<electronic_eel>
bvernoux: some stuff from their datasheet that might be interesting for CONWAY:
<bvernoux>
yes
<electronic_eel>
the probes are connected with coax wire to small termination circuits that sit on the grabber probes
<bvernoux>
I'm tempted to buy the DSLogic U3Pro16 to test it ;)
<bvernoux>
I really doubt it can capture 1GHz signal
<bvernoux>
as accuracy is 2ns ...
<electronic_eel>
they say 250MHz analog bw
<bvernoux>
in fact they speak about 1GHz ;)
<bvernoux>
but it is the sampling rate ;)
<electronic_eel>
I think they mix GHz and GSPS
<bvernoux>
so I even doubt it can capture correctly 500MHz
<electronic_eel>
250 MHz bw and 1 GSPS seems realistic to me
<bvernoux>
yes exactly they mix GSPS vs GHz to confuse people
<bvernoux>
it is marketing hint ...
<electronic_eel>
the threshold voltages are 0-5V and the probes are protected for +-30V
<bvernoux>
yes
<bvernoux>
they have nice caracteristics
<electronic_eel>
they claim 250Kohm || 13pF impedance for the probes
<bvernoux>
in fact the same as Saleae 3 time less expensive
<electronic_eel>
I do not see if the threshold is settable per channel, only for the whole la or something in between
<bvernoux>
Saleae LogicPro16 spec is 500MSPS ;)
<bvernoux>
thye are more realistic
<bvernoux>
I really doubt about 1GSPS (specified as 1GHz) for DSLogic U3Pro
<bvernoux>
yes I suspect it is for whole LA or in best case by group of 4 or 8chan
<electronic_eel>
now isn't that something that is easily verifiable?
<electronic_eel>
I mean the 1 GSPS
<bvernoux>
yes ;)
<bvernoux>
I have signal generator up to 6GHz
<bvernoux>
can be easily checked
<bvernoux>
even just a basic clock @500MHz ;)
<bvernoux>
especially it will be interesting to capture some DDR signals
<bvernoux>
but 13pF capacitance / 250kOhms could be a problem too
<electronic_eel>
yes, 13pF is not a light load for a fast digital circuit
<bvernoux>
LogicPro16 specify fastest Digital Signal 100MHz with 500MSPS
<Degi>
Huh
<Degi>
Why are they so pricy?
<Degi>
Like if they even confuse GSPS and GHz?
<bvernoux>
check Saleae LogicPro ;)
<Degi>
I mean ok it says the sample rate in ghz..
<bvernoux>
they are 3x more expensive
<bvernoux>
for the same things
<electronic_eel>
Degi: compared to other LAs (that are not just rebadged FX2s) they are cheap
<bvernoux>
also DSView Sw is nice and have all Sigrok Decoders
<bvernoux>
with nice gui
<bvernoux>
better than Saleae at end with more decoders as they are inherited from Sigrok/PulseView ...
<bvernoux>
yes especially with USB 3.0 Controller
<bvernoux>
and the 1GSPS ;)
<bvernoux>
I have never seen that on a FPGA
<bvernoux>
IIRC max is 500MHz
<bvernoux>
so 500MSPS
<bvernoux>
and I really doubt they use an ASIC ;)
<electronic_eel>
maybe a cypress FX3?
<bvernoux>
FX3 is bad for capture
<bvernoux>
but yes for USB3 side they are ;)
<electronic_eel>
not directly, coupled with fpga
<bvernoux>
there is not lot of choice for USB3 controller ;)
<Degi>
Like you could get 1 GS/s and stick a gig of DRAM onto it for maybe 50 €? And then coax cables to probes and a digital buffer IC in the probe... that should be doable
<bvernoux>
they have 2Gbit of RAM
<bvernoux>
digital buffer IC running at 1GSPS is not cheap
<bvernoux>
so I really think it does not run at such speed ;)
<bvernoux>
500MSPS ;)
<Degi>
I mean you can get 4 Gbit for 12.93 € for the chip (of DDR4, maybe 3/2 is cheaper?)
<electronic_eel>
bvernoux: if you buy one, I'd be interested to see some high-res photos of the pcb and also s-parameters of the probes/coaxes, to get an idea what they are doing there
<bvernoux>
yes
<electronic_eel>
just to get some more ideas for CONWAY
<bvernoux>
I'm very tempted by the 16chan version ;)
<bvernoux>
I have already a DSLogic Pro
<bvernoux>
bought with the KS years ago
<bvernoux>
and it work not too bad
<Degi>
Hm on CONWAY how long do we plan the wires on the comparator part to be?
