberg>
Hmmm
<azonenberg>
how much of the pd protocol would we actually have to implement though?
<azonenberg>
especially if only between our endpoints>?
<electronic_eel>
also I think it is an advantage if you can contact the probe mcu without a special scope, just by plugging it in a regular pc
<azonenberg>
Hmmmm
<azonenberg>
what would the use case for that be?
<azonenberg>
we cant power up the probe without a custom host
<electronic_eel>
true, but you can read out stuff from the flash (like cal data) or update the firmware
<azonenberg>
seems like a bit of a niche requirement
<azonenberg>
we can just have the fpga do that
<azonenberg>
or the stm32f7
<electronic_eel>
but we wouldn't have that when connecting to a scope of another brand. putting this into the external power interface makes this more complicated
<azonenberg>
i guess the other option is to go full standard usb 2.0 over a C connector
<azonenberg>
and put a dc-dc on the probe to generate all our rails from vbus
<azonenberg>
at that point using these probes with a third party scope is plausible
<electronic_eel>
why not use the usb 3 ss data lines for non-standard powering, once the connection to the mcu is established and it is ensured that it is a direct connection without hub in between?
<azonenberg>
what i mean is, if you want compatibility with third party scopes
<azonenberg>
we have to be powered entirely by vbus
<azonenberg>
If we are only using our own scopes thats not a problem
<electronic_eel>
yes, but that could be done with a external power adapter
<azonenberg>
hmm
<electronic_eel>
so we just need the dc/dc for other scopes
<azonenberg>
True
<electronic_eel>
the external power adapter would need a small mcu which is otg capable
<electronic_eel>
but that is very common
<azonenberg>
so i guess we're back to the original goal of an otg capable mcu on the host and a tiny usb device-only mcu on the probe
<electronic_eel>
I think this checks all the requirements without adding lot's of complexity needed for implementing both sides of pd
<azonenberg>
so suggest parts for host and device?
<azonenberg>
QFN stm32, as small as possible while having the resources we need
<electronic_eel>
the STM32F042 in qfn28 as device in the probe
<electronic_eel>
the STM32F401 in qfn48 as host in the scope and power adapter for other scope brands
<azonenberg>
ok i'll start poking at that and see how far i get
<azonenberg>
any suggestions on a preferred usb connector?
<electronic_eel>
I just have experience with usb-c connectors that don't have the ss data lines populated
<electronic_eel>
I have tried to simulate the afe of the rigol la
<electronic_eel>
either my sim is wrong or the performance of their circuit
<electronic_eel>
it looks more like a notch filter at about 50MHz than an afe
<electronic_eel>
I'll have lunch now and can share the data later
<azonenberg>
I want to build something competitive with this
<azonenberg>
probably a standalone 1U instrument with SMA inputs on the front, as many as i can fit
<electronic_eel>
the solder-in probes, do you want to change them to sma or stick with mmcx?
<electronic_eel>
sma would have the advantage that it is very common and you won't need to add adapter wires
<azonenberg>
Unsure at this point
<azonenberg>
yes
<azonenberg>
I am thinking of switching
<electronic_eel>
if symbiotic eda wants cheap giveaways, they would probably prefer not to have expensive cables
<azonenberg>
Yeah
<azonenberg>
i guess the question then is, do i want to use the cheap samtec sma i've been using?
<azonenberg>
or the higher end amphenol
<azonenberg>
Or even make variants with both?
<electronic_eel>
didn't bvernoux suggest a cheap + good alternative?
<azonenberg>
the one he suggested costs more than the amphenol i picked
<azonenberg>
he just found a good price on them from some sketchy source
<electronic_eel>
how sketchy? is anything below the likes of digikey sketchy?
<azonenberg>
i didnt look into it in detail
<azonenberg>
anyway, another thought i had was to increase the length of the flex probes i have now
<electronic_eel>
how long are they now?
<azonenberg>
about 75 mm ish iirc?
<azonenberg>
also we won't be able to use unmasked CPWs because of the risk of shorts
<azonenberg>
so i'll have to keep coverlay over it. Which might increase loss
<azonenberg>
But it might also be more predictable impedance properties than soldermask?
<electronic_eel>
75mm seems not too bad, but if it will be sma I think it should become a bit longer
<electronic_eel>
yeah, shorts are an issue. it should be isolated except the tip section and the actual sma connector on the back
<azonenberg>
Yeah. in fact i shorted a board with the mmcx ground today on my current solder-in probe lol
<azonenberg>
i would like a flex tail long enough to not hang over the DUT
<azonenberg>
lecroy's QuickLink probe tips have 9 inch leads built in
<azonenberg>
which then plug into an active amplifier pod
<electronic_eel>
ouch. would it be an option to add a small 3d printed housing around the sma connector, something you clip on or similar?
<azonenberg>
at that point i'd just use kapton tape or something
<azonenberg>
but that wont help with the mating half
<azonenberg>
no, i'd like to see it just get a bit longer
<azonenberg>
maybe even make several size variants eventually
<electronic_eel>
yeah, that is why I thought to clip it on, it would cover the connector and the mating part on the cable
<electronic_eel>
the mmcx connector was fully smd, not edge mount, correct?
