Jon Klein @loxodes

🛰️📡💻🏔️ github.com/loxodes California, USA Joined May 2011

Tweets

93
Followers

308
Following

290
Likes

140

Jon Klein @loxodes

5 years ago

@arturo182 Would placing the crystal on the bottom side of the PCB help protect it?

1 0 1 0 0

View Details

@PhantomDeadline @tdsepsilon Yeah, it is not cheap.. I'm only building a few, so I'm not super price sensitive. At least it isn't the ADF4371... Cool! The last few times I checked STuW81300 wasn't in stock, looks like it is in distribution now.

0 0 1 0 0

View Details

Jon Klein @loxodes

6 years ago

Testing LiteX on an ECP5 for future use on test equipment control boards. It blinks! (The board is adapted from the icebreaker++-ram with RF bias circuitry thrown in. I was going to add SYZYGY and ethernet too, but after starting routing I decided to buy a ButterStick instead.)

2 8 53 0 2

View Details

Jon Klein @loxodes

6 years ago

@mzollingr FR408 is lossy, but it seems to work well enough over short distances. I'll switch a lower loss substrate on the final design, but it is fine eval boards like this to verify footprints and bias networks. See harmoninstruments.com/posts/oshpark_… for measurements on FR408 out to 50 GHz.

0 0 1 0 0

View Details

Jon Klein @loxodes

6 years ago

@PhantomDeadline @tdsepsilon I'm working on building up a synthesizer board around a TI LMX2820. For testing, I plan to use a Hittite T-2000 with one or two multipliers. (I haven't started RF testing on the mixer board yet, just checked voltage regulation and supply sequencing.)

1 0 0 0 0

View Details

Jon Klein @loxodes

6 years ago

@tdsepsilon The board hiding in the corner is a MAMX-011036 8-43 GHz mixer with a MAAM-011109 as a LO buffer amplifier. (github.com/loxodes/vna/tr…)

2 1 8 0 0

View Details

Jon Klein @loxodes

6 years ago

@GregDavill One of the bugs I caught might be applicable to the ECPBreaker. The LM3880 sequencer has open drain outputs that needs pullup resistors. I caught that - but then I pulled up the 3.3V enable line to the output of the 3.3V regulator 🤦

0 0 2 0 0

View Details

Jon Klein @loxodes

6 years ago

@PhantomDeadline @azonenberg Looks like HASL holds up surprisingly well: ece.uwaterloo.ca/~oramahi/IEEE-…

2 0 2 0 0

View Details

Jon Klein @loxodes

6 years ago

@azonenberg You might have already come across this, but reducing the surface roughness with electrodeposited instead of rolled copper can difference too.

1 0 0 0 0

View Details

Jon Klein @loxodes

6 years ago

@GregDavill @oshpark There may still be reasons to leave out the solder mask. The thickness and dielectric properties of the solder mask may not be well controlled, so it will be harder to control impedance. Also, the solder mask may absorb moisture making your impedance humidity sensitive.

1 1 4 0 0

View Details

Jon Klein @loxodes

6 years ago

@GregDavill @oshpark In the case of oshpark, solder mask reduces transmission line losses. oshpark places soldermask over bare copper, so masked areas don't have ENIG plating. The nickel in ENIG can increase losses. harmoninstruments.com/posts/oshpark_… sonnetsoftware.com/pdf/sonnet-sui…

3 2 15 0 0

View Details

Jon Klein @loxodes

6 years ago

@grajohnt I'm doing my part!

0 0 1 0 0

View Details

Jon Klein @loxodes

6 years ago

I just found my near field probe design cloned on ebay. The kicad files are open too, they could have at least swapped the vertical SMA footprint for an edge launch one 🤦 (github.com/loxodes/emc_to…)

1 4 35 0 1

View Details

Jon Klein @loxodes

6 years ago

@alt_kia @bjt2n3904 @tedyapo (Disclaimers: I typically use ADS and have no idea how good the qucs models are. I made up the 0402 footprint dimensions. Use a non-ideal component model like an s-parameter block from the capacitor vendor. However, extended to 50 GHz the discontinuity is noticeable: )

1 0 2 0 0

View Details

Jon Klein @loxodes

6 years ago

@alt_kia @bjt2n3904 @tedyapo Simulating, e.g with qucs, is a way to investigate tradeoffs. Qucs doesn't have CPWG, so approximating with microstrip... Here a simulation comparing a transmission line with and without a 0402 cap in the middle. The shunt capacitance from the footprint isn't a problem at 1.1 GHz

2 0 2 0 0

View Details

Jon Klein @loxodes

6 years ago

@alt_kia @awygle @kbeckmann λ/10 spacing is great until the vias turn your ceramic substrate into brittle swiss cheese that all the fabs no-bid because they're too scared to ship it... (Yeah, I shoot for λ/10 with FR-4)

1 0 3 0 0

View Details

Jon Klein @loxodes

6 years ago

@alt_kia @kbeckmann Southwest Microwave has a appnote investigating this empirically (via sweep starts on page 14), but generally keeping via spacing to less than a quarter wavelength apart is a good start: mpd.southwestmicrowave.com/wp-content/upl…

1 0 5 0 0

View Details

Jon Klein @loxodes

6 years ago

@bis0nmakinm00 The golden areas are exposed traces for shielding. Soldermask is removed so a metal cover could make contact with the board. This board has an ENIG (electroless nickel immersion gold) finish, so outer layer is about 0.1 µm of gold, over 2-5 µm nickel, over 30-50 µm of copper.

0 0 2 0 0

View Details

Jon Klein @loxodes

6 years ago

Low phase noise, 10 MHz to 6 GHz frequency-locked downconverter passes the smoke test! (The board is part of a phase noise test set, github.com/loxodes/phasen…)