azonenberg changed the topic of #scopehal to: libscopehal, libscopeprotocols, and glscopeclient development and testing | https://github.com/azonenberg/scopehal-apps | Logs: https://libera.irclog.whitequark.org/scopehal
esden has quit [Ping timeout: 240 seconds]
sorear has quit [Ping timeout: 240 seconds]
esden has joined #scopehal
sorear has joined #scopehal
<_whitenotifier-b> [scopehal-apps] hansfbaier opened issue #393: starting glscopeclient fails (Option 'strip-debug' registered more than once) - https://git.io/JM3Wa
<_whitenotifier-b> [scopehal-apps] hansfbaier commented on issue #393: starting glscopeclient fails (Option 'strip-debug' registered more than once) - https://git.io/JM3WH
Degi_ has joined #scopehal
Degi has quit [Ping timeout: 250 seconds]
Degi_ is now known as Degi
someone-else has quit [Quit: Connection closed]
<azonenberg> So i'm starting to work on porting my scope VNA design from glscopeclient to the standalone scopevna test app
<azonenberg> with any luck i'll get much more stable results this way
<_whitenotifier-b> [scopehal] azonenberg pushed 2 commits to master [+0/-0/±2] https://git.io/JMsYT
<_whitenotifier-b> [scopehal] azonenberg e8b09f5 - FIRFilter: fixed bug causing coefficient calculation to hang if a zero-span bandpass configuration is requested
<_whitenotifier-b> [scopehal] azonenberg 530165b - MagnitudeFilter: fixed typo causing garbage output
<azonenberg> Measured S21 of PP023 + WaveRunner 8404M-MS 1MΩ frontend
<azonenberg> Uncalibrated ratiometric data, so mismatch between scope CH1/CH2, splitter ports 1 and 2, and the two FL086-24SM+ coax cables all contribute to measurement error
<_whitenotifier-b> [scopehal] azonenberg pushed 1 commit to master [+0/-0/±1] https://git.io/JMsBX
<_whitenotifier-b> [scopehal] azonenberg beb024e - LeCroyOscilloscope: added support for "ring" name on passive probes (does not have auto zero)
someone-else has joined #scopehal
<electronic_eel> @Aleksorsist will the thunderscope have switchable 1 meg / 50 ohms inputs or is it just 1meg or just 50 ohms like what azonenberg plans for his scopes?
<d1b2> <Aleksorsist> Just 1 Meg inputs, which is inline with other 1000 and 2000 series scopes. It's scopes that can do 500 MHz + that you generally see switchable 50 ohm termination crop up.
<d1b2> <Aleksorsist> I've been using an inline Terminator when tested the front end with my VNA tho
<azonenberg> Are you planning on doing flatness correction for the frontend/probes?
<azonenberg> if you can measure the vector response of the channel we can use the de-embed filter to correct for that
<azonenberg> That's what the data I linked above was for
<azonenberg> it lets you move the reference plane for a measurement from the scope front panel out to the probe tip
<d1b2> <Aleksorsist> That's really cool! You just need the touchstone S files right?
<azonenberg> Yep. as of now it only supports .s2p files, it can't parse higher port counts, but that will change later
<azonenberg> You can pick any path you want to de-embed. Or do the reverse, channel emulation
<azonenberg> so for example you can generate an ideal unit step then do channel emulation of S11 on it to see the TDR response of a touchstone file
<electronic_eel> hmm, i did some comparisons with the true 50 ohms input of my scope and some bnc inline termination resistors. there was some difference, even at 350 mhz. so the missing 50 ohms is one of the two big issues i see for example with the rigol mso5000 series (the other being data export speed to scopehal)
<azonenberg> or create an ideal PRBS and then do channel emulation and clock recovery to see an eye pattern through the channel
<someone-else> electronic_eel: I'm working on a mso5000-level frontend which will include both 1meg and 50ohm
<someone-else> hopefully waveforms in a month or two
<someone-else> wait, mso5000 doesn't have 50ohms?
