<Degi>
Hm glscopeclient seems to connect to the rigol sometimes but makes warnings of unknown trigger source, socket read/write failed etc
<Degi>
Ill figure that out tomorrow
futarisIRCcloud has quit [Quit: Connection closed for inactivity]
futarisIRCcloud has joined #scopehal
Degi has quit [Ping timeout: 265 seconds]
Degi has joined #scopehal
juli963 has joined #scopehal
<azonenberg>
Ooook so let's see... Probes are still a month out, probe dummy is held up in Chinese customs. We have the AFE+ADC talking to scopehal, albeit with a very stripped down, simplified controller
<azonenberg>
We're now up to 8 issues found on the AFE board
<azonenberg>
The TCP stack still needs a fair bit of work
<azonenberg>
The SMA test board is at oshpark
<azonenberg>
BLONDEL will probably still use the cheap SMA because it's good enough for 100 MHz, but starting around ZENNECK I'm thinking of moving to the nicer one
<azonenberg>
At some point i want to do full VNA measurements of the AFE
<azonenberg>
But i want to wait until we've finished the testing and rework
<azonenberg>
Sooo i guess one big thing we've kinda neglected is the active probe subsystem
<_whitenotifier-9>
[starshipraider] azonenberg pushed 1 commit to master [+1/-0/±1] https://git.io/Jfsyq
<_whitenotifier-9>
[starshipraider] azonenberg 5462572 - Updated design notes with current bug list
<azonenberg>
oh and we also haven't yet tested the input stage overload protection on the AFE yet
<azonenberg>
So I made a full model of the probe, simplified a bit (e.g. via arrays merged and the grounding simplified) in sonnet
<azonenberg>
with the medium accuracy meshing option it needs ~550 MB of ram, with high accuracy it needs 3.7 GB
<azonenberg>
the diagonals on the tapers are the big thing that bumps the cell count through the roof
<azonenberg>
Trying to split it up and see if i can model at least part of it ok-ish
<azonenberg>
Right now it looks like we have a sizeable impedance mismatch in the 2-3 GHz range with return loss up to about -16 dB then staying pretty high for a while after that
<azonenberg>
This causes about a dB of loss at 2 GHz
<azonenberg>
of insertion loss*
<azonenberg>
just at the SMA transition
futarisIRCcloud has quit [Quit: Connection closed for inactivity]
bvernoux has joined #scopehal
<bvernoux>
hello
<bvernoux>
I have just received the DSLogic U3Pro16
<bvernoux>
I can say the USB3.0 is fully exploited after a fast test of 10s @1GSMPS on 8chan ;)
<bvernoux>
I see with processexplorer that retrieving data is done at an average of more than 400MBytes/s
<azonenberg>
:)
<azonenberg>
nice
<bvernoux>
yes very smooth
<bvernoux>
and supporting RLE
<bvernoux>
I can capture use 8chan @1GSPS with Buffer Mode+RLE with duration up to 17.18s
<bvernoux>
so let's look inside ;)
<bvernoux>
I confirm the "probes" have some resistor soldered ;)
<bvernoux>
I will cut one spare to see exactly what they have do to adapt impedance...
futarisIRCcloud has joined #scopehal
<azonenberg>
Nice
<bvernoux>
But the cable are really thin I need to build a test fixture to check them with my VNA
<bvernoux>
I'm pretty sure they are bad and do not exceed 200MHz
<bvernoux>
I'm not sure Agilent or others do better anyway (especially the ones provide with Scope)
<azonenberg>
Yes. Stock LA pods generally are not super fast
<azonenberg>
For a future project i want to build a LA that can go out to the GHz range
<azonenberg>
With SMA inputs
<bvernoux>
ha yes
<azonenberg>
or perhaps MMCX if we want to have a lot of them in a dense area
<bvernoux>
with your flex PCB probes ;)
<bvernoux>
Also a fun stuff I have received a cheap DC-4.0GHz Fixed Attenuator
<bvernoux>
I have caracterized it up to 6GHz
<bvernoux>
and it is really not bad for something which cost <10USD ;)
<bvernoux>
it even reach about 5GHz
<azonenberg>
bvernoux: I plan to respin the flex pcb probe for higher bandwidth at some point
<azonenberg>
designation will be AKL-PT2
<bvernoux>
except the 20dB Attenuator which does not respect the spec +/-1.1dB ;)
<bvernoux>
I will probably do a web page later ...