<azonenberg>
i have two different mmcx's
<azonenberg>
the current iteration of the solder in probe uses a right angle smd
<azonenberg>
they're not that great past a hundred MHz or two
<azonenberg>
the single cheap resistor is a contributing factor too
<azonenberg>
the edge launch mmcx i have sits in a cutout on the board and is a pain to do footprints for
<azonenberg>
so i'm thinking of switching to edge launch SMA for the next board
<electronic_eel>
ok, so for edge mount you'd have to think about isolating the part below the probe pcb too
<azonenberg>
Yeah
<electronic_eel>
so that will be a lot of work, your idea of making it longer is probably easier
<azonenberg>
Yeah. Think i should switch to GCPW or stick with microstrip?
<azonenberg>
or even move to a pure CPW construction without a ground plane
<electronic_eel>
a edge mount connector means the probe will need to be flex in the front section and rigid at the connector
<azonenberg>
That's what all of my probes were. Polyimide PCB with a non-patterned fr4 stiffener at the connector area
<azonenberg>
This wouldn't change anything
<azonenberg>
there'd be no vias or copper on the fr4, just a little square glued on
<electronic_eel>
how well impedance controlled is the flex stuff?
<azonenberg>
Good question
<azonenberg>
i would expect +/- 10% or better
<electronic_eel>
hmm, that wouldn't be that bad
<electronic_eel>
so how would be the stackup, with a isolating polyimide layer top and bottom?
<azonenberg>
Yeah thick polyimide substrate then coverlay on the top over the copper
<azonenberg>
Also, thoughts on naming the high speed SERDES-based LA pod MEAD?
<azonenberg>
as in Conway's coauthor
<electronic_eel>
for me that is a project far in the future, so maybe naming is a bit too early, but that is your call
<azonenberg>
i want to start making notes on the roadmap etc
<azonenberg>
one big question is, do we have one serdes LA or two?
<azonenberg>
the xc7a200t in ffg1156 has 16 GTPs, if we use one lane of each for sync we can do 12 channels @ 6 Gsps for quite cheap
<electronic_eel>
12 ch, is that enough?
<electronic_eel>
if you want to look at ddr memory, I think you'd need a bit more
<azonenberg>
the HDA125 has 18. but that would need a much higher end fpga
<azonenberg>
i'm proposing an intermediate option
<azonenberg>
It's a $293 FPGA, then we'd need a bunch of DDR3 for buffering waveforms and the pcb itself
<electronic_eel>
maybe two fpgas in one box for 24 ch?
<azonenberg>
That's actually a decent option. But in that case i'd want to steal one gtp quad and use it for 10GbE
<azonenberg>
giving us 21
<azonenberg>
The other issue is, though... the physical logistics of 24 probes
<electronic_eel>
oh, right, 24 smas
<azonenberg>
yes
<azonenberg>
That is a nontrivial front panel design if you want to keep it 1U
<azonenberg>
and if you make it bigger now you need pigtails inside the box for the second row, you have to skew match them, etc
<electronic_eel>
1u won't work I'd say
<azonenberg>
12ch should
<electronic_eel>
yeah, so 12ch and some way to properly sync them
<azonenberg>
Yeah that makes more sense imo
<electronic_eel>
so you'd need two of these 1u boxes to get 24ch
<azonenberg>
Yep
<electronic_eel>
but does 12ch inputs allow for a 10gbe connection?
<electronic_eel>
with just 1gbe this doesn't look very practical
<azonenberg>
No. the FPGA has 16 GTPs in 4 quads, and we need one lane from each quad to sync
<azonenberg>
however we have lots of LVDS IOs
<azonenberg>
so we could slap a smaller kintex7 on and just use it as a 10g bridge?
<azonenberg>
i'm not sure yet. I wanted to name the project so we'd be able to start writing up notes and discussing it
<electronic_eel>
I think we should first get blondel, active and passive probe, conway and scopehal into shape, then we can think of future roadmaps
<azonenberg>
fair point
<azonenberg>
So... blondel's next step is continuing tweaks of the AFE with the 93 ohm resistors arriving this afternoon
<electronic_eel>
by the time we begin working on the next-gen la, maybe there are new fpga models out that allow better options
<azonenberg>
active probe's next step is to design the usb altfunction testbed
<azonenberg>
which is a TODO for me in the near future unless somebody else has time to work on it
<azonenberg>
Passive probe's next step is to wait for the prototypes to arrive
<electronic_eel>
why do you start the active probe with usb and psu, not with the actual probing part?
