<azonenberg>
Flat +/- 1 dB to 1.13 GHz, -3 dB BW of 1.63 GHz
<azonenberg>
with the 018-291-105 ground leaf
<azonenberg>
I kinda expected somewhat worse performance when i moved the ground further away
<azonenberg>
i'm pretty confident i can push back above 2 GHz with the new SMA though. I may also add a second ground right next to the tip for higher frequency use but less flexibility
<azonenberg>
you'll be able to put a second tip needle or similar in it
<_whitenotifier-c>
[scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±5] https://git.io/Jf2E5
<_whitenotifier-c>
[scopehal-apps] azonenberg f521d50 - WaveformArea: dragging waveform to the far bottom/right of a waveform area now creates a new group. Fixes #102.
<_whitenotifier-c>
[scopehal-apps] azonenberg closed issue #102: Dragging a channel to the right/bottom of the current waveform area should create a split - https://git.io/Jf2mP
bvernoux has joined #scopehal
<bvernoux>
azonenberg, nice update on the probe on Tweeter& KS
<bvernoux>
It just need few backer to reach 4KUSD Goal
<azonenberg>
Yep
<bvernoux>
Your test board RO4350B + Immersion Silver look very nice
<bvernoux>
do you have S11 / S21 VNA results on it especially for THRU ?
<azonenberg>
Yes. and they're awful
<bvernoux>
ha really ?
<azonenberg>
the impedance match on the samtec SMA is not very good
<bvernoux>
ha yes
<azonenberg>
i'm getting much better results with the amphenol one
<azonenberg>
i plan to respin the test board with it soonish
<bvernoux>
I was also thinking it will be the limiting factor
<bvernoux>
so you will switch to 26.5GHz connector ?
<bvernoux>
which is clearly better matched by design
<azonenberg>
well its also just easier to match to CPW even at lower freqs
<azonenberg>
vs the samtec one the pin is so far away (and huge)
<bvernoux>
yes
<azonenberg>
the other limiting factor is the grounding accessories. i'm thinking of putting a second ground socket on the probe about where the old design had it
<bvernoux>
yes one very close to Signal Pin
<azonenberg>
so you have your choice of where to ground it and trade frequency response against convenience
<bvernoux>
yes it is clearly the best
<bvernoux>
to reach better BW shorter GND
<bvernoux>
I have designed a footprint to match both SMA ;)
<bvernoux>
the cheap one and the 26.5GHz one ;)
<azonenberg>
yeah but how clean is the impedance match on both? :P
<bvernoux>
in fact it is optimized for 26.5GHz SMA ;)
<bvernoux>
and I suspect it will not change a lot on cheap big SMA
<bvernoux>
It is the idea to check the impact to see if one footprint could be used for both SMA type connectors
<azonenberg>
well let me know
<azonenberg>
i have two test boards out at oshpark right now
<azonenberg>
one is a thru line with the new SMA after optimizing in sonnet
<bvernoux>
Anyway OSHPARK are very disappointing for 4 Layers ENIG especially ENIG
<azonenberg>
i want to compare to both the initial thru line using the default amphenol footprint, and the simulations of both
<azonenberg>
i am expecting enig loss
<azonenberg>
but they're a cheap way to do 4-layer boards with halfway decent specs
<bvernoux>
yes
<azonenberg>
and i can quantify the enig loss and remove it from my measurements to some extent
<bvernoux>
but it is definitely not good for expensive 26.5GHz SMA connectors
<azonenberg>
anyway, the other board i have is a new version of the full probe
<azonenberg>
optimized for the new SMA with a different CPW geometry
<bvernoux>
even if in my case they cost about 4USD each
<azonenberg>
also, i found a potential manufacturing problem on the current probe iteration
<bvernoux>
the loss is terrible up to 3dB for 60mm
<azonenberg>
multech sent me the TDR test strip from the edge of the panel. The trace is bare but the ground has soldermask on it
<azonenberg>
and they didn't expand the mask enough
<azonenberg>
meaning part of the CPW trench has mask instead of air dielectric
<bvernoux>
ha interesting
<azonenberg>
so if they calibrated the impedance based on that, the impedance on the actual pcb might be wrong
<bvernoux>
and you see big impact because of that ?