<azonenberg> no, it's like the #1 complaint people have with it
<azonenberg> rigol reserves 50 ohm input for their higher end SKUs
<electronic_eel> someone-else: no, the rigol mso5000 doesn't have 50 ohms. that is what i really dislike on it
<azonenberg> which i think actually use the same asic
<someone-else> nevermind, I'm working on a mso7000-level frontend then
<someone-else> yeah, the rigol asic - apparently they do some very interesting almost entirely solid state switching in the 1meg attenuator there
<someone-else> the same as tek does
<electronic_eel> what kind of switching do you mean? ac/dc coupling?
<someone-else> 1meg rc divider attenuation
<someone-else> 1:2-1:100 or so
<someone-else> which is needed for high input voltage ranges
<someone-else> traditionally this was done with relays, since voltages there can be up to +-50v
<someone-else> but now at least rigol and tek have asics which only need 1 relay and the rest of the switching is done inside the asic
<someone-else> example
<someone-else> from mso5000, I think
<someone-else> no relays there at all
<electronic_eel> hmm, even ac/dc coupling seems to be done with a solid state relay
<someone-else> ac/dc coupling was always done this way
<someone-else> tek has 1 relay (for 1:1 attenuation, I think)
<someone-else> anyway I possibly have found a way to replicate this without an asic
<someone-else> works in sim, hardware testing soon
<someone-else> having no relays in 1meg path help signal integrity very much since non-coaxial relays are really terrible at 0.5-1GHz up to which the 1meg path is supposed to be functional
<electronic_eel> i must confess that i just looked at older generation frontends, where 50/1meg, ac/dc and dividers were all done with mechanical relays
<someone-else> yep, that's the normal way, but it's very difficult to achieve 500 or more MHz this way
<electronic_eel> properly using 500 mhz with 1meg is hard. i think tek has recently developed some new 1meg probes that can work at these frequencies. but usually you are better off using 50 ohms at these freqs
<someone-else> I know, the bragging rights though..
<someone-else> I think tek has 1ghz 1meg probes, but they cost a lot
<electronic_eel> i think these are the ones i meant. they were able to significantly reduce the capacitive loading somehow.
<someone-else> 3.9pf is still a lot though
<someone-else> as is $911 MSRP
<electronic_eel> you can get a full rigol mso5000 for that msrp (and hack it to 350)
<someone-else> with 4 channels no less
<electronic_eel> now if they would just add 50 ohms inputs and improve remote data readout speed...
<someone-else> I think it's called mso7000
<electronic_eel> don't think the remote data readout speed is much better with the 7000 series
<electronic_eel> i'm eying the rigol ds8000-r. they have 10gig sfp+ output
<someone-else> well, at least you get the 50ohm
<someone-else> 10gbe doesn't guarantee higher readout speed though..
<electronic_eel> no, that is why i haven't bought one yet. they write something about "opensource remote management software" on their website, but haven't published it anywhere yet.
<someone-else> would be classic rigol to lure everybody with the nice shiny 10gbe sfp+ but then leave the firmware-limited readout speed the same lol
<someone-else> would be nice if it works of course
<electronic_eel> yeah, then you would get a 2 ghz / 10gsps / 4ch scope for about 3500 eur (given that you can hack the 350 model, but that is very very likely)
<electronic_eel> if the sfp+ output is fast, then it would be a really nice scope for glscopeclient
<someone-else> is it 8 bit or more btw?
<electronic_eel> i think just 8bit
<electronic_eel> they have some high res mode, but that is just software averaging
<someone-else> I see
Yamakaja_ has quit [Quit: Bye]
Yamakaja has joined #scopehal
Yamakaja has joined #scopehal
Yamakaja has quit [Changing host]
<Degi> I guess we could replicate the high res modes in scopehal by adding a sinc brick wall filter?