<bvernoux>
and my probes was bad they have changed the design after
<bvernoux>
like we see now
<electronic_eel>
does it have the same frontend circuit, with the small coaxes and so on?
<bvernoux>
yes with the small coax
<bvernoux>
they have a filter ;)
<bvernoux>
in the termination of each probe interesting
<electronic_eel>
ah, so they already had the coax, with the filter, in their first design, but it didn't work well for you?
<bvernoux>
I'm just interested by their fact 1GSPS ;)
<bvernoux>
in theory with niquist it should do 500MHz (in best case)
<bvernoux>
but I even doubt it can capture 250MHz Analog Signal
<bvernoux>
because of the probes/filter ;)
<bvernoux>
anyway they have capture of the signal
<bvernoux>
Dupont fly wire vs Shielded fly wire
<electronic_eel>
you say "they have changed the design after", what details did they change regarding the probes, that you can see from the datasheet/photos?
<bvernoux>
anyway Dupont fly wire are awfull and does not work with > 50MHz Analog signal ;)
<bvernoux>
electronic_eel, I speak about my first DSLogic Pro from KS which was provided with just dupont wires very thin and not good
<bvernoux>
electronic_eel, after they have changed to 1signal+ground per wire
<bvernoux>
provided in DSLogic Plus
<electronic_eel>
ah, ok, so I misunderstood you, I thought they already had the small coax in the first version
<bvernoux>
no it is new
<bvernoux>
their stream mode is interesting too
<bvernoux>
even if the best if BufferMode especiallu with 2GBit+RLE
<Degi>
Nah wait its the same speed, but has 2 channels
<bvernoux>
It will be fun to check their coax cables ;)
<bvernoux>
with the VNA
<bvernoux>
I'm pretty sure they do not exceed 200MHz
<bvernoux>
who know ;)
<bvernoux>
ok so let's buy it
<Degi>
Huh yeah
<Degi>
I have some coax like that too
<Degi>
It has dubious damping at 1 GHz
<bvernoux>
to have express shipping I will take additional probes ;)
<bvernoux>
the famous Shielded fly wires
<bvernoux>
they do not look like coax ;)
<Degi>
Huh?
<Degi>
Mine have some weird plastic inside that totally melts during sodlering...
<bvernoux>
I will test them like that ;)
<bvernoux>
torture test until death ;)
<bvernoux>
ok so bought DSLogic U3Pro16 + Shielded fly wires ;)
<Degi>
Stick them to a magnetron to measure 2.45 GHz dampening
<bvernoux>
to check in how many time they smell bad and burn in fire ;)
<bvernoux>
ok payment done ;)
<Degi>
Hmh kinda wanna build a 32 channel 1 GS/s adjustable offset LA for < 200 €... Like that should be possible. Otherwise you could just add ESD diodes onto an FPGA input lol
<bvernoux>
The challenge is to drive the IO @1GSPS
<bvernoux>
I do not know any cheap FPGA doing that
<bvernoux>
ha or using a trick with special IO for PCIE ...
<electronic_eel>
you'd need some SERDES input for that I think?
<bvernoux>
but they are limited in voltage
<bvernoux>
something like 1V ...
<electronic_eel>
yes, you need a comparator with a compatible output
<Degi>
You mean inputs? Like you can use ddr lol
<bvernoux>
and it is tricky to reach 1GSPS at the end
<bvernoux>
but interesting
<bvernoux>
especially at low cost
<Degi>
Like th 5G-8 graded ECP5 part I have here can do 1.6 Gbit/s maximally (though not sure if thats an input limitation, I was probably somewhat overclocking the fabric) and the cheapest one is 12? or 25 bucks
<bvernoux>
anyway I will do high resolution photo of U3Pro16 ;)
<bvernoux>
I'm pretty sure they have an ultra low cost BOM something like 50USD max
<Degi>
Yes it uses a x4 serdes as an input ofc
<Degi>
Yes a photo would be neat
<Degi>
I wonder if I can disassemble MSO5074 without breaking warranty...
<bvernoux>
and I will publish test to check the limit ;)
<bvernoux>
to see if it really do 1GSPS ;)
<bvernoux>
as it as a CLK IN
<Degi>
This 520 MHz counter can do 700+ MHz when driven with higher amplitudes...
<bvernoux>
so in theory if you provide a CLK + DATA it shall recover the 1GSPS
<electronic_eel>
bvernoux: and if you have an extra set of the wires, could you also try to dismantle their filter circuit and try to find out whats on it?
<bvernoux>
in worst case 1/2
<bvernoux>
yes I will dismantle the wires ;)
<electronic_eel>
cool. looking forward to your findings
<bvernoux>
before I will do full test with the VNA from 30KHz up to 2GHz (I can go up to 6GHz for fun too)
<electronic_eel>
hehe
<Degi>
Wonder if it can work with a 1 GHz clock...
<bvernoux>
the EXT Clk is interesting
<bvernoux>
and it have also EXT OUT
<bvernoux>
Trigger
<bvernoux>
Which does not exist on Saleae
<Degi>
Can it do pattern triggering?
<bvernoux>
No
<Degi>
What does the trigger do?
<bvernoux>
or in SW ;)
<bvernoux>
trigger are just basic IIRC
<bvernoux>
like Saleae
<bvernoux>
all is done in PC side
<bvernoux>
the demo do not show that
<Degi>
Hm I think scopes should have derivative limits on the trigger too
<Degi>
Like "when the voltage crosses 2.5 V at at least 1 V/ns" or something
<bvernoux>
hmm there is advanced Trigger
<bvernoux>
in fact in HW
<bvernoux>
I see the option but when I choose it in Demo it says it requires real HW
<bvernoux>
other interesting point is voltage is configurable by 0.1V step
<azonenberg>
bvernoux: artix7 -2 can do 1250 Mbps on DDR LVDS inputs
<Degi>
Huh weird
<bvernoux>
azonenberg, I doubt they have put an artix7 inside but interesting
<bvernoux>
maybe there will be a good surprise as I do not know what they could use as cheap FPGA for 1GSPS SA
<bvernoux>
Spartan6 does not support that ;)
<Degi>
Huh yeah artix 7 -2 is relatively cheap (at least compared to the LA...)
<Degi>
Dont a lot of FPGAs support 1 GS/s with a SERDES?
<azonenberg>
This is using the ISERDESE2, not a full gigabit transceiver
<azonenberg>
Doing oversampling with transceivers is considerably more work
<Degi>
Yes I mean something like that, not the thingies with clock recovery lol
<bvernoux>
for a 299USD LA the margin is thin ;)
<bvernoux>
I doubt they have a >100USD FPGA
<bvernoux>
Anyway in China they have better price than us
<Degi>
Like for 12 bucks you can get an FPGA which can do 1 gigabit on inputs
<electronic_eel>
azonenberg: nice data from you probes and the VNA
<electronic_eel>
you wrote that you suspect the pmk pins don't fit as well a you initially thought?
<azonenberg>
i know for a fact they dont fit super well. They wobble a little bit. I think they can be held in place with a tiny drop of loctite but still be removable if damaged
<azonenberg>
i may at some point try and find a better tip as well
<azonenberg>
tip socket*
<bvernoux>
azonenberg, do you have some capture with the VNA of latest probe ?
<electronic_eel>
ah, so the socket from mill max isn't tight enough?
<azonenberg>
electronic_eel: its tight but its a single point of support
<azonenberg>
the pin can rotate about that point until the tail hits the walls of the socket
<azonenberg>
so it moves several degrees
<electronic_eel>
ah, ok. do you know how the socket from pmk is built internally? how do they hold the tip?
<azonenberg>
No i don't. it's been hard to get a look inside since its attached to the tetris via a cable i cant remove
<electronic_eel>
hmm, a tip would be something you could maybe use loctite on, but the ground connectors need to be changed and rotated, and they use the same socket and have the same problem, right?
<azonenberg>
The grounds it doesnt matter as much IME
<azonenberg>
as they're under some sideways force typically which holds them in place
<azonenberg>
the tetris ground actually feels a little loose too
<bvernoux>
great
<azonenberg>
vs the tip which is solid
<electronic_eel>
could you put some spring on top that puts this sideways force also on the pin?
<azonenberg>
bvernoux: also it looks like the "gray" probe, which i think is using five resistors (have to check which values) is a little flatter in this testing
<azonenberg>
(good news is it should be fairly easy to stick a 0R across an extra pad if we need five instead of six - this also will allow us to easily make an AC coupled version)
<bvernoux>
it is quite stable up to 1.5GHz
<Degi>
Hm now is the version with 3 or 6 better?
<bvernoux>
and better than TA061
<bvernoux>
I speak about probe-centerpin-bladeground-into-coupler.s2p
<azonenberg>
Yes
<azonenberg>
Degi: i've done quite a few tests and am still not confident exactly what is "best" lol. I think when i get the final PCB i will do a couple of tests to compare
<Degi>
Yes play around with the resistors
<bvernoux>
6res-tinytip-rodground.s2p have 6dB less un S21
<bvernoux>
in
<bvernoux>
but it is very good
<azonenberg>
That is measured across a terminator
<bvernoux>
up to 2GHz
<azonenberg>
and includes non-de-embedded fixture effects
<bvernoux>
so the 6dB are normal
<azonenberg>
so it's going to be 6 dB lower, yes
<bvernoux>
in that case it is the best
<azonenberg>
i think measuring right into the SMA is better for characterization
<azonenberg>
this one looks really good too though
<azonenberg>
Flat +/- 0.5 dB of -14.5 dB S21 out to 1.2 GHz then starts rolling off to -3 dB @ 2.29 GHz
<bvernoux>
yes grey probe is good up to 2.5GHz
<azonenberg>
input impedance is a bit higher than i'd like, but not really worse than the pico
<azonenberg>
just hits minimum at a slightly higher frequency
<azonenberg>
the ta061 hits 22 ohms @ 950 MHz
<azonenberg>
i hit about the same but bottom out at 1.35 GHz
<azonenberg>
these are all reworked boards and i have hopes that the final revision will be better since it has a properly designed footprint for the tip
<bvernoux>
impedance on grey probe is very stable upto 4.5Ghz
<bvernoux>
2.18Ghz have Z=89 OHms
<azonenberg>
yeah but it dips lower before that
<azonenberg>
i think i can modify the tip geometry to have less capacitance
<bvernoux>
but else yes
<bvernoux>
Z=23Ohms @1.33GHz
<bvernoux>
we see 5 spikes on group Delay in ns
<bvernoux>
probably the connectors + resistors
<azonenberg>
Yeah but compare to group delay of the ta061
<azonenberg>
Anyway it's definitely better than the ta061 as is. My tentative plan is to sell the first round of 100 boards via some sort of crowdfunding campaign to recover the cost of the VNA, using whatever combination of resistors works best in my testing
<azonenberg>
Then probably do some fine tuning of the layout to try and flatten it out even more
<bvernoux>
yes it is very nice
<bvernoux>
it exceed the expected 1GHz BW
<azonenberg>
you expected 1 GHz?
<bvernoux>
up to 1.37Ghz it is very nice
<azonenberg>
i was targeting at least 2 since that's the limit of my scope
<bvernoux>
yes anyway it is good up to 2Ghz too
<azonenberg>
It's not bad, but not as flat as i want from 1.5-2
<azonenberg>
which is the real benchmark i aspire to be competitive with. And i'm nowhere close
<azonenberg>
that is a NICE PROBE
<bvernoux>
anyway clearly better than TA061
<azonenberg>
So we'll see how far i can push this design next rev
<azonenberg>
but i think this one is good enough to sell
<bvernoux>
hmm let check it
<bvernoux>
ha yes very nice
<bvernoux>
a very good 3GHz
<bvernoux>
the 921
<bvernoux>
I do not know what is defined in spec
<bvernoux>
but I see a clean 3GHz
<Degi>
Hm what will the price point be? I mean the TA061 costs like 350 $ and this one is already better
<azonenberg>
They sell it as a 6 GHz probe
<azonenberg>
Degi: i'm probably going to do $150 for the student edition (was targeting $125 but i think i'd lose money on that unless i can cut something or negotiate a better price for the PMK stuff), $250 basic commercial, $500 fully characterized
<azonenberg>
for the crowdfunding on the first 100 pcs at least
<azonenberg>
after that i'll respin the board and see where we can push it
<bvernoux>
at 3.5Ghz there is 10dB loss
<Degi>
Hm the TA061 doesnt say anything about characterization...
<bvernoux>
on the 921
<Degi>
So yeah this probe is pretty nic
<Degi>
e
<bvernoux>
they consider that like a 6GHz probe ;) ?