<azonenberg>
Because we need the host side interface designed to do the AFE board for blondel
<azonenberg>
the probe side stuff can be figured out later but we need to have all of the alt function power stuff pinned down
<azonenberg>
we know the RF interface is SMA
<azonenberg>
And we can fine tune the probe design in parallel with spinning the 3 other boards we're doing for blondel
<electronic_eel>
yeah, but we don't know what the optimum voltages are for the probe
<azonenberg>
Fair point
<azonenberg>
although if we pick stuff right we should be able to just tweak a resistor on the host to change it up to +/- 12V
<electronic_eel>
true
<azonenberg>
So, the initial active probe. What specs are we targeting for it?
<azonenberg>
did you have an amplifier chipset in mind?
<azonenberg>
what kind of gain/offset range?
<azonenberg>
(i assume we're going to need an offset stage in the probe buffer)
<electronic_eel>
if we target around 500MHz, we'll have no gap when combined with the passive probe
<azonenberg>
That was my thought too. Make it usable out to ZENNECK as a primary probe
<electronic_eel>
the big issue for me is be max usable voltage
<azonenberg>
We will need attenuation on the input, i think that much is certain
<electronic_eel>
lot's of active probes just allow +-5v, but that is too low for a primary probe
<electronic_eel>
if we can get a 10:1 divider working on the input, without adding too much capacitance, that would be nice
<electronic_eel>
would allow +-50v, and be usable for most stuff that is not mains voltage
<electronic_eel>
maybe we can push it to 60v, then 48v systems would be covered too
<azonenberg>
So one thing i noticed about the passive probe a while back btw
<azonenberg>
We're thermally limited to about 11V with the current arrangement of resistors
<azonenberg>
the 100R starts to dissipate 1/20W at about 22 mA
<azonenberg>
over 500 ohms that's 11V
<azonenberg>
we can go higher in pulses i imagine but that will be the datasheet max Vrms
<electronic_eel>
for the passive probe I think this isn't really an issue, as you probably won't have high voltages at high frequency
<electronic_eel>
but for the active probe with slower stuff I think voltage tolerance is an important issue
<azonenberg>
Yeah.
<azonenberg>
So what kind of input attenuator were you thinking?
<azonenberg>
maybe 90K : 10K or so?
<azonenberg>
we need to find an amplifier without a built in differential terminator on the input
<azonenberg>
actually i think we need more like 20:1 on the input as most of the nice looking amplifiers are +/- 2.5V not higher voltages
<azonenberg>
THS4520?
<azonenberg>
So 95K : 5K or thereabouts
<electronic_eel>
hmm, the THS4520 with -+2.5v supply has common mode input voltage from +1.75v to -1.3v
<azonenberg>
yeah i just found that spec and didnt like it :p
<electronic_eel>
if I read that correctly
<azonenberg>
hmmm
<electronic_eel>
the THS4520 has a differential input. do you plan to build a fully differential probe?
<azonenberg>
THS4303
<azonenberg>
I would like to at some point. initial version will be single ended
<azonenberg>
anyway yeah the 4303 looks nice so far
<electronic_eel>
the THS4303 has common mode of +-1.5v when powered at +-2.5v
<azonenberg>
So we need more attenuation
<azonenberg>
it's a fixed 10:1 gain anyway
<azonenberg>
Do you think 100:1 input attenuation + 10:1 gain would give us too much noise?
<azonenberg>
that would make it a 10:1 probe net
<azonenberg>
Say 99K : 1K or thereabouts
<azonenberg>
The example circuits all show 50 ohm inputs and i'm trying to see what kind of bias current it needs etc
<electronic_eel>
I'm not sure about the noise, don't have experience with it in such an application
<electronic_eel>
but having to add a shielding cage makes it more complicated
<azonenberg>
13 uA max over the commercial temp range
<azonenberg>
Which seems like too much with a 100K input impedance?
<azonenberg>
Hmmmm this is more like it
<azonenberg>
ADA4817-1ACPZ-R7
<azonenberg>
1 GHz -3dB b/w in unity gain configuration, 2 pA bias current
<azonenberg>
1.3 pF input cap
<azonenberg>
max bias current across temp range 135 pA
<azonenberg>
max 20 pA at 25C
<azonenberg>
input is 1.3 pF || 500G ohms
<azonenberg>
And can run on a +/- 5V supply
<electronic_eel>
input common mode at +-5v is -5v to +2.2v
<azonenberg>
Yes but we're attenuating the input anyway right?
<electronic_eel>
yes. if we build a 10:1, it would be 22v max in
<azonenberg>
That sounds reasonable right?
<azonenberg>
for a general purpose probe?
<azonenberg>
then we can do a higher attenuation lower bandwidth version for special purpose stuff
<electronic_eel>
for a active probe it is very good, but for a general purpose probe just meh
<electronic_eel>
you can't measure dips and stuff on a very common 24v supply
<azonenberg>
Fine, make it 20:1?
<azonenberg>
then it'll be 44V max
<electronic_eel>
that would make it much more usable for measuring on psu rails
<electronic_eel>
question is how it fares with low voltage stuff
<azonenberg>
Yeah
<azonenberg>
Well, i guess the question is 20:1 with what impedance?
<electronic_eel>
maybe 20:1 with a switchable integrated amp?