<electronic_eel>
bvernoux: thanks for the pictures
<electronic_eel>
it looks like they are going nearly directly into the fpga and are not using extra comparators
<electronic_eel>
there are just some resistors, probably tvs diode arrays (U5, U6, U13, U19, U20) and thats it
<electronic_eel>
such fpgas usually have differential inputs that can be used as comparators, but they often don't have the same performance as dedicated ones
<electronic_eel>
I also don't see any array of capacitors for all the inputs, to build a rc divider
<electronic_eel>
so either that is all in the circuit at the front of the probe, or they are using a different architecture
<azonenberg>
it claims adjustable threshold in 100 mV steps
<azonenberg>
inputs look to be single ended juding by the resistors
<azonenberg>
so yeah i bet they're using lvds inputs as comparators
<azonenberg>
its a good hack for cheap hardware but not the kind of performance i want to reach
<electronic_eel>
maybe the 100 mV steps are not the limit of the dac, but the accuracy limit of the lvds inputs
<azonenberg>
lol
<azonenberg>
that is hilarious but plausible
<Degi>
Huhh
<bvernoux>
re
<bvernoux>
hmm this famous XC6SLX16 seems to have GTP transceiver
<bvernoux>
as I do not see how they can do 1GSPS without Gigabit Transceiver
<bvernoux>
up to 3.2Gb/s
<bvernoux>
ha strange the LX16 does not have GTP
<azonenberg>
bvernoux: the ftg256 package never has gtps
<azonenberg>
also xc6slx does not have them
<azonenberg>
xc6slx*T does
<bvernoux>
yes it is why I do not understand how they can do 1GSPS
<bvernoux>
I plan to do test with different signal to check that
<bvernoux>
the must have will be SPI but there is no SPI working at >100MHz ;)
<azonenberg>
DS162 says a -2 spartan6 can do 950 Mbps on DDR LVDS inputs using ISERDES2
<azonenberg>
-3 can do 1080
<azonenberg>
page 18
<bvernoux>
ha ok so it is probably the trick
<bvernoux>
it is a 2C ;)
<azonenberg>
So they are probably running slightly over spec and/or only sampling at 950 Msps
<lain>
950, 1000, close enough :P
<bvernoux>
they are using Micron DDR3
<azonenberg>
lain: to be fair if you're oversampling timing is less critical
<lain>
azonenberg: true
<azonenberg>
and if its a product meant to be used at room temp and where an occasional bitflip isnt a huge deal
<azonenberg>
aka budget equipment not for serious users
<azonenberg>
i mean rigol OC'd their adcs for a long time iirc
<azonenberg>
i wouldnt put my name on a product that did that, but it's been done
<lain>
I'd do it under two conditions
<lain>
1/ every OC'd device is binned according to performance testing under the full specified temperature range
<lain>
2/ it provides enough cost saving to justify 1
<monochroma>
some companies get explicit information/permission from the part vendor that they can safely run a part past the datasheet spec (basically what the /REAL/ limit is), but who knows :3
<lain>
yeah
<lain>
at an old job, one of our customers was making a board that needed a part spec'd to run up to like 150C
<lain>
the closest thing available that wasn't also, like, rad-hard, aerospace, military, etc etc, was a TI part spec'd to something very close, I think 130C
<lain>
(and all those aerospace crazy parts were like 100x more expensive)
<azonenberg>
So you tested them all at 150 to see what happened?
<azonenberg>
or contacted ti to ask?
<lain>
yeah they setup a fixture to test huge batches of them at once and they had I think 100% success rate even after extended testing
<lain>
so they went about their business doing that
<lain>
until one day one of /their/ customers was like "we can't buy your product because it uses this one chip outside of the specified range"
<lain>
more specifically, the customer required certification of every part inside their device to work at the full range
<azonenberg>
was the customer DoD or an agency thereof?
<lain>
they went to TI and said "here is YEARS of testing on this chip, we know for sure it works at 150C, please update your spec so we can sell to this customer"
<azonenberg>
most folks wouldn't do that level of due diligence on a product lol
<lain>
azonenberg: probably, or nuclear power
<lain>
TI refused though :<
<lain>
they tried several times but TI wasn't budging
<bvernoux>
lain yes TI like ti keep things and do not move ;)
<bvernoux>
I hate they politic towards silicon bugs it is the same they never fix !!!
