<azonenberg>
Looks good but feels a bit flimsy in spots. Definitely will want to thicken up the walls
<azonenberg>
I think i can improve the LCD mounting, it looks like there's a spot to put a screw on
<_whitenotifier-f>
[starshipraider] azonenberg pushed 1 commit to master [+0/-0/±2] https://git.io/JfH9l
<_whitenotifier-f>
[starshipraider] azonenberg bf81392 - Final schematic for sfp-to-sma test fixture
Degi has quit [Ping timeout: 256 seconds]
Degi has joined #scopehal
<_whitenotifier-f>
[starshipraider] azonenberg pushed 2 commits to master [+2/-0/±3] https://git.io/JfH7X
<_whitenotifier-f>
[starshipraider] azonenberg 337123b - Routed most of sfp-to-sma
<_whitenotifier-f>
[starshipraider] azonenberg 75473ad - Added oshpark probe tip simulation
_whitelogger has joined #scopehal
<azonenberg>
Still debating whether i'm going to want to make the full metal enclosure or not. I'm probably going to hold off on a decision there until the MEAD boards come in next weekend, and i can see how hot they run
<azonenberg>
and play with my new EMC probes and see how leaky they are :p
<azonenberg>
Also just got an email from kickstarter, things have cleared on their end and they're sending me payment from the backers shortly. Once that comes in I'll be doing the PMK order to get all of the tips, grounds, etc
<monochroma>
:D
<azonenberg>
Next probe pcb rev is still a few days out as well, so we'll see how it works
<azonenberg>
Then the SFP+ test fixture is coming along nicely and will hopefully be done in another hour or two
<azonenberg>
then i can get back to MAXWELL schematic symbol creation
<azonenberg>
and start thinking about approximate bank assignments etc
<_whitenotifier-f>
[starshipraider] azonenberg pushed 1 commit to master [+0/-0/±2] https://git.io/JfHFP
<_whitenotifier-f>
[starshipraider] azonenberg 9466ecc - Continued layout on sfp-to-sma board. Fully routed, need some plane tweaks and stitching vias.
<azonenberg>
ready for final signoff review, i think
<azonenberg>
maybe some tweaks to the power planes
<monochroma>
sniff all the high speed busses! :D
<monochroma>
azonenberg: what do you think needs to be done to the planes?
<azonenberg>
monochroma: just rearranging a bit of layer 3 so i had solid ground under the back side diffpairs
<azonenberg>
done now
<azonenberg>
Ready for sneef sneef
<monochroma>
yay
<_whitenotifier-f>
[starshipraider] azonenberg pushed 1 commit to master [+0/-0/±1] https://git.io/JfHNL
<_whitenotifier-f>
[starshipraider] azonenberg 1543647 - Final version of sfp-to-sma, ready for layout review
<azonenberg>
If you wanna have a look at the board before i send it out
<Degi>
Oh neat
<Degi>
Is that for 10G?
<azonenberg>
Degi: in theory it should be able to do 10G, yes. Although i dont have a scope fast enough to see that
<azonenberg>
monochroma's scope is 5 GHz bandwidth so just barely too slow
<azonenberg>
however it's not a brick wall frequency resopnse
<Degi>
5 GHz is suffiicient
<Degi>
Heh yeah
<azonenberg>
10G is 10.3125 Gbps
<Degi>
If it was brick wall then maybe not
<azonenberg>
if it can still read 5.15625 GHz without too much attenuation
<azonenberg>
then we might be able to pull it off
<azonenberg>
i intend to try
<Degi>
(I mean like my rigol can show 250 MHz just fine, with a few dB damping despite having 70 MHz BW)
<azonenberg>
she wants to use it for debugging some fpga stuff, i personally plan to use it to develop a 1000base-X protocol decoder
<Degi>
Like thats 3% more
<azonenberg>
and possibly 10Gbase-R and 64b66b eventually
<azonenberg>
Yeah. I fully expect it to work
<azonenberg>
it's out of spec but lecroy gear is usually pretty good about being slightly faster than advertised
<azonenberg>
they dont cut corners
<Degi>
Inb4 it turns out that there's some 100th order filter or so in there
<azonenberg>
no i've seen actual cal data off my scopes
<azonenberg>
they're a few percent above the advertised bw cap on both channels
<azonenberg>
its a decently steep rolloff after that but 150 MHz past 5 GHz i think will work
<Degi>
For the -3dB?