<azonenberg>
Let's think about doing this as 20:1 first
<azonenberg>
what input impedances would you want to target?
<electronic_eel>
1meg would be what people are used to from other scopes
<azonenberg>
at 100k ohms, 10 mV gives us 10 nA current flow, we need only 100 pA to bias the amplifier so we're two OOMs beyond there
<azonenberg>
sorry at 1M ohms
<azonenberg>
Is the 1.3 pF of input C a problem though?
<azonenberg>
my naive calculation modeling this as an RC LPF gives 122 kHz cutoff frequency
<azonenberg>
but i'm not sure if that's accurate
<electronic_eel>
I think this should be modeled a bit like what I'v been doing for the la afe, with a rc divider
<azonenberg>
So you want R||C on the input instead of just R
<azonenberg>
sized to match the input C of the amp?
<electronic_eel>
that would go into the ff range for this amp, I think we'd need to add some more capacitance to make it work
<azonenberg>
how low fF? you can get 100 fF discrete caps. Or we could just rely on the ~30 fF parasitic C of the 1M resistor lol
<electronic_eel>
yeah, that is the problem - if you rely on parasitics, the data changes dramatically changes when you wave your hand near it
<electronic_eel>
so we need to increase capacitance to make it more robust against such things
<electronic_eel>
we could start modeling for example with 8p || 1p (opamp) and 1p to give a 10:1
<electronic_eel>
ah, no, we want a 20:1
<electronic_eel>
18p || 1p (opamp) and 1p then.
<electronic_eel>
but I think you'll have to start simulating this to see if it works out
<electronic_eel>
but I wouldn't go far below 1p, because stray capacitances and parasitics then begin to play a too large role
<azonenberg>
So you want 1p in addition to the 1.3p parasitic on the amp
<azonenberg>
?
<electronic_eel>
hmm, for a rc-divider, don't you put the c's in the opposite direction? like 1p || big R on one side, 18p || low R || opamp on the other side
<electronic_eel>
just like the schematics of my la input sim
<azonenberg>
hmm
<azonenberg>
i guess you're right, since bigger C is less reactance
<azonenberg>
But at this point aren't we just copying the standard cheap 1M scope probe circuit?
<azonenberg>
Shouldn't we be able to get much lower parasitics since it's a fet buffer input stage?
<electronic_eel>
I think that is what common active probes do, but then they don't have a divider in front anymore, meaning much lower allowed voltage range
<azonenberg>
i dont mean like the super high end stuff like the lecroy wavelink
<azonenberg>
the tetris is 1M || 900 fF, 10:1, has a +/- 8V dynamic range and +/- 12V offset range
<azonenberg>
and will survive +/- 20V
<electronic_eel>
ok, so if we were able to replicate their design with a 20:1, that would bring us in the right direction
<azonenberg>
if they have a 10:1 attenuation then they only need to handle +/- 1.2V at the amplifier assuming the active component is in unity gain mode
<electronic_eel>
the question is how do they do their 10:1 divider without adding lot's of capacitance
<electronic_eel>
and being usable in the ghz-range
<azonenberg>
Correct
<azonenberg>
I'm targeting a few hundred MHz here with the ADA4817
<azonenberg>
but even that will take some work
<electronic_eel>
have you opened your tetris and took a look at the basic input configuration?
<electronic_eel>
might give us some insight
<azonenberg>
No
<azonenberg>
the housing that connects to the scope is easily openable but probably just contains the id eeprom and some bias generation stuff
<azonenberg>
the fun bits are all in the handpiece and that appears to be welded shut
<electronic_eel>
damn
<azonenberg>
there's no obvious fasteners and just what looks like mold flash around the perimeter
<azonenberg>
it almost looks like they just cast the board into a block of plastic
<azonenberg>
that probably isn't what actually happened, but i see no way to open it short of a dremel
<azonenberg>
and i am not doing that to a $2000 probe that works fine
<electronic_eel>
maybe something below a label?
<azonenberg>
there's no label or anything
<electronic_eel>
I have seen lot's of screws hidden behind labels
<electronic_eel>
do you have a xray?
<electronic_eel>
that should also show the traces
<azonenberg>
I do not have an xray
<azonenberg>
sooo i took a closer look at the tetris handpiece
<azonenberg>
it looks like there are two pieces of plastic
<azonenberg>
one is the strain relief and base area, the other is the body and tip
<azonenberg>
they slide over the board and envelop it
<azonenberg>
right at the junction where i would expect to see a little plastic snap or something there is a hole in the light gray plastic
<azonenberg>
on either side
<azonenberg>
with dark gray, very smooth material showing through
<azonenberg>
it looks like they put the strain relief on the cable, soldered the cable to the board, put the body on, pushed them togther
<azonenberg>
then filled the probe with potting compound through the two big holes and two smaller (air relief?) holes closer to the tip
<electronic_eel>
hmm, wouldn't filling the probe with goo affect the rf performance quite a bit, and in irregular ways?
<electronic_eel>
or did the put some kind of shield around the actual traces?
