Friday, 24 April 2020

New Raspberry Pi 4B up and running!

For about two years now, I've been running a Raspberry Pi 3B+ as my portable/mobile computer for digital operations. In all that time, in all that banging around and being cooked, then frozen inside a car every day, the Pi has never failed.

Fairly slow, but painless test installation of the new Pi 4B, already running WSPR.

With this success in the background, I thought I'd try a Pi 4B, with 4GB RAM.  One thing I expected would make the 4B less attractive than the 3B+, itself a very capable and fast machine, is the added power consumption, fed by a lot of anxious online talk - as there also was when the 3B+ came out - about CPU heat.  I elected to fit a passive aluminium heatsink case because of this, instead of a standard plastic or metal case.  I used heat transfer compound to make contact with the chips, rather than the supplied pads.

It turns out, though, that at a fairly typical resorce load, the 4B only uses about 6% more power than the 3B+.  With all the cores pushed to the limit, the difference increases to 31%.  But use of WSJT-X is fairly undemanding, and so we can assume something like the lower limit for this application.  The heatsink proves to be only hand warm when running WSJT-X - nothing to worry about there.

What does this mean in practice?  Well, I can run the 3B+ plus its 7" Waveshare screen for at least 5 hours, and probably much longer from a small ~20Ah Li USB battery pack.  So the 4B should not suddenly see usable times come down dramatically.

Pi 4B decoding nicely!  Waveshare 7" LED touchscreen (with my own timber frame, to attach to a 'go-box' lid later on)

The real question is: do you need a 4B and its slightly higher power use for field digital operations?

My view is: no, you don't, even though the two versions cost much the same.  The 3B+ is perfectly good enough for all digital modes and has, I think, the perfect balance between performance and power use.

I can't see anything other than a marginal increase in decode speed of weak WSPR signals with the 4B.  The 3B+ has also proved itself over and over again, although there's no reason to doubt this won't also be true of the 4B.

UPDATE: I ran the Pi 4B for 4 hours and 10 minutes later in the day.  At this stage, the USB battery was still only half empty, although it may be that non-linearity in discharge means we can only expect maybe another 60-90 minutes of further reliable operation.

So, roughly 5-6 hours' operation from a hand-sized power pack, with plenty of waking the screen up during this period, is a very good result and indeed much the same as that seen with the Pi 3B+.

Accurate temperature measurement with a DMM probe inserted into the heatsink fins.

Decoding speed is a couple of seconds faster than my HP laptop with decoding set to 'deep' on both systems.

Interestingly, the Pi 4B doesn't seem to produce the RFI that is often seen with a 3B+, which can be quite strong, even when housed in a metal casing.  This strongly tips things in favour of the 4B for radio work. 

Despite having all that aluminium in intimate contact with the board, wifi connection reliability remains excellent.  Similarly, the Bluetooth connection outdoors is good out to around 15-20 metres in the clear.

Temperature measurements with a multimeter probe inserted deep into the aluminium heatsink gave 37 degrees Celsius after an hour of operation (ambient = 18.5 degrees C), confirming the low resource demand of WSJT-X.

Anyway, I've a fairly exhaustive page on setting up the Waveshare screen and some advice on WSJT-X set up here.


No comments: