azonenberg_work has quit [Ping timeout: 260 seconds]
bvernoux has joined #scopehal
sorear has quit [Read error: Connection reset by peer]
elms has quit [Ping timeout: 264 seconds]
elms has joined #scopehal
sorear has joined #scopehal
sorear has quit [Ping timeout: 268 seconds]
<bvernoux>
azonenberg, hello do you have done other tests with Sonnet ?
<bvernoux>
I'm in contact with QucsStudio creator and he has integrated OpenEMS but so far the latest version supporting it is not released as he want more support with developers ...
<azonenberg>
bvernoux: i've done a lot, i use it for stackup design on almost all of my boards now
<bvernoux>
Ha great and what are the result simulation vs reality ?
<azonenberg>
found another minor bug and had a response back in like 2 hours that it was already a known issue with a fix scheduled for the next point release in a few weeks. There's a workaround but i had discovered that myself before sending in the ticket
<azonenberg>
let me put it this way, the results are good enough that i caught my fab making a mistake with impedance control
<bvernoux>
and your previous bug is fixed ?
<bvernoux>
ha nice
<azonenberg>
The previous bug is scheduled for the same point release. There's a good enough workaround that they didnt release an out of cycle patch
<bvernoux>
ha ok great
<azonenberg>
According to my support contact, they have a "five minute rule" they recommend to customers
<azonenberg>
If you can't solve your issue in five minutes, talk to them
<bvernoux>
Very nice support
<azonenberg>
yes i'm highly impressed
<bvernoux>
It seems Sonnet Method Of Moment is more accurate then FDTD method
<bvernoux>
But I would like to compare them ;)
<azonenberg>
I'm using sonnet right now on a customer's board with a cellular radio module
<bvernoux>
ha nice
<azonenberg>
For optimizing a ground plane cutout for return loss of the antenna connector as well as general stackup design
<bvernoux>
it is clearly a must have for things like that
<azonenberg>
it's just another tool in the toolbox now
sorear has joined #scopehal
<azonenberg>
i dont even close it anymore, it's open 24/7 since i use it so much lol
<azonenberg>
Also, re probes
<azonenberg>
my new scope (lecroy hdo9204) came in a few days ago
<azonenberg>
I did some more tests with my old Z0 probes
<bvernoux>
Do you have finished the probes with Sonnet simulation with some new improvements ?
<azonenberg>
as well as the stock probes
<azonenberg>
no screenshots but the zs1500 was the best in terms of flatness and loading (duh, its a $2000 active probe)
<azonenberg>
the stock 1M probes were the worst
<azonenberg>
the best variant my homebrew probes came in #2, but had a significant impedance mismatch i think at the mmcx connector. The new variant (at fab now, should be done soon) should fix that
<bvernoux>
ha great
<bvernoux>
Yes it seems major issues often comes from connectors too
<bvernoux>
I was not suspecting there is such effect on connectors
<azonenberg>
anyway, what's funny is
<azonenberg>
the pico ta061 was way better than the stock probes (duh)
<azonenberg>
But not nearly as good as i had hoped
<azonenberg>
(this is their 1.5 GHz Z0 probe, not the premium 5+ GHz picoconnect 900 series)
<azonenberg>
in particular it has a signficant amount of overshoot and ringing on the rising edge of my 40ps risetime test signal
<bvernoux>
I'm impatient to test the probes with my VNA ;)
<azonenberg>
i think while it's pretty flat to 1.5 GHz, it has lots of peaking beyond that
<bvernoux>
I have characterized lot of NFC Antenna and the results are amazing ;)
<azonenberg>
and there's no bw limiting filter on it
<azonenberg>
Which means that if you probe signals with fast risetimes, the >1.5 GHz frequency components will be attenuated much less than the lower freq
<azonenberg>
and you'll see a huge overshoot
<bvernoux>
very interesting
<azonenberg>
This is likely the result of capacitive coupling across the attenuator resistor. The exact same reason that i use flip chip resistors, and three of them in series, on my new design
<bvernoux>
So far I'm testing JLCPCB service too
<azonenberg>
for a low cost probe they probably didnt bother
<azonenberg>
it's better than the OEM probes for sure, and i think if i apply a 1 GHz bw limiting filter in postprocessing i can make them quite usable
<azonenberg>
but i think my probes will be better :p
<azonenberg>
the 3d printed shells for my prototypes come in on monday
<azonenberg>
already shipped
<bvernoux>
ha great
<bvernoux>
do you have picture of the case ?
<azonenberg>
New probe shells are one piece, not two. Let me grab a screenshot
<bvernoux>
but with 40GSPS it can do easily 4GHz up to 6GHz
<azonenberg>
the same series goes faster
sorear has quit [Ping timeout: 260 seconds]
<azonenberg>
this is the lower cost edition within the family
<azonenberg>
it's a rental fleet unit that had no software option upgrades, so wasn't getting a lot of people renting it
<azonenberg>
so they sold it off
<azonenberg>
I got it for 15 kUSD
<bvernoux>
ha very nice
<bvernoux>
for a 40GSPS 10bits it is clearly a very good price
<azonenberg>
mid 2017 manufacture date, less than 3 yrs old
<bvernoux>
with warranty ?