<electronic_eel>
just read the tetris datasheet from pmk. there is nothing about the +-12v offset range you wrote about
<electronic_eel>
just the +-8v dynamic range
<electronic_eel>
is that maybe some lecroy thing they do in the scope?
<azonenberg>
this was the lecroy zs1500 datasheet
<azonenberg>
its possible the zs1500 is customized to include an offset amplifier that the base tetris lacks?
<azonenberg>
it does have a big probus box at the other end
<azonenberg>
maybe the base tetris ties the other amp input to ground and the zs1500 has an i2c dac driving it?
<electronic_eel>
maybe something like this
<electronic_eel>
when controlled conveniently from the scope software this would be a nice feature to have too
<azonenberg>
yeah i mean i know for a fact the zs1500 has scope controllable offset
<azonenberg>
i've never used an oem PMK version
<azonenberg>
ooh
<azonenberg>
lecroy apparently has a product called "maui studio" which is basically the full scope UI you can run on any windows PC, for offline waveform analysis
<azonenberg>
apparently it comes with demos of a lot of software options
<azonenberg>
i'm gonna grab a copy of it and play with it to see how scopehal stacks up
<bvernoux>
it is old high end SA up to 26.5GHz defective on YIG ...
<bvernoux>
so far 658USD ...
<bvernoux>
but I think it will go up to 2KUSD or more
<bvernoux>
woo wining bid 1691USD ;)
<bvernoux>
depending on what is defective it can be expensive ...
<miek>
i've given up on looking at US deals - they're tempting but always end up too expensive after taxes/shipping
<bvernoux>
yes
<bvernoux>
each time they ask 400USD for Shipping ...
<bvernoux>
in EU
<bvernoux>
my VNA ask cost a lot because of that ;)
<bvernoux>
1/4 of the price ;)
<bvernoux>
as I was very lucky to obtain it for 2100 USD fully calibrated with all options ;)
<miek>
nice
<bvernoux>
as I see today the price is more >4KUSD for something not calibrated ...
<bvernoux>
Yes it take lot of time to have the good deal ;)
<bvernoux>
especially for those very old HW which can break every day ....
<bvernoux>
especially SA E440x they have lot of issues
<bvernoux>
I will do not recommend anyone to spend more than 2KUSD in an E4405B or more
<miek>
i took a rather large risk on my vna, it was £1k faulty.... :p though i've seen them go for £5-10k working
<bvernoux>
yes
<bvernoux>
sometime it pay to take risks as often it can be repaired
<bvernoux>
but you need the hw to check and repair it ;)
<bvernoux>
on SA 26.5GHz it is tricky sometimes ;)
<bvernoux>
So far I do not have other HW to check at such high freq
<bvernoux>
miek, what the freq range on your VNA ?
<miek>
3GHz, but it's a newer series - E5062A
<bvernoux>
ha ok
<miek>
i also did a bit of reverse-engineering and enabled the time-domain option on it :)
<bvernoux>
hehe great
<bvernoux>
I suspect it is again License code to enter ?
<bvernoux>
If anyone is interested I can generate any license code on Spectrum Analyzer E440x ;)
<bvernoux>
I have fully reversed the ROM code in 68040 ASM ;)
<miek>
yeah, though i worked out that there's a fixed secret key that lets you either enable it over scpi or generate the key over scpi :D
<bvernoux>
haha
<bvernoux>
for E440x it is a bit more complex ;)
<bvernoux>
but it is based on FlexLM license
<bvernoux>
like tons of old or new instrument using license in fact
<bvernoux>
What I like on those old Instrument is they are robust and we can find CLIP with full schematic and documentations are amazing
<bvernoux>
but bad things is they are heavy loudy pretty slow ;)
<bvernoux>
and it is often required to write our own sw to communicate with them over GPIB ;)
<miek>
yeah, i do like that about the older HPAK gear
<bvernoux>
it let me think about old school demo on Amiga, Atari ;)
<bvernoux>
where everything was optimized and even done in ASM ;)
<azonenberg>
also wow
<azonenberg>
looking at the setup
<miek>
i was looking through the vna service manual and realised that the replacement part for the test set is the same whether you have the T/R option or the full s-params option... so i enabled that too :) i still need to cal it to get it working right though
<azonenberg>
maui studio is literally just xstreamdso with different license config
<azonenberg>
they have checkboxes for installing touchscreen drivers, updating DSO board microcode, wavepulser stuff, etc
<azonenberg>
it's literally just their oscilloscope firmware
<bvernoux>
miek, yes the cal can be tricky as it often requires lot of different Instrument ...
<azonenberg>
that's what i expected, but still nice to see
<bvernoux>
miek, but up to 3GHz I think it is still possible with not a huge budget
<bvernoux>
miek, it is harder for 6GHz or more ;)
<miek>
bvernoux: yeah, as far as i can see it just needs a power-output cal for the other port so hopefully it should be doable. just not had time to write the scripts yet :)
<miek>
azonenberg: hah, nice
<bvernoux>
miek, it is good to have some nice reference attenuator... with S2P expected values ;)
<bvernoux>
azonenberg, yes ultra configurable SW
<bvernoux>
azonenberg, I think HP was the first on their old HW to do that ;)
<bvernoux>
which was requiring to change EEPROM in very old days ;)
<azonenberg>
so what i want to see is, can i connect it to a real scope?
