Making the rev0 more usable

[tnt]

Here I'll present the few mods I made to my parallella rev0 to improve it's usability.

This first post will focus on the hardware side. And I'll do a second post as a reply on this thread for the software side later this week.

So let's dive in:

UART/Console port

To play with the kernel or bootloader or just general debug, it's essential to have access to the serial port. Unfortunately on the rev0 it's only accessible through the PEC and not via a standard .1 inch header like it will be on rev1.

Solution is simples, just add the header ! On the picture below you can see where to connect, I just soldered a 3 pin header using the center pin as ground soldered onto one of the GND 'post' for mechanical support and then just ran a couple of 30AWG wire to the buffer chip.

UART connector

uart_full.jpg (183.01 KiB) Viewed 27971 times

Force IDSEL low

If you want the use the USB host port, you need a cable that pulls the IDSEL pin low for it to work currently. Mine didn't have that which meant the port didn't really work (no power basically). I'm not planning on using that port as a device port, so the solution is simple, just force it low in all cases.

Again, AWG30 to the rescue and there is a resistor in the IDSEL path which is conveniently located near a big GND pad.

IDSel forcing

idsel_full.jpg (51.97 KiB) Viewed 27971 times

Acrylic support

Using the bare board wasn't really practical, I wanted something to hold it a bit better. So I took a scrap of thick acrylic and made a small base for it using metal standoffs. The tolerance of the hole is pretty tight, they are M3 and drill on the board is like 3.2mm so not much room ... hopefully if the holes aren't perfectly aligned, just use a soldering iron to heat the standoff and move it a bit :p

Acrylic base

base_thumb.jpg (67.94 KiB) Viewed 27971 times

[tnt]

Heatsinking

As mentionned elsewhere, the board heats ... a lot ...

The rev1 should be much better but I still wanted to be able to use this rev0 normally without worrying about thermal issues. So I added heatsinks to all the main source of heat. The zynq is easy, it's nice and very flat. The epiphany is not flat at all so it's not great. I used thicker thermal adhesive but still the contact area could be better. The other chips that heat a lot are the PMIC (voltage regulators). I used two very small heatsinks for them, just beware not to short any of the capacitors nearby. I had to cut a bit of the aluminum on the bottom of those to avoid issues.

Heatsinking

heatsink_full.jpg (137.8 KiB) Viewed 27969 times

Fan

Despite the heatsink, if the air is stagnant, they get hot. Too hot. The epiphany will start producing wrong results and the zynq has weird behavior. So I took a sunon quiet 5V fan and stuck it to the side of the board, connected in parallell to the 5V input supply.

You can barely hear it and the board is now very cool and very stable.

Fan

fan_thumb.jpg (95.04 KiB) Viewed 27969 times

Power input

As can be seen in previous pictures, I don't use the barrel jack. The one on the rev0 doesn't make good contact and makes it unstable (i.e. you just touch the cble and the board reboots ...).

So I basically soldered two wires directly to the board and I'm feeding 5V from a lab supply. Beware that at nearly 1.5A of current the resistance of wire is not negligible, you can easily get a few hundred millivolts of drop ...



That's it for the hardware side. Hopefully this will help/inspire other people


Cheers,

Sylvain

[Sundance]

[mark03]

Hmmm... as someone with a gen1 board on pre-order, I'm curious, how has the thermal management improved gen0 -> gen1? What's different?

[tnt]

Here are the improvements I know about:

* The rev0 uses LDO to drop 600mA of 5V to 1.8V. That's about 2W of pure heat waste. The rev1 uses a switching regulator for that, so you could expect ~ 85% efficiency, reducing heat losses to like 0.2W on that rail. That's 1.8W of heat less generated. On a board that currently draws 6.6W that should be more than 25% less heat being generated in the first place.

* The rev0 epiphany package is not great thermally. High theta_ja and not flat, so not easy to passively cool. The final chip will have a much better package with better thermal dissipation and a flat top allowing passive cooling if one wants to.

* Thicker copper on the ground plane, allowing better thermal conductivity, to spread the heat over the whole board, thus offering more surface area to exchange that heat with the ambient air.

* More thermal vias and via plugging (fill them with solder) to conduct more of the heat to the ground plane

* More bare metal exposed area around the mounting hole so that you can conduct the heat away to a case for example.

[mark03]

Mmmm, interesting. I assume the Epiphany Pdiss is dependent on what code it is running. If I'm only using the Zynq for a while, can the Epiphany be put into a some form of power-down?

[aolofsson]

mark03,

There are a few different power saving options for the Epiphany:

1.) Enable the aggressive clock gating mode in the CONFIG register to reduce dynamic power
2.) Turn off the unused link ports to reduce static power on the IO
3:) Turn off the clock to the Epiphany in the FPGA logic (work in progress, need to hack the FPGA logic for this not implemented in official repo yet)
4.) Use the ARM to program the ISL9305 with the I2C interface to turn off the VDD_DSP rail all together or to reduce the operating voltage to something that still retains state (most significant of the mentioned options)

Andreas

[pilloutche]

Hello,

thank you for this post !

can you give us the size and (if you have) a reference of all thermal dissipator (heatsink) you used ?

thank you in advance

[9600]

[notzed]

Had a bit of scare the other day when i rotated the monitor it's under to show something to a mate and I must've knocked something - some time later a nasty cracking sound and i saw the red light on and a 'hot' smell. But it was just the cd it was mounted on cracking, the mounting standoffs were too hot to touch even though the fan was still running. I rebooted it while still hot and it has worked ok and has ever since. Not sure what happened, maybe a jiggle of the power crashed it into some high-power state.

But really i'm here to ask: will HDMI ever work on the rev0 boards? I seem to remember it just being fpga-firmware related but I admit I haven't been keeping track of whether there's been any updates.