<azonenberg>
manufacturer warranty has a few months left, the rental company has their own 1-year warranty on anything they sell from the fleet
<bvernoux>
ha very nice
<azonenberg>
I'm selling my old 350 MHz wavesurfer 3034 to $dayjob to cover part of that cost, and keeping the waverunner 8104
<azonenberg>
So now i'll have a 1 GHz 10/20 Gsps and a 2 GHz 20/40 Gsps scope
<bvernoux>
Especially 10bits for the 2nd one ;)
<azonenberg>
From what i hear the 10 bits is actually a bit of a hack. The other HDOs use true 12 bit ADCs
<bvernoux>
ha yes maybe but it is always better than 8bits which are often 7bits ENOB or less ;)
<azonenberg>
it sounds like the HDO9000 series is doing some kind of DSP to recover more bits of resolution than the ADC physically has? unclear on details but from what i hear it uses a lot of CPU and doesn't buy much more performance than 8 bit
<bvernoux>
12bits is clearly a game changer for some stuff but so expensive for > 2GHz ...
<azonenberg>
which is one of the reasons it was discontinued (but existing units are still supported for quite a few years)
<azonenberg>
luckily you can switch the HD mode off and go to 8-bit operation too :p
<azonenberg>
at which point it effectively turns into a waverunner 9000 series
<azonenberg>
i dont have the rack kit for it yet, but i stuck it on my bench and did some preliminary testing plugged into a random ethernet port from the next lab bench. Libscopehal, unsurprisingly, worked the first time without any code changes
<azonenberg>
also at some point soon i plan to add support for "unit conversion" filters and non-voltage/frequency graphs to scopehal
<bvernoux>
ha great
<azonenberg>
for example, there will be a filter for "current shunt" that, given the value of a resistor and a trace of voltage across it, gives you a current waveform in amps
<azonenberg>
(this same API will also eventually support current probes, once i get a CP030/CP031 to test with)
<azonenberg>
Another thing on the todo is trending of parameters over time
<azonenberg>
so i can plot, say, frequency of a clock from one waveform to another
<azonenberg>
eventually i'll add support for measurements to have both instantaneous and whole-waveform outputs
<azonenberg>
so you can measure cycle by cycle frequency of a single capture and plot that, to look at a spread spectrum clock or something
<bvernoux>
Look their video I have never see this feature where channel are independant like that thanks to their tricks in HW
<bvernoux>
independent time and frequency controls
<azonenberg>
I dont like the whole concept of adjusting time/div in general. I find it kinda annoying
<bvernoux>
check the video you will see
<azonenberg>
Which is why scopehal hides all of that from you. i care about sample rate and record length
<azonenberg>
(i'm watching)
<bvernoux>
the must is you can see the Spectrum related to slow SPI for example
<azonenberg>
in particular, glscopeclient lets you arbitrarily zoom waveforms independently of the record length (so you can have captures much longer than what fits in the screen)
<bvernoux>
and Tek are the only one using Linux (Closed) on their high end when others use Windoz ;)
<azonenberg>
if somebody has a high end tek i'll gladly help add support for it :p
<azonenberg>
what i am more impressed by looking at that video is the 8 analog channels... although lecroy's waverunner 8000HD series has that too
<azonenberg>
I really want one of those. Saw a 1 GHz 8ch 10 Gsps one on ebay for about $25K
<azonenberg>
Which i didn't have, especially since i had just bought the hdo9204 :p
<azonenberg>
and given the choice of only one upgrade i needed the 2 GHz more than four extra 1 GHz channels
<bvernoux>
yes
<azonenberg>
anyway, re the video basically i think they're just working around the awful hack that is time/div
<bvernoux>
me too I prefer >=2GHz with less chan than 1GHz on 8chan
<bvernoux>
for my usage
<azonenberg>
pretending to be an analog oscilloscope
<azonenberg>
The only thing that is cool is that they can vary sample rate per channel
<azonenberg>
which scopehal supports, but very few scopes do
<azonenberg>
what would be nice is if you could see a low speed time domain capture of the slow rs232 and a frequency domain plot *of the same channel* with a much higher sample rate to see risetime harmonics etc
<azonenberg>
so basically streaming fft of full bandwidth data, plus decimated data going to the time domain plot
<bvernoux>
yes will be amazing
<bvernoux>
with FFT ...
<bvernoux>
GPU FFT ;)
<azonenberg>
When i build a scope fft will be in the FPGA at full bandwidth
<azonenberg>
and independent of any decimation done for deep-memory captures
<azonenberg>
The adcs will also be fed by separate pll outputs
<azonenberg>
so if i have say 1M points per channel of memory, i'll be able to have 1M points at 10 Gsps on one channel but 1M points at 500 Msps on another
<azonenberg>
phase locked 20:1
<azonenberg>
And you'll have individual horizontal offset control of each
<azonenberg>
so i could have 1M points at 500 Msps, with 1M points at 10 Gsps on another channel arbitrarily positioned within that window
<azonenberg>
All pretense of pretending to be an analog scope will be dead and gone
<azonenberg>
there will be no concept of scale divisions, you'll control record length and sample rate only
<bvernoux>
ha yes very nice
<bvernoux>
cool => sma-probetarget_oshpark_bve_v0_3_01Jan2020 shipped from OSHPark ;)
<bvernoux>
it shall be exactly 50ohms impedance ;)
<azonenberg>
:)
<bvernoux>
based on your probe v0.1
<bvernoux>
but this one is CGPW
<azonenberg>
Nice. If it works out well i want to make one
<azonenberg>
What SMA are you using?
<bvernoux>
the cheap one ;)
<bvernoux>
But I'm planning testing with more expensive ones too for an other design