<bvernoux>
I'm impatient to receive my cheap "2.92mm Edge Launch Connector Southwest Microwave" ;)
<bvernoux>
first time I have 2.9mm things haha
<miek>
bvernoux: yeah. i sent off my DIY cal kit to a friend with a much nicer vna, but i wish i'd sent a nice attenuator too! next time..
<azonenberg>
if not, offline analysis of waveforms from the real scopes with lots of software options will be cool to try
<bvernoux>
they mate with SMA but we shall be very careful to do not destroy it ...
<azonenberg>
see how all of the stuff i cant afford works :p
<azonenberg>
bvernoux: 2.92 mm mates with sma?
<azonenberg>
i thought only 3.5 did
<bvernoux>
azonenberg, yes
<azonenberg>
3.5 is just sma with air dielectric right?
<bvernoux>
it is specified 2.92mm mate with 3.5mm and SMA
<bvernoux>
but with good SMA ;)
<bvernoux>
as tolerance are very important
<azonenberg>
do you have proper connector gauges? :p
<bvernoux>
2.92mm is also like 3.5mm ;)
<bvernoux>
azonenberg, so far I need to buy some gauges ;)
<miek>
so i think you'll find this funny: the original vna fault symptom was that on power-on it would spin the fans and show nothing on the screen, so i was pretty confident that it wasn't going to be a problem in the scary RF side
<bvernoux>
if anyone have link to good Dial Gauges ;)
<miek>
when i got it, i tried the VGA out in case it was a bad screen: nothing. i followed the troubleshooting guide and it narrowed down to either a faulty motherboard or bad bios settings
<bvernoux>
ha
<miek>
i couldn't get into the bios, so of course tried the next best thing: pulled the bios battery, waited a while, put it back and booted up....
<miek>
and it worked.
<bvernoux>
haha
<bvernoux>
so there was no any HW issue ?
<bvernoux>
just a bug with BIOS ?
<miek>
yep!
<bvernoux>
excellent
<bvernoux>
as the worst case is when something is fried on the input
<bvernoux>
there is often some custom chipset with protection/attenuator ...
<bvernoux>
for info about mating different connectors
<bvernoux>
so in the table we can mate 2.92mm to SMA ;)
<bvernoux>
they are defined as compatible
<bvernoux>
without adaptor
<bvernoux>
Anyway so far I only have SMA cables ;)
<bvernoux>
I have not even bought Type N as I use converter Type N to SMA on my instruments
<bvernoux>
I will probably win to use 7mm to type N on my VNA instead of 7mm to SMA as performance shall be a bit degraded especially VSWR ...
<bvernoux>
or even better direct 7mm to 7mm with 7mm cables hahaha
<bvernoux>
the price are just crazy for such old connectors
<azonenberg>
ok maui studio installed, waiting for my license file (sounds like it's a manual review process that can take up to 24h)
<bvernoux>
azonenberg, what have you upgraded ?
<azonenberg>
bvernoux: maui studio is basically the lecroy scope firmware packaged up for standalone use on a windows PC
<azonenberg>
for offline analysis
<bvernoux>
ha ok
<azonenberg>
I just found out about it today, not sure if it's a new thing they released because so many of their customers were working from home or what
<bvernoux>
they ask money for that too ?
<azonenberg>
the base software is free, and it includes demos until the end of the year of "50+ software options"
<azonenberg>
unclear which ones
<azonenberg>
But i intend to find out
<bvernoux>
ha ok
<bvernoux>
so at least you can evaluate it freely at start
<azonenberg>
My plan is to compare capabilities of it to glscopeclient for the various protocols we've got decodes for, etc
<azonenberg>
figure out what we do better, what we can improve on, etc
<bvernoux>
yes very interesting
<bvernoux>
as it seems LECROY are one of the best for Scope
<bvernoux>
I have seen some training ;)
<bvernoux>
and the UI seems pretty nice
<azonenberg>
compared to pico's ui?