<bvernoux>
I have a very bad example with trf7970a
<bvernoux>
it does not respect the timing for ISO
<bvernoux>
so you cannot use it in product certified for NFC ...
<bvernoux>
and they NEVER fixed that
<azonenberg>
lol
<bvernoux>
what a shame
<bvernoux>
it is in an errata which specify they are not compliant ...
electronic_eel has quit [Ping timeout: 258 seconds]
<azonenberg>
lool
electronic_eel has joined #scopehal
<azonenberg>
reminds me of errata on a couple of chips "feature X is nonfunctional"
<azonenberg>
"wontfix"
<lain>
there's an errata for an early revision of the atmel avr32 uc3a3 series where multiply is not functional
<lain>
none of the multiply instructions work
<azonenberg>
yeah but they fixed it right?
<lain>
yeah
<azonenberg>
they didnt simply update the ISA spec to say there's no mutliply instruction
<lain>
yeah lol
<lain>
but if that got to tapeout with an entire instruction non-functional, what the hell lol
<bvernoux>
I have found it
<bvernoux>
The bit duration(1 etu) as definedby ISO14443A/NFC-Aair interfaceis supposedto be9.44μs (±0.5%)
<bvernoux>
it is 9.57us ;)
<bvernoux>
The device fails this criterion.Measuring this value(for example,for EMVCOcompliancetesting),the 1 etu is approximately≥9.587μs and,hence,outsidethe givenspecificationlimits.
<bvernoux>
what a joke
<bvernoux>
so it cannot be EMVCO compliant !!
<bvernoux>
in clear it cannot be used for any payment and more ...
<bvernoux>
I do not speak of all other crazy bugs ;)
<bvernoux>
even if the features of this chipset was amazing ;)
electronic_eel has quit [Ping timeout: 260 seconds]
electronic_eel has joined #scopehal
juli964 has joined #scopehal
juli963 has quit [Ping timeout: 265 seconds]
<bvernoux>
so let's test different clock to test this USB3 Analyzer ;)
bvernoux has quit [Read error: Connection reset by peer]
bvernoux has joined #scopehal
<bvernoux>
re
<bvernoux>
test with 250MHz clock signal on DSLogic U3Pro16
<bvernoux>
it recover perfectly the frequency
<electronic_eel>
so they are probably slightly overclocking the transceiver, I see no problem with that in itself
<electronic_eel>
the question is more how accurate the "comparators" are
<Degi>
Re OC: For the ECP5 there was a document somewhere that showed that it can reach way faster data rates
<electronic_eel>
do you have an arb gen where you could try that with a slower signal?
<bvernoux>
i'm checking the threshold
<electronic_eel>
like forming a slower signal so that it goes near the defined threshold, but not over? or just over the threshold for a short time
<bvernoux>
and it is tricky
<bvernoux>
but there is clearly an effect but I really doubt it is 0.1v step ;)
<Degi>
Hm maybe just a rc cicuit if you dont have a awg...
<Degi>
And connect all channels to the voltage, that way you can see differences
<bvernoux>
I have an AWG ;)
<bvernoux>
but it is more for RF stuff ;)
<bvernoux>
so output level is pretty limited
<Degi>
huh RF awg?
<bvernoux>
yes
<electronic_eel>
integrated in an rf gen?
<bvernoux>
it is more a Signal Generator in fact ;)
<bvernoux>
Agilent E4432B
<bvernoux>
it can do also basic stuff like square wave ;)
<electronic_eel>
yeah, that is no arb gen, but a rf gen
<electronic_eel>
nice rf gen
<bvernoux>
yep ;)
<bvernoux>
I have also ERA SynthMicro but it is more for RF
<bvernoux>
and we see the load of the probe have big impact on the ouput
<electronic_eel>
hmm, maybe you can hack something together with some 74 logic at different vcc levels?
<bvernoux>
output
<bvernoux>
as I connect my scope + probe to check the effect
<bvernoux>
as it is 50Ohm output too ;)
<bvernoux>
it is a bit a mess
<bvernoux>
especially for high freq >100MHz
<bvernoux>
other example is USB HS
<bvernoux>
but that will kill the signal ;)
<bvernoux>
just when i connect the probe
<electronic_eel>
yeah, sure
<electronic_eel>
what did they spec, 13pf or similar?