<azonenberg>
I dont know because i have no way to measure just the impedance of my board easily. I guess i can try the time domain view in the VNA
<azonenberg>
to look at just the line and not the connector or attenuator response
<bvernoux>
Does they will do new board for free as they have done mistake ?
<azonenberg>
I'm talking with the engineers at the fab to see what the situation is
<bvernoux>
it will be interesting to compare
<azonenberg>
it's possible they only did one spin of the board and it was still within tolerance regardless
<bvernoux>
mistake always allow to understand new things ;)
<azonenberg>
anyway, the fab is aware of the issue and investigating
<azonenberg>
i sent them photos under the microscope
<bvernoux>
the fab is multec in USA ?
<azonenberg>
no
<azonenberg>
Multech PCB Technologies Co, Ltd in Shenzhen
<bvernoux>
ha ok
<azonenberg>
also have you seen any of the recent UI tweaks to glscopeclient?
<bvernoux>
I'm tempted to launch your test probe board v0.1b RO4350B Immersion Silver with my SMA connectors
<bvernoux>
I have some free board for RF test with PCBWay
<azonenberg>
i'm probably going to do a new video soonish showing some of the capabilities i've added lately
<azonenberg>
timebase configuration, drag and drop of waveforms
<azonenberg>
drag and drop of protocol decode overlays is next on the TODO
<bvernoux>
ha yes very nice
<bvernoux>
and something to import sigrok files ;)
<bvernoux>
maybe just VCD
<azonenberg>
yes vcd import is nice to have
<azonenberg>
oh and just general massive performance speedups etc
<bvernoux>
I could test that natively soon with my next SSD ;)
juli964 has joined #scopehal
<bvernoux>
I will have 500GB for Linux Ubuntu 18.04LTS and 1GB on other side for zindoz stuff
<azonenberg>
you mean 1 TB? :p
<azonenberg>
Speaking of which, at some point i would like someone to work on getting glscopeclient to run on windows
<azonenberg>
the GTK and OpenGL stuff is all cross platform but i know of at least a few little bits that will need some work to run on win
<azonenberg>
likely not a huge amount of work for someone who does windows dev though
<bvernoux>
I have never used GTK on my side
<bvernoux>
I mainly use Qt for UI stuff
<bvernoux>
Also so far the only things which could work is my Rigol Scope
<azonenberg>
None of the UI stuff should need porting, at least once you get a build environment set up
<bvernoux>
but IIRC USB is not supported ?
<azonenberg>
A USBTMC driver needs to happen
<azonenberg>
I don't know what the current status of it is
<azonenberg>
afaik right now we have raw TCP and VICP and somebody is working on LXI
<bvernoux>
IIRC there is a way to replay files too ?
<azonenberg>
There is file load/save support, yes. Including waveform history and all instrument settings
<azonenberg>
when opening a file you can either reconnect to the scope or work offline
<azonenberg>
not all scope settings, most notably timebase and trigger config, are serialized at the moment
<bvernoux>
ha ok great so it is a way to check basic things with that without HW
<azonenberg>
Yes
<bvernoux>
even if the most important is to check realtime stuff with scope in 2nd step
<bvernoux>
Support for Picoscope is also hard I think as the Pico lib are bloated
<bvernoux>
and not open source
<bvernoux>
it's like binary blob
<azonenberg>
Yeah. honestly i wonder if the best option would be to RE the usb protocol :p
<azonenberg>
but as long as the blob can be linked to without any licensing issues, it's not unreasonable to use it if available
<bvernoux>
yes and doing a libusb driver for that
<bvernoux>
with advantage as it could work on all platforms
<azonenberg>
at least until an open driver can be made
<bvernoux>
So far I do not have lot of time to do that but I can help with USB 3.0 traces ...
<bvernoux>
using Beagle USB 5000
<bvernoux>
If I'm not alone to do all ;)
<azonenberg>
feel free to acquire some data, but probably not worth your time until someone has time to work on a driver
<azonenberg>
most likely though, that person will need a picoscope to test on
<azonenberg>
at which point your data isnt going to be too useful since they can just sniff themself
<bvernoux>
yes
<bvernoux>
I'm also afraid different Picoscope version have big difference in USB packets ...
<azonenberg>
Very likely if that was not intended to be a stable API
<bvernoux>
as there is lot of lib to be used for each Pico serie
<bvernoux>
but at the end it is just a Python 2.x wrapper on Pico lib ...