<bvernoux>
what is awful is the time to start the Application
<azonenberg>
Which makes me want to smash the thing with a hammer? :p
<bvernoux>
haha yes Picoscope UI is horrible in comparison ;)
<azonenberg>
no picovna ui
<azonenberg>
it's worse
<bvernoux>
yes picovna seems the be the worst of all ;)
<azonenberg>
And actually glscopeclient has got a bit slow to initialize lately. I have a ticket filed to investigate this, but haven't sat down to debug
<bvernoux>
but Picoscope is pretty ugly too
<bvernoux>
but usable ;)
<azonenberg>
picovna's hardware is at least better than the xavna
<bvernoux>
ha yes sure
<azonenberg>
When it works, it gives much better data
<bvernoux>
but it is not same price
<azonenberg>
the software just makes me want to tear my hair out
<bvernoux>
I shall do a comparison ;)
<bvernoux>
I have my attenuator to check ;)
<azonenberg>
i was watching some training videos on the lecroy vna/tdr (wavepulser 40ix)
<bvernoux>
as they are pretty good up to 4GHz
<azonenberg>
which goes out to 40 GHz and has a MSRP around 50 kUSD
<azonenberg>
but it looks AMAZING
<bvernoux>
I still do not understand why S21 is so good ;)
<bvernoux>
I cannot think the PCB is not FR4
<azonenberg>
I'll probably never be able to afford one
<azonenberg>
But i can dream :p
<bvernoux>
azonenberg, yes I'm registered to their training ;)
<bvernoux>
done by a french guy
<bvernoux>
in french haha
<azonenberg>
oh a locwal apps engineer?
<bvernoux>
I have missed De-embedding methods in frequency and time domain – TDR advanced
<azonenberg>
afaik their only R&D site is in cherry hill NY
<azonenberg>
and so all of the stuff i've seen on youtube by their engineers was in english
<azonenberg>
Several of the wavepulser videos were presented by none other than eric bogatin
<bvernoux>
I have done the training with Eye Diagram on a big Lecroy scope ;)
<bvernoux>
40GSPS ;)
<bvernoux>
pretty nice
<bvernoux>
it was quite intuitive to use it
<azonenberg>
My HDO9204 gets 40 Gsps. But only 2 GHz b/w
<azonenberg>
and yes MAUI is a very nice UI. My favorite scope UI by *far*
<bvernoux>
the 2GHz BW is just a SW limitation
<bvernoux>
40GSPS can do at least 5GHz BW
<azonenberg>
i'm trying to replicate that intuitive feel in glscopeclient while optimizing for mouse and keyboard instead of touchscreen
<azonenberg>
bvernoux: the hdo9k series only goes up to 4 GHz
<azonenberg>
i suspect it's software limited below there but the frontend is the limit past that
<bvernoux>
ha they have limited it by SW ?
<azonenberg>
well fpga filter limited*
<bvernoux>
as all scope I have seen so far have GSPS/10 for their BW
<bvernoux>
and it is always SW option
<azonenberg>
you should see lain/monochroma's dda5005a lol
<azonenberg>
it's 10 Gsps, i think maybe 20 Gsps interleaved
<azonenberg>
and 5 GHz bw
<bvernoux>
I do not speak about sampling scope of course ;)
<Degi>
re probes azonenberg: For usb data you can use softusb
<azonenberg>
with all 4 channels, you are literally right at nyquist
<azonenberg>
so really the only way you can actually use the 5 GHz bw is equivalent time mode
<bvernoux>
yes 5GHz BW on 10GSPS is not realist ;)
<azonenberg>
That said, it does 200 Gsps equivalent so it's quite nice for SI work
<miek>
mine's got a fixed filter on the diff pairs to the adc :( i haven't been brave enough to open it up and upgrade it yet..
<bvernoux>
miek, what is your scope ?
<azonenberg>
There is a guy in another channel who is INSANE
<miek>
agilent mso6034a
<azonenberg>
he had a keysight (well, agilent) scope that was i think 200 MHz
<bvernoux>
I know rigol use a chipset to limit the BW which is programable
<bvernoux>
I suspect lot of other manufacturer do that too
<azonenberg>
He upgraded it to 1 GHz
<bvernoux>
like that it is upgradable by SW
<Degi>
Hmh instead of all that usb stuff we could just detect if a probe is attached (for example checking if a resistor is on a pin) and if so use a custom protocol (I2C?)
<azonenberg>
it was about 150 components swapped in the frontend
<miek>
oof
<Degi>
But usb would be nice too then you can directly attach the probe to pc
<Degi>
lol
<azonenberg>
then he went to 1.5 GHz
<azonenberg>
there was never a 1.5 GHz version of that scope
<Degi>
haha
<miek>
lol
<azonenberg>
but if you set strap pins right you could tell the software it was 1.5
<azonenberg>
and he managed to get frequency response kinda decent
<azonenberg>
it seems like for marketing reasons that version was never released
<bvernoux>
azonenberg, haha nice hack ;)
<miek>
hah, wow
<miek>
mine's 300MHz 2GS/s. the 500MHz 4GS/s version is just a filter change & some strap pins. the 1GHz one has a lot different, probably similar to that guy's hack
<azonenberg>
that may be the same line
<azonenberg>
there was a very long thread on the eevblog forums about it
<azonenberg>
... you gotta be kidding me, lol
<azonenberg>
decided to poke around a bit in the maui install directory while waiting for my key so i can actually run it
<azonenberg>
they even have fpga bitstreams for some of the scopes ("Janus" and "Yarmouth") in there
<miek>
lol
<miek>
encrypted?