<electronic_eel>
that will instantly kill usb hs
<bvernoux>
yes 13pF / 250KOhms
<bvernoux>
I'm testing with my GPSDO ;)
<bvernoux>
mini GPS from LeoBodnar
<bvernoux>
but same effect as it is 50Ohms outputs the probe attenuate a lot the signal when exceding 100MHz
<electronic_eel>
hmm, the E4432B goes down to 250kHz and also has a wide range of output power. can't you use that to create a sine signal into 50ohms and probe it?
<bvernoux>
yes I can ;)
<electronic_eel>
regulating the output power should give you a way to test the threshold
<bvernoux>
it is just overkill for that ;)
<bvernoux>
and it is 50Ohms output
<bvernoux>
for RF stuff
<bvernoux>
except the LF Out
<bvernoux>
which is limited
<electronic_eel>
overkill of course, but it is otherwise just getting lonely on your bench
<bvernoux>
I will test on real signals on HydraBus anyway it is limited to 42MHz max ;)
<electronic_eel>
but can you set arbitrary output levels on hydrabus?
<bvernoux>
I need to buy an AWG ;)
<bvernoux>
no I cannot set output level at high freq
<bvernoux>
the DAC is limited to 1MSPS ;)
<bvernoux>
else it is pure IO with PWM to go up to 42MHz but we can change only duty cycle and it is 3.3V out ;)
<electronic_eel>
have a look at the siglent SDG6000x series
<bvernoux>
could be convenient as so far I'm doing mainly RF stuff
<bvernoux>
or digital ;)
<electronic_eel>
I have used it for some analog stuff, but also for adding programmed noise to digital circuits to test the limits
<bvernoux>
yes it is nice for that
<bvernoux>
as so far I'm pretty limited for such type of test
<bvernoux>
especially to inject noise on digital side
<bvernoux>
but to test a LA it is even more complex as it requires multiple chan ;)
<bvernoux>
if we want to test it correctly it will requires 16 chan with pattern where we can change the level ;)
<bvernoux>
to test trigger
<bvernoux>
a FPGA could do the trick to generate multiple signal up to 500MSPS
<bvernoux>
but not to test the trigger
<bvernoux>
-trigger +thresholds
<bvernoux>
I confirm also the thresholds with 0.1V increment is for all channels
<electronic_eel>
yeah, there would have been a block of several dacs on the pcb otherwise
<electronic_eel>
or one bigger one with several outputs
<bvernoux>
there is sig gen on the PCB ;)
<electronic_eel>
hmm, which part are you talking about?
<bvernoux>
it is a freq synthetizer ADF4360-7
<electronic_eel>
yeah, the vco
<bvernoux>
it is probably to have an accurate sampling rate
<electronic_eel>
they probably use it for proper clocking
<bvernoux>
it is configurable from 380MHz to 1800MHz
<bvernoux>
350Mhz
<electronic_eel>
to generate the comparator voltages they could also use pwm from the fpga, but they should have some opamps to do that properly
<electronic_eel>
and I don't see them
<electronic_eel>
them = a set of opamps for all the channels
<bvernoux>
do you have checked carefully all photo ?
<bvernoux>
it is clear there is not lot of things ...
<electronic_eel>
maybe you just forgot to publish the pictures of the second pcb ;)
<bvernoux>
there is only 1 PCB
<bvernoux>
;)
<bvernoux>
I was searching too ;)
<bvernoux>
where is the other one
<bvernoux>
I'm searching what is this SOT23-6 1201 marking
<bvernoux>
I suspect it is related to the threshold ;)
<bvernoux>
it is clearly something related to voltage as there is inductor+capacitor
<electronic_eel>
how about U1? couldn't that be a dac?
<electronic_eel>
something with an inductor is probably a dc/dc converter, too much power for just setting the threshold
<electronic_eel>
a dac also needs some digital input (spi or i2c), so some more pins than sot23-6
<bvernoux>
the famous 1200 9AB A17X
<bvernoux>
ha yes that could match
<bvernoux>
with 5 blocks
<bvernoux>
as there is 5 blocks of probes
<bvernoux>
it do not take into account IOC
<bvernoux>
OIC
<bvernoux>
it is special for ext clk ...
<bvernoux>
bye
bvernoux has quit [Quit: Leaving]
<Degi>
You can charge a capacitor through a resistor to make a ramp...