<azonenberg>
Degi: the ta061 doesnt even have an input impedance chart
<azonenberg>
bvernoux: they consider the ta061 1.5 GHz
<azonenberg>
the s2p says otherwise
<Degi>
Hmm maybe its fully molded and they got a bad plastic?
<bvernoux>
TA061 is more a 1.5Ghz Pribe than 921 a 6GHz probe ;)
<azonenberg>
lol
<bvernoux>
but yes 921 is very good
<bvernoux>
the best
<bvernoux>
I'm curious to check my 10Ghz probe ;)
<bvernoux>
12Ghz Probe ;)
<bvernoux>
Auburn Tech Model P-20B ;)
<bvernoux>
to test it in same condition as you
<bvernoux>
I really doubt it is better than 921
<bvernoux>
or even your probe
<Degi>
Huh that probe is even cheaper than the TA061 and claims 12 GHz...
<bvernoux>
yes ;)
<azonenberg>
Degi: i mean i'm sure S21 is better than -100 dB at 12 GHz... :p
<Degi>
lol
<bvernoux>
they have not published S2P anyway ;)
<Degi>
Hm it says that its for 50 ohm sources tho
<azonenberg>
they have "typical probe response" but i'm not sure how much i trust that
<Degi>
And its not down to DC...
<Degi>
100k to 12G
<azonenberg>
well its ac coupled
<bvernoux>
they just show that with Direct Barrel Grounding the response is ultra flat max +3dB ;)
<azonenberg>
and lol that probe is built out of a ballpoint pen
<azonenberg>
as far as i ca nsee
<bvernoux>
from 100Khz to 12GHz ;)
<Degi>
loll
<bvernoux>
yes I really doubt about that ;)
<bvernoux>
yes it is a ballpoint pen ;)
<Degi>
"Light RF Loading Characteristics"
<Degi>
What is its input Z???
<azonenberg>
The power rating they give is massive
<Degi>
"Low Distortion"
<azonenberg>
1W continuous, 25W for 1 sec
<Degi>
1 Watt yeah
<Degi>
25 W for 1 s is a bunch for SMDs
<azonenberg>
i suspect this is using a large carbon resistor or something like that
<azonenberg>
maybe the same Caddock resistor most of the other higher end probes use?
<bvernoux>
azonenberg, does it is fast to capture S2P with 4001pts ?
<bvernoux>
from Picoscope info PicoVNA can capture 10Kpoints in 1s IIRC
<azonenberg>
It's not that fast although vmware usb may be part of the limitation
<bvernoux>
about 10s ?
<azonenberg>
it's maybe 2-3 sec at 4001 points on all 4 channels
<bvernoux>
ha ok
<azonenberg>
per sweep
<bvernoux>
so it is good
<azonenberg>
1 sec i'd believe with only one channel active
<bvernoux>
I was checking PicoVNA forum it is not active and the support is not existing in Forum
<bvernoux>
with no answer during more than 4months ...
<Degi>
Hmm a VNA which scans all frequencies at once would be fun
<bvernoux>
it is strange they do not support better the SW part as the HW seems very good
<azonenberg>
the sw appears to be written in visual basic
<azonenberg>
the original vb, not vb.net
<bvernoux>
haha
<Degi>
lol
<bvernoux>
ha ok ;)
<azonenberg>
judging by the fact that they use mscomctl.ocx as their ui widgets
<bvernoux>
anyway the most important point is to retrieve the SxP data ;)
<bvernoux>
to display SxP file nanovna-saver-v0.3Dev is not too bad
<azonenberg>
i use sonnet
<azonenberg>
lol
<bvernoux>
as it is hard to find a good view for SxP
<bvernoux>
viewer
<bvernoux>
I use pocketVNA viewer for multi file to compare them
<bvernoux>
it is quite good for that
<bvernoux>
It is crazy that there is no any good SParam viewer ...
<bvernoux>
the worst are from R&S or other ;)
<bvernoux>
ultra old like VB6 stuff but even worse and ultra slow and big
<bvernoux>
what a shame
<Degi>
Maybe somebody can hack a python script with pyplot together
<bvernoux>
yes it is what nanova-saver do ;)
<bvernoux>
and the display is quite nice Python+Qt
<bvernoux>
working on Linux, Windows ...