<azonenberg>
Yes
<azonenberg>
the cal report includes actual -3 dB bw for each channel
<azonenberg>
my VNA also can be used as a sinewave signal generator, no modulation support so won't replace a proper vector siggen
<azonenberg>
but can be used to test freq response of a scope after a fashion
<Degi>
My rigol has rather slow rolloff, dont really remember how much, maybe 18 dB per decade or so
<azonenberg>
i intend to characterize her scope to find the actual response curve
<azonenberg>
Degi: anyway if you have a bit of time to do a layout review before i send it out, i wouldnt mind a second set of eyes
<Degi>
Hm yeah ill git clone it
<Degi>
Ahh its a thingie which connects SMA 1 to 2 and splits out the signals on the SMA
<Degi>
Those splitters look neat for paralleling HMCADs
<Degi>
Maybe a tad high BW
<azonenberg>
for $2 each is it worth going lower end?
<Degi>
Oh
<azonenberg>
Lol
<Degi>
Ok thats p cheap
<azonenberg>
Exactly
<Degi>
And its a passive. With active elements that makes more sense due to noise.
<azonenberg>
But yes for ADC paralleling this is a good choice
<azonenberg>
i actually picked it out originally for FREESAMPLE
<azonenberg>
the very first thing the signal path hits on FREESAMPLE is one of these, wtih one port going to the trigger circuit and one to the "ADC"
<Degi>
Huh those splitters are quite tight underneath
* monochroma
looks it over rl quick
<azonenberg>
it passes the oshpark drc
<azonenberg>
let me know when you two have finished looking at it and i'll run through my layout review checklist
<Degi>
I mean I hope it passes the solder paste "drc" heh
<azonenberg>
They come in a range of sizes, this is one of the smaller ones
<azonenberg>
lower power handling capacity but a non-issue for low voltage like this
<Degi>
Like that time we had a diode shorted from solder...
<azonenberg>
i think when i get a proper stencil printer that holds tension on the stencil i will get much better prints
<azonenberg>
i've been eyeing one but need to figure out how to free up some bench space to set it up before i order
<Degi>
Tension on the stencil? As in stretchingi t lightly?
<azonenberg>
yes
<Degi>
Because otherwise it kinda bends and moves?
<azonenberg>
so it cant flex and move around
<azonenberg>
i dont get good gasketing
<Degi>
I mean theres stencils with big metal frames too, at least at JLC
<azonenberg>
it's a lot better with steel than kapton but there's still room to improve
<azonenberg>
there is one stencil printer i've found that works with frameless stencils
<Degi>
Huh, Kapton stencils sound weird
<azonenberg>
much cheaper and more storable
<azonenberg>
laser cut kapton is the poor man's stencil
<Degi>
Hm yeah
<azonenberg>
they're ultra cheap and better than using a syringe to put down dots of paste
<azonenberg>
but vastly inferior to steel
<Degi>
I mean theyre only 7 bucks non-framed at jlc anyway
<azonenberg>
only recently did steel become affordable to the average hobbyist for one-off boards
<Degi>
And with frame theyre 9.43 $ and its like 1 kg or so oof
<azonenberg>
oshstencils is a bit more than jlc but its also a us-based shop so quick turn
<azonenberg>
and yes, that is the big issue. frames are big and heavy
<Degi>
Tbh not sure if jlc or oshpark ships to me faster heh
<azonenberg>
imagine dozens of them on a shelf somewhere
<azonenberg>
my lab is tight enough on space already
<Degi>
Yeh like foils are easier to stack
<azonenberg>
i store my small stencils in a baseball card organizer binder
<azonenberg>
lol
<Degi>
Is the rx pins on the SFP where the power comes out?
<Degi>
The diff pairs are routed kinda near to cutouts on In2.Cu
<Degi>
And how much current and precision do they need on the 3V3 rail? I guess not so much?
<azonenberg>
rx pins are outbound from the optic, yes
<azonenberg>
max current is i think ~300 mA total between tx and rx?
