15:05 <darthrake> @azonenberg While following your search for SMA end-launch connectors on twitter I was wondering if you have any experience or opinion with solderless compression connectors. E.g. HRS HK-R-SR2-S (https://www.hirose.com/en/product/p/CL0338-0006-0-00) or similar. As the connectors are quite often the most expensive parts on e.g. fixtures or one off filters, the reusability of these solderless

15:05 <darthrake> connectors is quite interesting to me. Of course this doesn't help with more permanent things like probes...

15:38 <azonenberg> darthrake: they can work and give good results. you need very good board planarity (e.g. ENIG not HASL) and repeatable torque on the screws to work well

16:51 <azonenberg> mxshift: so it turns out that i had the baud rate set wrong when i tried decoding your 56G PAM4 capture

16:52 <azonenberg> i'm less familiar with 100GbE and didn't realize that when going from 25GNRZ to 50GPAM4 the symbol rate actually goes up slightly due to increased FEC

16:52 <azonenberg> https://www.antikernel.net/temp/56g-pam4.png

16:52 <azonenberg> This is probably clean enough to actually decode the PRBS9

16:53 <azonenberg> even without an improved CDR

17:02 <azonenberg> https://www.antikernel.net/temp/56g-pam4-viridis.png changed color ramp to show off the transitions better

17:03 <mxshift> Oh, yeah, that looks pretty close to what we saw on the lab master

17:06 <azonenberg> This is using a single threshold. not good enough for jitter measurements

17:06 <azonenberg> because you're measuring at different points in the rising edge for 0-3, 1-3, 2-3 crossings

17:06 <azonenberg> But i think its enough to do decoding

20:33 <darthrake> azonenberg: thx. will give it a try then. Buying a 9Ncm torque "wrench" makes all savings on connectors irrelevant for a while, but I guess it's a good excuse to buy one ;)

20:34 <azonenberg> Lol. Yeah good torque wrenches are not cheap

20:34 <azonenberg> On that note, generally you will not damage an RF connector by finger tightening it

20:34 <azonenberg> you can easily over torque with leverage from a fixed bar wrench

20:34 <azonenberg> but with fingers you will likely be too loose but not dangrously tight

20:34 <azonenberg> This results in a less repeatable connection, which can come loose more easily and not give as good VNA measurements

20:35 <azonenberg> but for quick tests you can get away with it

20:35 <azonenberg> especially for lower dynamic range stuff

20:44 <darthrake> i would like to try de-embedding the connectors with some 2X-Through structure. So I guess investing in some "wrench" will be worthwile. The moment when everything is questioned will come no matter what.

20:45 <azonenberg> yeah for that kind of repeatability a proper torque wrench is the way to go

20:45 <azonenberg> out of curiosity what VNA / software are you planning to use for the de-embed?

20:45 <azonenberg> 2x thru separation and de-embedding has been high on my wishlist for a while now

20:46 <azonenberg> but i havent managed to wrap my head around the math yet

20:48 <darthrake> we are thinking about getting a siglent sna5000. right now we have a used JD785A, but its quite tedious to use. exporting everything and then using scripts to postprocess is kind of non-interactive

20:48 <darthrake> are you happy with the pico? we had a demo units a few years ago, when it was brand new. and the software was quite incomplete.

20:52 <azonenberg> We'd like to support the siglent VNAs in scopehal

20:53 <azonenberg> i think someone was talking about buying one and writing a driver eventually