<azonenberg> And yeah i designed my prototype front end using a coaxial relay IIRC
<someone-else> yep, but coax relays don't really help with 1meg path
<someone-else> Degi: if willing to trade bandwidth for resolution, yes
<someone-else> could also average several waveforms together with no loss of bandwidth if the signal is repetitive
<azonenberg> someone-else: well in my case it was a 50 ohm path only
<azonenberg> and the relay was a SPST not a SPDT
<azonenberg> it was there as part of the protection circuit
<azonenberg> basically it opens if the input is overloaded to prevent the clipping diodes from melting
<someone-else> yeah, in 50ohm environment proper relays are very nice
<azonenberg> yeah i consider 1M probes to be deprecated :P
<someone-else> can replace the relays for +-5V 50ohm input, I think
<azonenberg> I needed something that could sustain much higher voltage in the open state
<azonenberg> for the protection circuit
<someone-else> in that case, yeah
<azonenberg> IIRC my initial frontend was designed to survive a dead short to +/- 50V DC
<azonenberg> The relay was the first thing in the signal path, followed by a 50 ohm 6 dB pi attenuator using big chonker 1206 resistors
<azonenberg> then fast ESD clamp diodes clipping the output to the +/- 5V power rails
<azonenberg> I sized the series resistor in the attenuator, and the diodes, to survive the current you'd see from a 50V short for about a millisecond
<azonenberg> the relay opens in a few hundred us
<someone-else> sounds good
<azonenberg> and is activated by a comparator, no firmware or gateware in the feedback path
<azonenberg> A high power RF overload would probably not have enough DC component to trip the comparator, so it would still cause damage
<azonenberg> but a DC transient would be almost instantaneously cut off
<someone-else> I just had an idea of putting a smd fuse of roughly similar width to the microstrip into the 50ohm path lol
<azonenberg> lol
<someone-else> would be very cheap
<azonenberg> Yeah this was designed to be a fully resettable "electronic fuse" style inpuit protection
<azonenberg> it would light up a red overload LED and require the channel to be manually reset if it popped
<azonenberg> but nothing would be permanently damaged
<azonenberg> I might try and see about adding some kind of RF power sensor to detect up to a few hundred MHz of high power RF as well and also trigger the shutdown
<someone-else> I wonder what kind of circuits you plan to work on to warrant this level of protection :-)
<azonenberg> Test equipment inputs tend to get a lot of abuse
<azonenberg> I want to design something that users can be a little rough on and not kill it
<someone-else> makes sense
<Degi> Hmm, maybe we can add a GDT
<Degi> I think some might even go down to 60 V or so
<someone-else> although the standard way for 50ohm scopes is adding a beefy external attenuator when working with >5V circuits
<someone-else> I think
<Degi> Hmm, the good thing about GDTs is that they have a high damage threshold
<Degi> Otherwise shorting the input might be an option
<Degi> Depending on the type of fault applied
<Degi> Wouldn't want that with a huge capacitor
<azonenberg> someone-else: yeah that was something i wanted to avoid here
<azonenberg> the intention is, you're probing a low voltage signal
<azonenberg> but if your probe slips you might accidentally hit a 12V power rail or something
<azonenberg> and i want to not blow the frontend
<Degi> ohh right
<azonenberg> my main target fault was "dead short from SMA input to power rail"
<azonenberg> having 50V tolerance means even a 48V DC rail won't be fatal if you hit it with a 1x probe
<Degi> Hmm yes
<azonenberg> i figured that would be more than enough for any realistic use cse
<Degi> So nothing like accidentally connecting the input to a magnetron or 230 V power rails
<azonenberg> Yeah
<azonenberg> i would like to be able to hit a DC-DC inductor by accident and not die though
<azonenberg> so a ~2 MHz 12V squarewave should be something it will survive
<azonenberg> i'm not sure if my current protectiono is fast enough to detect that
<Degi> Hmm yes, I guess 1 meg and 50 ohm have very different use cases (like usually 1 meg inputs are rated to 400 V or so but if you switch to 50 ohm its like 5 V)
<azonenberg> Yeah