<azonenberg>
checking the idcode (they're raw binaries. not .bit's) Yarmouth has an xcku085
<azonenberg>
So does Janus
<azonenberg>
They do not appear to be encrypted, there's lots of 0x00 bytes i found while grepping for 0x93 which is the xilinx jedec id
<monochroma>
:D
<monochroma>
azonenberg: yeah they upgrade the FPGA firmware on software updates
<monochroma>
or, optionally can
<monochroma>
there is a tool that gets installed you can access from the desktop iirc that does the FW updoots
<azonenberg>
i just find it amusing how much bloat there is for a software distribution that doesnt actually run on a real scope
<azonenberg>
There's also a directory "SerialFPGAs"
<azonenberg>
which contains half a gigabyte of bitstreams
<monochroma>
i think they just have like, one SW distribution across all their HW
<azonenberg>
looks like all of the 64/66 and 8b10b triggers are a) in hardware and b) different bitstreasms for each config
<azonenberg>
using GTXes
* monochroma
suddenly notices the lack of zigggy
<Degi>
GTX?
<azonenberg>
(judging solely by filenames)
<azonenberg>
xilinx serdes
<Degi>
Ah yes
<azonenberg>
monochroma: ziggy hasn't been here for a while
<Degi>
Oh neat they can do 6466
<azonenberg>
I will be implementing that in scopehal as soon as i get some waveforms of 10GbE to play with
<azonenberg>
But i havent got my hands on a fast enough scope yet :p
* monochroma
ponders tapping a 10GbE line with her scope
<azonenberg>
monochroma: you only have 20 Gsps right?
<azonenberg>
10.3125 Gbps seems like you wouldn't quite have the b/w to decode
<monochroma>
bah
<azonenberg>
(and that's interleaved)
<azonenberg>
sata, pcie gen 1/2, etc should be doable
<azonenberg>
10GbE is just a touch over your limit
<monochroma>
yeah
<Degi>
Hm how much BW do you have
<Degi>
I mean 10.3125 is still below nyquist for 20 GS/s
<azonenberg>
Degi: to do clock recovery you're going to need more than one sample per bit
<monochroma>
20GS on 2 channels, 5GHz frontend
<azonenberg>
unless you sample phase coherent
<azonenberg>
at 10.3125 you get 2 samples on some bits, 1 on others, and i dont think you could reconstruct the waveforms fully
<Degi>
I mean isnt below nyquist sufficient
<Degi>
At 10.3125 the frequency would be like 5 GHz and nyquist at 20 GS/s is 10 GHz
<azonenberg>
i would want at least 4x oversampling to do reliable CDR
<azonenberg>
you're just below 4x
<azonenberg>
10.00 Gbps is what i would consider the limit for 20 Gsps
<azonenberg>
My waverunner 8104 does 20 Gsps on 2 channels / 10 Gsps on 4, with 1 GHz bandwidth, and i was able to do 1.25 Gbps comfortably but i wouldn't have wanted to push much higher
<azonenberg>
... oh god "maskdatabaseeditor.exe" has the visual studio '98 MFC icon
<azonenberg>
"MSO.bit" - no idea for what - is an xc5vlx30t
<azonenberg>
obviously some kind of LA
<azonenberg>
wonder which gen of scope
<azonenberg>
also i see what looks like a DLL for each software option
<azonenberg>
they probably have a COM server for each decode or measurement/group of measurements, based on what i've seen in the past
<azonenberg>
yes, this is a nightmare :p
<azonenberg>
93R resistors are here
<azonenberg>
Will try them out after work
<azonenberg>
also poking around in device manager... looks like the waverunner8 acquisition board is named Magellan
<azonenberg>
but so is the HDO9. so that might be a family, not sure. The WR8 and HDO9 have a lot in common but are not the same exact analog board
<azonenberg>
also it looks like the MSO on my wr8 connects via USB
<azonenberg>
anyway from this quick delve into it, xstream/maui is built exactly how i would expect something made of legacy piled on top of legacy going back to the visual studio 98 days to look
<monochroma>
hehe
<azonenberg>
who needs an anti-reversing clause in the EULA when you can just make your code such a trainwreck nobody *wants* to reverse it? :p
<monochroma>
haha
<azonenberg>
i put maui in the same class as ISE
<azonenberg>
by all means figure out interesting things from looking at high level inputs and outputs, UI, etc
<azonenberg>
but don't even think about getting lower level or you'll regret it
<azonenberg>
Not that i had any interest in REing maui before this, but any doubts i had are now gone
<electronic_eel>
azonenberg: can you load arbitrary sample data (like not taken from an lecroy scope) into the maui sw and analyze it?
<electronic_eel>
or do they ensure that the data was sampled by a lecroy scope, for example by using special file formats and such?
<azonenberg>
They expect their "trace" or "labnotebook" file formats
<azonenberg>
that said, i suspect the trace format would be pretty trivial to RE. might even be documented
<azonenberg>
initial observation: zooming in maui with a mouse is *painful*
<azonenberg>
Some decodes are drawn over waveforms and actually look kinda ugly