<azonenberg>
lol
<bvernoux>
yes ...
<bvernoux>
It is main issue of Pico stuff they are so closed bloated to be used freely at low level
<bvernoux>
with blobs
<bvernoux>
Pico=Good Hardware for the price but very bad SW stuff ;)
<bvernoux>
I will probably end to buy a RS Oscilloscope ;)
<bvernoux>
as Anyway Picoscope are clearly limited for GHz BW stuff
<bvernoux>
they sell SamplingScope ..
<azonenberg>
yeah i am not a fan of the pico vna software either
<azonenberg>
why R&S?
<azonenberg>
so far i prefer lecroy's UI to any other scopes i've used recently
<bvernoux>
not R&S but Lecroy ;)
<bvernoux>
as they have some good offer
<azonenberg>
R&S seems to make really nice RF and power stuff
<azonenberg>
i've been happy with their multimeters and power supplies so far
<bvernoux>
latest offer from them is WAVESURFER 4104HD for 9995Euros ;)
<azonenberg>
that's a pretty wimpy scope by my standards
<bvernoux>
yes it is their low end scope
<bvernoux>
IIRC with 5GSPS
<azonenberg>
yeah. but 12 bits which is unique among entry level bandwidth scopes
<bvernoux>
If I buy one I prefer 20GSPS
<bvernoux>
20GSPS
<azonenberg>
Lol yeah
<azonenberg>
i see a waverunner 8054 on ebay (sold by lecroy as a factory refurb) for $9800
<bvernoux>
I think I prefer 8bits 20GSPS or more than 12bits 5GSPS ;)
<azonenberg>
that's 500 MHz (lowest bandwidth in the line) but a midrange scope, 8 bit 10/20 Gsps
<azonenberg>
i have the 1 GHz version of the same scope, cost me $12K as a retired rental fleet unit with a bunch of software options on it
<azonenberg>
been very happy with it
<bvernoux>
ha yes it seems to be a good things
<Degi>
Ugh kinda hate that scope manufacturers do this whole thing of making good hardware and then software or simple hardware change limiting the bandwidth etc. (like swapping resistors)
<bvernoux>
and could be even better if it can be hacked to unlock 2GHz ;)
<bvernoux>
as 20GSPS can clearly do 2GHz
<azonenberg>
on the higher end, lecroy ebay also has a waverunner 8404M-MS which is 4 GHz, 8 bit, 20/40 Gsps, with a 16-channel 1.25 Gsps LA
<Degi>
They have one with 30 GHz too lol for like 500k
<azonenberg>
They want $20K for that and if i had just a little more spare cash i'd be seriously tempted to sell my 8104 and trade up to that
<azonenberg>
it would probably only cost me about $10K to upgrade if i sold the 8104 at fair market value
<bvernoux>
ha yes it is clearly a beast
<azonenberg>
it's the same series as the one i have, just the full bandwidth model
<azonenberg>
note that the waverunner 8k series actually has two sub-models
<azonenberg>
the 8000 and 8000M
<bvernoux>
and it seems pretty compact too
<azonenberg>
the -M has more memory and double the sample rate
<azonenberg>
The other option i see on ebay in the same roguh performance class is a HDO9304 for $21623
<bvernoux>
what is strange is they advertize 4GHZ BW but I see there is BNC connectors
<azonenberg>
that's a slightly larger screen (same resolution, just larger)
<azonenberg>
only 3 GHz BW but 10 bit resolution
<azonenberg>
my other scope is a 9204, the 2 GHz version of that line
<azonenberg>
i wish the display was 1080p at that size and not 720p, but it's not BAD
<azonenberg>
just seems like it'd be an easy upgrade
<Degi>
Hmm if scopehal can be used that good with the mouse, it should work fine for touchscreens too, right? (I haven't tried it yet since it didnt wanna work with the rigol, might try to fix that sometime)
<bvernoux>
yes 10bits vs 8bits is clearly not bad
<azonenberg>
Degi: i make heavy use of right click and mouse wheel etc
<azonenberg>
i expect a full 3 button mouse
<azonenberg>
we'd need to come up with gestures etc to replace all of that
<azonenberg>
it likely could be done but is not a current priority
<azonenberg>
bvernoux: and yes. Lecroy uses BNC connectors, or more precisely ProBus, up to 4 GHz inputs
<Degi>
Hm yeah I guessed. But for later on that would be great
<bvernoux>
azonenberg, BNC connector are not designed to exceed 3GHz in theory IIRC it is why it is strange to see that
<azonenberg>
(ProBus = BNC, selectable 1M/50 ohm termination, and six pin +/- 12V, presence detect, ground, I2C active probe connector)
<azonenberg>
i see a lot of BNC stuff labeled as 4+ GHz
<Degi>
I once soldered a magnetron to a BNC... Still untested at high power for leakage
<bvernoux>
so BNC is intended to go up to 4GHz max
<azonenberg>
digikey has one that claims 18 GHz although i doubt THAT :p
<Degi>
>Loll
<bvernoux>
yes 18GHz BNC is clearly a big joke ;)
<azonenberg>
its a 75 ohm which is odd
<Degi>
Maybe with several dB los
<azonenberg>
115-031-6861-ND is a "12 GHz" 50 ohm BNC
<bvernoux>
I see in the list HD-BNC which can go up to 6GHz
<azonenberg>
bvernoux: anyway, beyond 4 GHz lecroy moves to ProLink inputs which are SMB based and 50 ohm only
<Degi>
Huh 1.3 VSWR to 12.4?