<azonenberg>
i have a 2A fuse on the input and will be feeding it from a regulated bench supply
<Degi>
So in1 is GND, in2 is 3v3...
<Degi>
Hm yeah
<azonenberg>
in2 is a mix of 3v3 and ground
<azonenberg>
its ground anywhere i have RF signals referenced to it
<azonenberg>
3.3 around the perimeter
<Degi>
Like if you have 1 Ohm cable drop, you'd get like 3.0 V
<azonenberg>
and idgaf about ref planes for the i2c etc :p
<azonenberg>
My PSU supports remote sense if it comes to that, but i expect nowhere near 1 ohm of drop
<azonenberg>
i can also just monitor the actual load and bump up the voltage a bit
<azonenberg>
this is meant for benchtop use in a controlled lab environment
<azonenberg>
and it was a quick test fixture, i didnt want to bother throwing regulators on it
<Degi>
The RF splitters with the 100 µm solder mask look kinda tight tbh. But if that works out that'd be great, I guess that means that smaller pitch BGAs would work fine too
<monochroma>
azonenberg: not seeing anything jumping out at me, but i'm not quick enough to properly review to get it out tonight
<azonenberg>
Degi: my min design rule for oshpark soldermask is 50 um
<azonenberg>
monochroma: yeah i'll probably wait till the morning to make sure
<azonenberg>
not like they'll run the board overnight anyway
<electronic_eel>
but since it is closed, it all depends on the quality of their firmware - and you can't easily improve/fix it without reversing it all first
<Degi>
1300 MHz input bandwidth
<Degi>
Sadly 200 € per piece
<bvernoux>
electronic_eel, Yes I just say for 199USD he can sell probably 100x more products
<Degi>
Awwh only 11 ENOB, why even make it 16 bit at taht point
<bvernoux>
Degi, yes Fake 16bits ADC ;)
<bvernoux>
Degi, when marketing is too strong ;)
<electronic_eel>
bvernoux: I'm not so sure about that. The julescope grabbed some of the market a year or so ago, so I don't know how many people needing such a product are left
<bvernoux>
electronic_eel, Yes maybe but I say for a CrowdSupply which is pretty limited audience the best is to sell for price lower
<Degi>
Lets do something like that but with the 32 bit ADC
<Degi>
Dynamic range from 1 nA to 2 A
<bvernoux>
electronic_eel, but it seems clearly better than Joulscope
<electronic_eel>
also it is just 6V input, which limits the use severly
<bvernoux>
yes 6V max is pretty limited ...
<electronic_eel>
you can't use it with 12v or 2S lipo
<bvernoux>
If it will be 40V/2A it will be a different story ;)
<bvernoux>
to test any HW
<electronic_eel>
even if it was 15V it would have been much more usable
<bvernoux>
yes especially for 499USD
<bvernoux>
it is why I say for such niche market it shall not exceed 199USD
<bvernoux>
to have way more guys buying it anyway we will see
<bvernoux>
it is made in India ;)
<bvernoux>
it is first time I see that
<bvernoux>
ha ok the guy name is Prajay Madhavan
<bvernoux>
;)
<Degi>
Sadly no fast ADCs with more than 12 ENOB
<bvernoux>
Degi, yes it is very hard to find that
<Degi>
I kinda wonder what the heck this PCIe slot is doing
<bvernoux>
as there is fake 16bits ADC with 10bits ENOB ;)
<Degi>
Like just hang some LFSR on the last 6 bits and tadaa 16 bit haha
<bvernoux>
also some 12bits ADC with 8bits ENOB ;)
<bvernoux>
electronic_eel, anyway what is great with CrowdSupply is price inlcude VAT
<bvernoux>
electronic_eel, so there is no surprise when you receive the HW ;)
<electronic_eel>
are you sure about that? do they now send DDP?
<azonenberg>
honestly i think if we ever wanted to do that, i think we would be better off building a from-scratch mobile client using libscopehal and a completely new UI around it
<azonenberg>
glscopeclient was intended for a large screen with mouse and keyboard
<Degi>
I can attach my smart TV with VNC or so probably
<bvernoux>
Degi, Yes SmartSCope ;) for the Price of a full Rigol 1GSPS ;)
<Degi>
Or a long HDMI cable
<azonenberg>
we could keep all of the protocol decodes etc but i don't think the current UI is practical to port just for ergonomic/UI constraints
<Degi>
You can get a rigol for 139 USD?