<bvernoux>
it is pretty limited to compare multiple Sparams
<bvernoux>
921 Probe have a huge loss at 5.4GHz
<bvernoux>
more than 26dB
<bvernoux>
how they can say it is a 6Ghz Probe when 5.4Ghz is so lossy
<Degi>
Is that the "rated for 6 GHz" one?
<azonenberg>
Yes
<bvernoux>
yes
<Degi>
Maybe if its above 30 dB?
<bvernoux>
5.4Ghz is clearly awfull
<bvernoux>
in fact after 3.5GHz there is more than 20dB loss
<bvernoux>
I will say it is a very good 3GHz probe ;p
<azonenberg>
lol
<azonenberg>
yeah
<bvernoux>
I imagine the signal with the scope ;)
<bvernoux>
as it is a scope probe correct ?
<azonenberg>
yes
<bvernoux>
you have 2V input -26dB ;)
<bvernoux>
at 3.7Ghz there is -12dB
<bvernoux>
-16dB at 5.42Ghz ;)
<bvernoux>
(not -26 but -16)
<Degi>
Hm what is it for low frequencies? Should it be 0 dB?
<bvernoux>
it is -20dB
<bvernoux>
I hav computed the delta ;)
<azonenberg>
bvernoux: the 921 is -26 i think? because it's a 20:1 probe
<azonenberg>
mine and the ta061 are -20 (10:1) nominal
<bvernoux>
strange I see -20 by default
<Degi>
So when you say "at 3.7 Ghz there is -12dB" it has -32 dB relative to input?
<azonenberg>
bvernoux: oh did you not compensate for the 6 dB from probing across an open circuit?
<bvernoux>
-20dB quite flat up to 3GHz
<bvernoux>
then after it fall
<azonenberg>
the 921 s2p i gave you was open, not across a terminator
<bvernoux>
-32dB @ 3.73Ghz
<bvernoux>
so -12dB loss
<bvernoux>
vs ref
<bvernoux>
ha ok
<Degi>
Ah yes
<Degi>
Hmm I dunno, maybe they mistyped when they wrote 6 GHz?
<Degi>
-12 dB is pretty unusable if thats not even half spec
<bvernoux>
will be interesting to put a 50 Ohms
<bvernoux>
Yes I will admit ma permitted loss is -3 or max -6dB
<bvernoux>
mainly something better than 6dB loss
<Degi>
-6 dB is kinda pushing it but would still be usable near 6 GHz...
<Degi>
But -12 at a bit over halfway of 6 GHz is ooof
<bvernoux>
yes unusable with -12dB ;)
<bvernoux>
and the worst is at 5.4Ghz
<bvernoux>
-16dB (-36dB)
<Degi>
Oh it doesnt get so much worse after 3.7? No linear rolloff?
<Degi>
I mean -12 dB is 1/4 and -16 something near 1/10 or so... Your square wave is probs gonna look weird
<bvernoux>
Probe was not connected to scope or something ?
<bvernoux>
during the test ?
<bvernoux>
we see Z start at 523 Ohms and fall to 126Ohms at 2GHz
<azonenberg>
bvernoux: this is a full 2-port measurement with the probe on port 2, a SMA coupler on port 1
<bvernoux>
Test has been done in same conditions for the 4 probes ?
<azonenberg>
and the probe across the coupler
<azonenberg>
6res-tinytip-rodground is across a terminatino
<azonenberg>
all the rest are same conditions
<bvernoux>
ha ok
<bvernoux>
the port1 & 2 have big impact too
<bvernoux>
maybe by removing them (de-embedding) we could have only the probe effect
<bvernoux>
do you have measurement of the coupler ?
<bvernoux>
without probe
<azonenberg>
the coupler is basically perfect IIRC
<bvernoux>
do you have a photo to see it ?
<azonenberg>
i dont have a measurement handy but i recall it being super flat
<bvernoux>
as you have adone a TRL calibration ?
<azonenberg>
i'll grab one in a bit
<azonenberg>
I did SOLT with a traceable standard that actually came with error data for the non-idealities of the loads, opens, electrical length of the thru, etc
<bvernoux>
so in theory with a TRL everything is removed except the trace in center
<azonenberg>
Full 12 term calibration
<bvernoux>
ok so we see the trace effect+probe
<azonenberg>
yeah you're seeing just the probe and the coupler
<azonenberg>
and i think the coupler is negligible
<bvernoux>
yes it should be
<bvernoux>
so the tip is on the trace and where does the probe is connected on other side ?