<azonenberg>
Some of their higher end scopes are dual input stage and for each channel have both a ProLink and ProBus frontend you can switch between, feeding through a mux to the same ADC
<azonenberg>
so you don't totally sacrifice the convenience of using lower end probes while still having high speed inputs
<Degi>
Hm I think for our scopes we should make a 1 MOhm to 50 ohm converter sometime, 1:1 or so. Because then most conventional accessoires can be used
<azonenberg>
Degi: yes, but would likely have to be active
<Degi>
Yes ofc lol
<Degi>
Otherwise it'd be like 1:20000
<_whitenotifier-c>
[scopehal-apps] azonenberg pushed 2 commits to master [+0/-0/±4] https://git.io/Jf2zz
<bvernoux>
Anyway before to buy a 20KUSD scope I will need to push lot of project ;)
<_whitenotifier-c>
[scopehal-apps] azonenberg 346a953 - WaveformArea: fixed some UI glitches around channel dragging
<_whitenotifier-c>
[scopehal-apps] azonenberg 525c19d - ChannelPropertiesDialog: added widget for selecting display color of waveforms. Fixes #34.
<_whitenotifier-c>
[scopehal-apps] azonenberg closed issue #34: Allow user to change channel color - https://git.io/Jf2zg
<bvernoux>
I checked the spec of VOLLUM it use AD9213
<bvernoux>
10G
<bvernoux>
this ADC cost about 3.9KUSD
<bvernoux>
glurps
<Degi>
Idk maybe by the time we get there there will be other ones
<Degi>
5470 € for 1 pc on mouser
<bvernoux>
yes it is just during test if you do a mistake you burn 4KUSD in 1s ;)
<Degi>
yes lol
<azonenberg>
Yes. VOLLUM is intended to be in the same performance class as the HDO8108A which is a $$$ scope
<Degi>
Hope by that time we wont make any mistakes or have a ton of money
<azonenberg>
it's 8 channels 1 GHz 10 Gsps 12 bit
<azonenberg>
i dont have a price for it, but the 350 MHz version costs $26600
<bvernoux>
Degi, we always make mistake ;)
<azonenberg>
so i expect the 1 GHz version is in excess of 30K USD
<Degi>
We didn't burn up any semiconductors yet
<Degi>
Huh wpw
<Degi>
We can parallel them too, 6.5 GHZ BW
<azonenberg>
probably 35-40
<azonenberg>
Yes. That's MURDOCK
<azonenberg>
interleaved AD9213s
<Degi>
Big monies
<azonenberg>
that scope will cost as much as a house if we pay MSRP for the ADCs
<Degi>
Maybe we can make our own
<azonenberg>
So it's very likely to never happen :p
<azonenberg>
But we're also a long ways from even attempting it
<azonenberg>
even VOLLUM will cost as much as a nice car and that's to build one unit assuming no respins
<Degi>
I dunno if it takes a few years maybe price will fall or better ADC out
<azonenberg>
we're talking somewhere around 30 kUSD of ADCs plus a bunch of virtex ultrascales and 12+ layer PCBs
<Degi>
oh geez
<Degi>
Can we have a rly thick PCB and use that as part of the housinjg
<azonenberg>
Degi: lol
<azonenberg>
let me put it this way, those ADCs spit out 120 Gbps of data *each*
<Degi>
Hm not sure if 12 layers make much sense
<azonenberg>
i'm thinking about fanning out all of the pins on the FPGA
<azonenberg>
just a random guess
<Degi>
We can make like 4+4+4 layers and inbetween stack components and PCB-to-PCB connectors
<azonenberg>
We're going to need probably a sodimm of ddr4 dedicated to each channel
<azonenberg>
the ADC is i think 16 lane JESD204
<Degi>
Hm yes, a bit of upgradeability
<Degi>
DDR4 is fast enough?