<Degi>
Well 179 $
<bvernoux>
Degi, A friend have bought a portable scope like that for 99USD too I said Him for 200USD you have a full Rigol 1GSPS 2Chan with Screen ...
<Degi>
Idk the UI seems fine even for touch. Maybe redesign the menus
<Degi>
200 USD for a rigol with 1 GSPS?
<azonenberg>
Degi: i rely on right clicks and things like that, and i have fairly small ui elements
<bvernoux>
Degi, yes there is special offer on the old ;)
<Degi>
Oh
<azonenberg>
and scroll wheel, etc
<azonenberg>
i do actually want to have PC touchscreen capability
<Degi>
Scroll weel can be replaced with pinching
<azonenberg>
been eyeing a 24" 1080p touch monitor on amazon to experiment with that
<Degi>
Thatd work just fine for time base and voltage
<azonenberg>
BUT i want to avoid making it less useful for mouse/keyboard with big chunky UI elements
<Degi>
Hmh
<azonenberg>
Since that is the primary use case
<Degi>
Can we change the UI bigly with the css files?
<azonenberg>
adding touch friendly alternatives under the hood like gestures is possible
<azonenberg>
The CSS only themes the GTK chrome
<bvernoux>
Degi, the DS1052E was 199USD few months ago
<azonenberg>
menus, toolbars, dialogs, etc
<bvernoux>
Degi, on a special offer
<azonenberg>
what it will not change is the custom rendered elements like the channel info boxes, the trigger arrow
<bvernoux>
Degi, DS1102Z is 299USD ;)
<Degi>
I kinda wanna build a good AWG sometime...
<azonenberg>
Degi: that's on my list
<Degi>
But theres some kinda inverse correlation between bandwidth and amplitude
<azonenberg>
i
<Degi>
Like I cant just have 100 Vpp at 1 GHz
<azonenberg>
i'd love to collaborate at some point but i wanted to get at least one scopehal instrment out first
<Degi>
Hm yes
<bvernoux>
Degi, hehe 100V @1GHZ ;)
<Degi>
Currently I have to do with other things too
<bvernoux>
Degi, it is not common such high voltage at such freq
<azonenberg>
and i was looking at 50 ohm output for the AWG, not high impedance / high voltage
<Degi>
Yes 50 ohm
<Degi>
Still 100 V 1 GHz :P
<azonenberg>
lol
<azonenberg>
how many kW is that?
<bvernoux>
yes 50Ohms AWG is perfect
<Degi>
0.2
<bvernoux>
especially reaching >6GHz ;)
<bvernoux>
as they are crazy epensive
<Degi>
My scope can only output like 2.5 V into 50
<bvernoux>
especially with nice filters
<bvernoux>
not a bad freq generator like we see on Ebay for 50USD ;)
<bvernoux>
with tons of harmonics everywhere
<Degi>
Huh, smartscope can only export to .csv and .mat and is pretty pricy for what it does
<Degi>
I mean AWG not freq gen
<bvernoux>
Degi, ha you want just AWG
<azonenberg>
see i would be fine with 10-20 dBm output power
<azonenberg>
maybe even a bit less, 18 dBm = 5V p-p
<azonenberg>
which is the max for most 50 ohm test equipment
<Degi>
5Vpp is kiinda low tbh
<Degi>
Id want to be able to drive TTL at least
<azonenberg>
5Vpp is ttl if you have a dc offset on it, no?
<Degi>
Yes if it can do that
<Degi>
The rigol cant somehow
<azonenberg>
i was targeting dc coupled output with a reasonable offset range
<azonenberg>
say max of 5V p-p and +/- 5V offset or something
<Degi>
Hmm ok...