<bvernoux>
a scope 50 Ohms input or other ?
<bvernoux>
as it is will be the most realistic
<azonenberg>
there's no trace
<azonenberg>
the SMA end of the probe is connected via the probe's supplied cable directly to the VNA
<bvernoux>
ha ok
<azonenberg>
the needle end of the probe is inserted into the center contact of the SMA coupler which is plugged into the other VNA port
<azonenberg>
with the ground on the shell of the coupler
<azonenberg>
it's as close to a "naked" measurement of the probe as i could get
<bvernoux>
yes
<bvernoux>
the 921 is very strange after 3Ghz it is why I ask
<bvernoux>
I do not understand how it is possible too loose 12dB
<bvernoux>
especially so fast it act like a low pass filter
<bvernoux>
between 3GHz to 3.7Ghz 12dB are lost
<bvernoux>
like there is a filter ...
<azonenberg>
Yeah i need to spend more time testing
<azonenberg>
It might have been bad contact somewhere that showed at higher freqs or something
<bvernoux>
it is why it is so strange for a 6GHz probe of that price
<azonenberg>
I am not yet 100% confident in my data
<bvernoux>
and I suspect you do not see that when probing with it up to 6GHz
<bvernoux>
to be tested with a SA ;)
<bvernoux>
as to find a 6GHz Scope is hard
<bvernoux>
even just a 4GHz scope
<bvernoux>
as I really doubt we see that effect between 3Ghz to 3.7Ghz with a loss of 12dB
<bvernoux>
with 2V in you have 0.5V on Scope ;)
<bvernoux>
signal is clearly destroyed ;)
<bvernoux>
I imagine the eye diagram
<bvernoux>
azonenberg, could you do the test on the TRL test board ?
<bvernoux>
to qualify the VNA ?
<bvernoux>
and also the SMA connectors ...
<bvernoux>
doing TRL calibration
<azonenberg>
I did. My footpritns have a fair bit of reflection on them
<bvernoux>
it is interesting to check the SMA
<azonenberg>
which is weird as i simulated them. Going to have a chat with sonnet support about it at some point
<bvernoux>
as I suspect they have some reflection and are not so good up to 6GHz
<bvernoux>
not flat
<bvernoux>
it is Ro4350B ?
<azonenberg>
yes
<bvernoux>
so it shall be a bit like OSHPark FR408
<bvernoux>
so such type of test should be done in anechoic chamber
<azonenberg>
this is input impedance on port 1 with no resistor
<azonenberg>
*2 ohms* at 1.6 GHz
<azonenberg>
where is that RF going??
<azonenberg>
best guess is it's being shunted to ground somehow
<azonenberg>
via capacitive coupling to the ground plane or something
<bvernoux>
yep
<bvernoux>
we see it clearly with nanvna saver ;)
<bvernoux>
the zoom are nice
<bvernoux>
to pop out any window
<azonenberg>
So the new probe has a big ground plane cutout under the tip
<azonenberg>
i hope this will improve the match there
<bvernoux>
at about 4.7cm ?
<bvernoux>
can be checked with TDR but accuracy is not very good
<bvernoux>
with actual points
<Degi>
Is that picture the S11?
<Degi>
Maybe some kinda resonances?
<azonenberg>
that's input impedance
<Degi>
hmh
<Degi>
How long is the metal part from the pad of the detached resistor to your measurement point?
<Degi>
Hm yes 4.7 cm
<azonenberg>
maybe a cm?
<Degi>
Oh
<Degi>
Weird
<miek>
can the pico do time domain analysis?
<azonenberg>
yes
<azonenberg>
although its a bit of a pain to set up and only works with power of two point sizes up to 4K
<Degi>
Hmh I mean if you can do frequency domain with phase, you should be able to fourier transform that?
<azonenberg>
That's exactly how it works
<Degi>
Hmh why is it complicated to setup... ugh software. Like they could have a "time domain" button lol
<azonenberg>
they do
<azonenberg>
the issue is, you have to have set it up with 4k points, done the cal there, etc
<Degi>
oh
<azonenberg>
any changes to these settings take LOTS of time
<Degi>
huh
<azonenberg>
and risk the software hanging
<Degi>
oh no
<azonenberg>
and the load-cal mode has caused hangs with 100% reliability in my experience
<azonenberg>
i have never been able to load a saved cal
<azonenberg>
so any time i close the software i need to re-cal it
<miek>
coming soon: vnahal :p
<Degi>
oof
<azonenberg>
Sometimes the firmware crashes and causes the signal generators to not work
<Degi>
lol
<azonenberg>
so you read noise on all channels
<azonenberg>
the only fix to this is to close the software, power down the vna
<azonenberg>
turn it back on again, and hope for the best
<Degi>
Somebody hasn't heard of fuzzers yet or was too lazy to implement them...