<Degi>
yes 16 lane
<azonenberg>
might have to go dual channel but if we have one per ADC should be OK
<Degi>
Hmm 10.25*12 = a lot indeed
<azonenberg>
the issue is, you need an fpga with 16x 10g serdes and enough GPIOs to run a 64-128 bit ddr4 memory bus
<azonenberg>
per channel
<azonenberg>
that's probably kintex ultrascale+ territory at minimum
<azonenberg>
and a package with high hundreds to low thousands of pins
<azonenberg>
which means a very multilayer board
<azonenberg>
oh and you need enough serdes left over to talk to the root fpga that runs the ethernet interface
<azonenberg>
because we are not going to hang eight of these adcs off one fpga, that would mean 128 serdes
<azonenberg>
and like 16 lanes of ddr4 on one fpga
<azonenberg>
i dont even think an xcvcu13p could do that :p
<azonenberg>
but like i said i havent done extensive design work on vollum/murdock precisely because i expect to learn so much on the lower end scopes that any design work i do at this point will be undone
<azonenberg>
The next step on the hardware side is, and continues to be, designing the MEAD module and an INTEGRALSTICK-based testbench for it
<azonenberg>
Which afaik has no blockers, i've just been busy working on glscopeclient and not doing hardware stuff the last few weeks
<azonenberg>
there is actually a decent bit of progress already from when i started on CONWAY, MEAD is currently just a fork of the CONWAY design before we figured out anything about the frontend
<azonenberg>
maybe THIS weekend i'll work on it...
<Degi>
Oh geez
<Degi>
Between my grounded oscilloscope and grounded shelf is 44 V
<azonenberg>
...
<Degi>
EMP deluxe
<Degi>
The charger for the laser makes more EMP than triggering the flashlamp discharge
<Degi>
C
<Degi>
There must have been a reason why it was inside of a metal box before.
<azonenberg>
lol
<azonenberg>
cant imagine why that would be
<Degi>
I think me adding a metal mesh below it helped, at least dmesg doesnt show any errors related to USB devices going bad
<Degi>
Did you do characterization on the AFE and on the whole system yet?
<azonenberg>
Some. But more is needed
<azonenberg>
i can't be in 10 places at once though
<azonenberg>
so TDMA is needed :)
<Degi>
Hmm
<Degi>
:/ something doesnt work
_whitelogger has joined #scopehal
lain has quit [Remote host closed the connection]
lain has joined #scopehal
laintoo has joined #scopehal
<Degi>
I got a weird IC. A TCD1703C. But it is 2 cm shorter than the datasheet says
<Degi>
40 instead of 66 mm...
<Degi>
It looks pretty weird too
<Degi>
Should I msg the ebay seller?
<monochroma>
probably :D
<Degi>
I really wanna know what that is
<Degi>
It says "TOSHIBA 1901E76" "TCD1703C JAPAN" on the backside
<monochroma>
probably best case you got a used pull that they shaved the case down and re-labeled it and it IS what it says it is. worst case it's not the same chip at all :3
<Degi>
How do you relabel a ceramic IC?
<Degi>
Like would you go thru all that effort?
<Degi>
I mean its a CCD, the actual die is shorter
<monochroma>
if it's just an ink stamp, they def relabel those
<sorear>
2 *cm* ?!
<Degi>
Huh why'd you do that
<sorear>
that's a lot of IC to shave off
<Degi>
Its 4 cm long, datasheet says 6.6 cm
<monochroma>
but they pretty much only "refurbish" the epoxy case ones
<Degi>
Weird, these are long and narrow photodiodes
<Degi>
Thats no 7*7 µm
<azonenberg>
so probably a random other ccd that's relabeled