<azonenberg>
we could then make an external PA/offset stage if we needed more
<Degi>
yeah
<azonenberg>
If/when i build an AWG, i was planning on using the AD9154
<azonenberg>
8 lane JESD204B interface, 16 bit, 2.4 Gsps
<bvernoux>
azonenberg, My Agilent E4432B have an AWG option but it is defective anyway I doubt it will be useful as the software to load waveform is horrible & obsolete
<azonenberg>
scopehal actually has APIs for function generators (not arbitrary yet) and in theory supports the wavesurfer 3000's integrated function gen
<azonenberg>
(although that may have bitrotted as i no longer own a ws3k and can't test)
<bvernoux>
Degi, 10GHz 30USD USB SCope ;)
<Degi>
Its 600 usd
<bvernoux>
Degi, Even 10GSPS is hard at that price ;)
<Degi>
Its a sampling scope
<bvernoux>
yes so it is not an Oscilloscope ;)
<bvernoux>
it was advertized as Oscilloscope ;)
<azonenberg>
bvernoux: thoughts on the ad9154 for an awg?
<Degi>
Its a sampling oscilloscope...
<bvernoux>
ha yes I see on 2nd line => USB Sampling Oscilloscope
<azonenberg>
we'd get four channels of 16 bit dac out of it at 2.4 Gsps so nyquist bandwidth over 1 GHz
<bvernoux>
azonenberg, yes it is a nice DAC
<Degi>
The ADRF6720 are external and not part of it right?
<azonenberg>
It's only $120 on digikey too
<azonenberg>
Degi: yes
<azonenberg>
it doesnt do any QAM internally
<azonenberg>
however it could easily be turned into a dual qam signal source with external components
<azonenberg>
This siggen would need to be kintex7 based and probably have a 1G ethernet interface
<azonenberg>
streaming of raw waveform data wouldnt be possible without a much beefier FPGA and if i use 8x GTX for the JESD204 I won't have any left for 10G ethernet
<bvernoux>
Supports input data rates up to 1 GSPS
<azonenberg>
unless i move to something bigger than the 7k160t which is the largest kintex supported by free vivado
<Degi>
Yeah what does that mean
<azonenberg>
i wonder if that means it does interpolation internally and you can't give it data faster than that
<azonenberg>
but the dac runs faster? let me look
<azonenberg>
even so a 1 Gsps AWG would be pretty nice
<Degi>
Uhm yeah look at page 4
<bvernoux>
I suspect it is what is used in Rigol AWG
<bvernoux>
the spec seems to match well
<azonenberg>
you think rigol's awg is using jesd204?
<azonenberg>
So let me add this project to the roadmap
<azonenberg>
since i think we all agree it's going to happen, we can figure out details later
<bvernoux>
fun things is the output can go to a Mixer for higher freq ;)
<bvernoux>
to do also RF stuff haha
<Degi>
And then into a TWT
<azonenberg>
bvernoux: that would be a separate standalone system IMO
<azonenberg>
I'd make the AWG differential output too
<bvernoux>
ha yes great
<azonenberg>
so you have four differential outputs and can terminate _N if you dont need them
<azonenberg>
then you can just run SMA-SMA cables into a separate I/Q mixer and upconverter if you want an RF signal source
<azonenberg>
i personally would be more interested in using it for baseband stuff
<bvernoux>
yes will be very nice
<bvernoux>
me too ;)
<azonenberg>
simulating lower speed signals like 100mbit ethernet with various signal distortions etc
<bvernoux>
but a good AWG is very nice to test electronic stuff too
<azonenberg>
Yeah
<Degi>
And TWT amplifiers
<azonenberg>
I figure if i have a sine source to 6 GHz and a full arbitrary to 500 MHz i'll be in pretty good shape
<bvernoux>
I still plan to build a compact YIG FIlter )
<Degi>
Didnt bvernoux find that one good 11 GHz mixeer or so
<bvernoux>
But I need more details on how to do the driver
<Degi>
For the magnet?
<bvernoux>
as I'm pretty sure a good one from 2GHz to 18GHz can be put in a 20cmx20cm*10cm box ;)
<Degi>
I think its mostly a constant current source
<bvernoux>
yes to drive the YIG with heater ...
<Degi>
Heater?
<bvernoux>
yes there is heater
<Degi>
Well two constant current sources
<azonenberg>
What do you think of FLEMING as the signal generator?