<azonenberg>
typically it takes 10+ reboots to solve this
<miek>
ouch
<azonenberg>
I will be getting rid of this thing as soon as the harmon VNA is ready, if it's 1/10 as good as this it will be less painful
<azonenberg>
but i need the pico for now because i need *a* vna and the xavna didnt cut it
<bvernoux>
azonenberg, you have all those issues with PicoVNA ?
<azonenberg>
Yes
<azonenberg>
it is literally the most painful piece of test equipment i've ever used
<azonenberg>
By a huge margin
<bvernoux>
It is a shame they have not improved /rewrite the APP since the launch of the PicoVNA 2 years ago
<azonenberg>
The software is written in visual basic 6, as far as i can tell
<azonenberg>
like windows 98 era tech
<bvernoux>
yes ...
<bvernoux>
I will say Picoscope SW look like also VB6 ;)
<azonenberg>
but given the firmware hangs, i dont even think open PC software will help
<azonenberg>
we'd need a new firmware too and at that point might as well just make a new VNA
<azonenberg>
Which is what darrell is doing
<bvernoux>
are you sure it is not related to VMWare ?
<azonenberg>
i wanted to buy his but couldnt wait any longer to get a vna
<bvernoux>
as USB is real crap on VM
<azonenberg>
I'm not sure, but vmware usb has been good in my experience
<azonenberg>
i do not have any native windows machines to test on
<bvernoux>
no news from Darell about his VNA
<bvernoux>
he was saying that it is possible to rent it at start
<bvernoux>
in 2020 ...
<bvernoux>
azonenberg, Do you have some feedback from PicoVNA support as they should provide C code using libusb
<bvernoux>
and everyone will be happy
<bvernoux>
at least a SDK with some API using libusb
<azonenberg>
AFAIK no c code exists
<azonenberg>
their documented API is an activex control
<azonenberg>
i wish i was kidding
<bvernoux>
I saw their Python is a crap
<bvernoux>
as it use COM components ...
<bvernoux>
so 100% Windoz
<azonenberg>
Yes
<azonenberg>
because the entire driver stack is writtne that way
<bvernoux>
like you said from Win98 Erra
<azonenberg>
beacuse that's how you talk to vb6
<azonenberg>
What i'm seeing from pico so far is that their software is utter garbage and their hardware is topnotch
<bvernoux>
they could provide at least just minimal API to drive it ...
<bvernoux>
and to calibrate it ;)
<bvernoux>
with my HP I do calibration on the screen as I have not coded it through SCPI ...
<bvernoux>
but after that I retrieve all S1P or S2P from my App else it is unusable ;)
<bvernoux>
The good points is documentation done in 1990 are very good ;)
<bvernoux>
and everything (or >95%) can be done through GPIB
<bvernoux>
it is crazy that PicoVNA does not have SCPI/GPIB to be driven by external SW
<bvernoux>
with standard commands
<azonenberg>
i suspect all pico stuff are "dumb" and most intelligence is pc-side
<azonenberg>
in their blbo
<azonenberg>
blob*
<azonenberg>
the cal, for example, is almost certainly software based postprocessing of raw sample data going over usb
<bvernoux>
yes clearly everything can be done on PC
<bvernoux>
using raw data
<bvernoux>
like that embedded SW is ultra basic
<bvernoux>
and HW can be cheaper but in that case PC SW shall be good enough and reliable ;)
<bvernoux>
anyway so far the good point is you can retrieve S2P files
<bvernoux>
the bad point is to do all the time the calibration ...
<bvernoux>
as recall with stable environment work very well on good VNA
<bvernoux>
even if it is recommended to do calibration from time to time I can tell you that even my default config is very nice for fast measurement without any cal especially for S1P
<bvernoux>
we are far from those cheap xaxavna which drift like crazy ;)
<bvernoux>
also my miniVNAPro2 has drift but not like xaVNA
<bvernoux>
with just a good VCTCXO (which cost 5 or 10USD) it could be really better I think