Parallella Community

Lucky guy! Same here, my wife has been extraordinarily patient during my 6 year Epiphany journey.

Although some presensitized pcb and chemical leftovers I had in a corner were well over past the expiry date, I gave it a shot. Checking with a microscope when done, tracks look like Swiss cheese but it works (after some repairs). Ain't pretty though.

For someone considering to make a homemade pcb for the Parallella, the thing to watch out is Vias' running in the middle of the Samtec connectors. To connect top and bottom sides, a wire has to be soldered and on the connector side, the soldering joint should be as low as possible so that there is no short with the contacts extending below the connector (I smashed the top of a 30 awg wire with a screwdriver and removed the excess solder with desoldering braid). Of course this is not an issue if the pcb is made by a shop.

Right now I'm using a 7020 Parallella but I'm gonna buy Samtec connectors for the microserver.

Currently QSPI flash contains FSBL, FPGA bistream, NETMF TinyCLR (cpu0), FreeRTOS (cpu1) and the deployed C# application and the complete stuff is up and running in about a second.

There are 4 SPDIF outputs and 1 input, a JTAG connector and the serial port connector to Visual Studio when deploying and debugging the application.

Also even running without a heatsink, after a couple of hours the Zynq is moderately hot. So far so good...

The Diptrace layout:

I didn't intend to write a post about this intermediate step but I had such a hard time with it that I decided to share my experience. Someone will certainly come with a much better way but albeit hard, at least the method described here works.

Say you have a micro-server Parallella and you want to access the signals on the Samtec connectors, probably the easiest way is to directly solder wires on the pcb tracks. It worked fine but then I decided I prefered to have a couple of Samtec connectors.

I don't have a stencil, besides even if the components are glued on the pcb other side, I wouldn't want to get the board through a reflow cycle again, so I went with hand-soldering.

Before on a homemade daughter board, whenever I got a bridge between the connector contacts, I added a bit of solder, increased the iron heat and sucked it all with a desoldering braid. It worked great so I felt pretty confident it would be okay with the micro-server as well. Couldn't not have been more wrong, it took me some 8 hours to get those 2 connectors (kind of) properly soldered !

I guess when a bridge develops between a GND contact and one next to it, the GND plane dissipates the heat and if the bridge is too far below the connector, no matter what I tried I couldn't get rid of it. For 3 times I had to remove the connectors with a heat-gun (the connector is trashed of course).

Eventually I used a nail-like tip for the soldering iron and 0.2mm solder (not lead-free). Also a microscope comes very handy.

- remove the excess solder on the pcb connector contacts
- apply solder to every connector contacts and remove it with desoldering braid
- apply flux on both the pcb and connector contacts with a coton tip
- soldering iron set to 310C(Celsius) apply a little bit of solder one contact at a time without touching the contacts on either sides
- if a bridge develops, immediately change the iron tip to a wider one (~1mm) and set to 370C
- apply a little bit of solder to the iron tip and suck the bridge with desoldering braid
- when a row is done, heat every contact separately to let the solder spread (310C)
- if there is no solder joint contact/pcb, add a bit of solder
- check for bridges with an Ohmmeter
- finally glue the connector to the pcb

I use non lead-free solder because I find it easier to use. A fan is next to my desk to avoid breathing too much of the toxic smoke.

The iron is set to 310C because if it is too hot, the solder being very thin will melt by just getting close to the tip, however a physical contact is needed between connector contact/pcb track/iron/solder.

I glued the connector because the amount of solder per contact is very small and I'm not sure the bond is very strong.

Although this method works, it is very hard and I don't recommend it even to non-beginners. Do at your own risks.

I figured out a significant improvement after my previous post regarding hand-soldering Samtec connectors.

Looking at the iron horizontally, I apply solder to only the top of the tip so that the bottom will quickly oxidize if there is no solder on it. Then I solder each contact with the top of the tip, adding solder as it flows onto the contact. If the bottom of the tip touches the contact below, it doesn't matter because there is no solder on it.

Microscope is a must to monitor the amount of solder flowing between the connector contact and the pcb track for a good bond.

I don't do step 2 and 3 any longer (apply solder to the connector contacts and apply flux). Also 0.5mm solder works fine, just wipe it out from the tip if there is too much.

(actually I'm even starting to enjoy it, even though originally it turned out to be a real nightmare!).

Coming up next: the enclosure.

Amazing to see you project coming together like this! I am impressed by your confidence and skills with the soldering iron. I would probably have given up and sent it to the professionals..

Can't wait to see what it looks like when it's finished!

Andreas

Thanks Andreas. Packed with features yet great form factor and great price factor. What you guys are doing and the price you're offering it are awesome. A heartful thank you.

All enclosure parts cut with a cheap Chinese CNC. It's probably not meant for metal because I couldn't cut a line straight when the thickness exceeded 0.5mm.

I don't have a proper tool to fold the copper and aluminium sheets so I used a hammer instead.

I made an additional small SPDIF with its voltage translators extension board in order to keep the daughter board somehow generic enough for future projects.

I had a couple of this heatsink: which I cut to the proper width. Unfortunately, even though I tried to make the SPDIF extension board as narrow as possible, the length was ~2 mm too short. Still the end-result is not visually too shocking.

(guess by now it's pretty clear why I went through the trouble of soldering the Samtec connectors to the micro-server instead of using a standard Parallella)

Very nice work indeed! There are few things more satisfying that seeing a project like this develop, combining many different skills to make a finished product.

Interesting to hear about the issues encountered with a budget Chinese CNC mill. For a while I couldn't decide between one of those and an old Denford Novamill (small machine used in schools). In the end I opted for the latter, over concerns about the rigidity of the frame with the former. Sounds like I made the right choice — the Novamill is small, but cast iron and IIRC weighs a few hundred kilos. Although the downside is that I haven't yet got round to ripping out it's old control board and wiring it up for LinuxCNC. And I'm presently distracted by a Chinese CNC plasma, with a hand controller that only seems to work with a very limited number of old/small USB flash drives (it's starting to look like I'm setting up a USB memory stick museum!)

Just noticed the heatsink arrangement. Nice touch.

Cheers,

Andrew

Thanks Andrew. The Denford Novamill looks like a VERY nice CNC.

Below is a picture of the one I bought on Ebay in 2012 to turn it into a pick-and-place machine


I wish you success with your project.