<Degi>
Heater is probably fine with a class D modulatedo ne
<bvernoux>
as the YIG shall be stable ;)
<azonenberg>
after the inventor of the vacuum tube
<Degi>
Heh nice
<bvernoux>
main issue is power consumption
<Degi>
Maybe use a peltier
<bvernoux>
I have not found something better than YIG for ultra sharp filter
<bvernoux>
with such tunable band
<bvernoux>
from 2GHz to 18Ghz or more
<Degi>
Maybe cooling helps
<azonenberg>
Degi: btw i think MAXWELL and FLEMING will have a *lot* in common architecturally
<Degi>
Is maxwell fleming^-1
<azonenberg>
no maxwell is the LA
<Degi>
Ah yes
<Degi>
If you want a better filter, get a superconducting cylinder where one end is height adjustable
<azonenberg>
But both will be an xc7k160t with all gtx used for high speed io, all HP banks used for a ddr3 sodimm, in the same 1U rack case, with the same stm32 based control system
<azonenberg>
speaking of which, we will have a lot of OSERDES unused on FLEMING
<azonenberg>
so i am considering throwing a "free" digital signal generator into the mix
<Degi>
Hmh yeah
<azonenberg>
a bunch of MMCX or SMA outputs with 1.25 Gbps LVDS and maybe some lower speed LVCMOS
<Degi>
Maybe something cheaper
<Degi>
ribbon cable socket or so idk
<azonenberg>
we could add some lower speed connectors too
<azonenberg>
actually we could do 1.5 Gbps
<Degi>
Or a pcie slot hahaha
<azonenberg>
because FLEMING will need a -3 speed fpga
<bvernoux>
Anyway for those which are interested I will open my KC908 when I receive it ;)
<azonenberg>
i think
<bvernoux>
there is lot of interesting things inside for RF stuff ;)
<Degi>
Wear gloves
<azonenberg>
need to fine tune that later on during the design
<azonenberg>
but it's on the roadmap now, will push later on
<azonenberg>
also assuming we use a 4GB SODIMM
<azonenberg>
that's 1GB per channel or 512M points per channel of AWG :D
<Degi>
Why not more heh
<azonenberg>
because i think when you go bigger you need to be dual rank
<azonenberg>
which hurts Fmax for the 7 series controller
<azonenberg>
it's faster in point to point mode
<azonenberg>
anyway so given the dram read latency and refreshes
<azonenberg>
i am probably going to want to do a prefetch of some sort
<azonenberg>
with a block ram fifo to cover latency bursts. We'll figure that out later
<azonenberg>
I also plan to build a DDS based sine, ramp, etc generator in there as well
<bvernoux>
yes DDS is a must have too
<bvernoux>
for fast signal generator
<azonenberg>
and of course have trigger input for syncing the sweep to other systems, trigger output for syncing DDS/repeating AWG pattern to an external scope
<azonenberg>
10 MHz refclk in
<azonenberg>
all the usual high end accessory ports
<bvernoux>
yes definitely 10MHz refclk ;)
<bvernoux>
for all instruments
<azonenberg>
Yes. no matter how low end
<azonenberg>
it's stupid not to
<bvernoux>
yes
<azonenberg>
anyway i'm not fully out of bed and dressed for the day (yes it's 14:00 here, but it's the weekend during quarantine :P)
<azonenberg>
So i think i'm gonna hop in the shower and find some breakfast :p
<bvernoux>
hehe
<_whitenotifier-f>
[starshipraider] azonenberg pushed 2 commits to master [+11/-0/±1] https://git.io/JfQ3p
<_whitenotifier-f>
[starshipraider] azonenberg 288238d - Updated README/design notes with signal generator concept ideas
<_whitenotifier-f>
[starshipraider] azonenberg 76f216f - Initial skeleton of MAXWELL main board
<bvernoux>
bye see you
bvernoux has quit [Quit: Leaving]
<Degi>
Woow I found a datasheet so old it advertises 500 kbps and 2.456 Mbps as high speed
<_whitenotifier-f>
[starshipraider] azonenberg pushed 1 commit to master [+0/-0/±2] https://git.io/JfQZf
<_whitenotifier-f>
[starshipraider] azonenberg d6be583 - Final version of sfp-to-sma after signoff review