When I started this project, I acquired 2 board sets; one to use as an everyday unit, and another to hack on. There is substantial potential in this design through software, and the cost per unit is reasonable, so obtaining a pair seemed wise. In these last few weeks, I’ve been assembling the first of the 2 units, the second to come later. That’s what this post is about.
Acquiring parts is the first task (unless you purchase the full kit from Chris; I elected not to). The project has origins in the UK, and the original Bill of Materials mostly references part numbers as sourced by Farnell/element-14 in Europe, with much of the remainder eBay sourced. Here in North America, most of the Farnell catalog is available through Newark/element-14. Yet, most of the BoM is fairly standard stuff available from other more familiar sources such as Mouser, where I sourced most of mine (and who I’m used to dealing with…heck, with Grant Imahara as a spokesperson, you have to be in good company…).
There are some exceptions and single source items, such as:
- HY28B LCD displays – single source (have to come from China, required about 3.5 weeks via Singapore Post registered service). I ordered 3; one as a spare, because once these become unavailable, I’ll be well and truly stuck.
- Si570 programmable oscillator – kind of the “heart” of this radio – this seems to only be available via DigiKey. I have not ordered just yet.
- Mitsubishi RD16-HHF1 Power MOSFETs for the transmitter final amplifier. There are a small number of North American resellers of these, and they’re not all that cheap. Fakes appear to be very common on eBay.
- Inductor cores – available via Amidon, Palomar, or others. Some, I already had in my possession.
- The specific 3.5mm connectors specified for this PCB needed to come from Newark (using the Farnell numbers), no other equivalents appeared to match the footprint.
The photo above shows the state of the boards as I currently have them, with approximately 20-25 hours invested in assembly so far. Most assembly is complete with a few exceptions:
- No display yet: I’m thinking of putting it in a socket, because of the damage that could be done if it needs to be removed. It connects using an uncommon 2mm pin pitch (instead of the typical 2.54mm/0.100″ header pitch). Sockets are available in that size, they’re just uncommon.
- No encoders: I’d prefer to solve the display socket question first, because encoder shaft length might be tied to socket height.
- No pushbutton switches on the UI. They’re backordered.
- Have not wound the BPF/SWR inductors yet; I don’t have them all.
- Do not have the finals yet, and I didn’t install the driver transistors yet; I think I’ll wait to get the receiver and UI working first.
- No relays for the BPF.
- I still need to get the Si570 oscillator. It will be difficult to solder correctly, so I’ve left off the nearby passives to make sure I have room to heat it well. Many builders report trouble soldering this 6-pin 5x7mm QFN part, which is not at all surprising; it was meant for wave soldering. So, I want to take some care with this.
- No serial EEPROM on the UI yet. Early builds of the firmware didn’t use it, but newer ones do.
- Not sure if I want to use the stock 30-pin header for the UI-RF board interconnect, or rig up something allowing more space between boards. They stack to 0.425″,which is just enough. However, I might want enough space between the boards for a metal sheet for RF quieting reasons, and that may require more space.
- There are still some optional component value selections from recommended mods, which I need to make final decisions on.
This is my first exposure to SMT assembly at this level; I’ve done minor rework to SMT before, but never assembly on this scale.
For assembly, I’m using a simple temp-controlled soldering station with a fine-point conical tip for most assembly (using .015 67/33 flux-core solder wire), and it’s working fine for most things. It took a little practice to get good quality joints, starting with the larger pitch SO-x chips, working my way up to the 100-pin LQFP CPU. That one took about 60 minutes to pin down clean one pin at a time (I didn’t like the drag method myself). I did bridge a few pins, but they cleaned up OK.
A friend lent me an unused hot air pencil for rework; I haven’t needed to use it so far. For unit #2, I may use it for some of the assembly.
Of course, I haven’t seen the unit run yet…until I do, no claims about my quality of workmanship…
I should note, it’s certainly not a project for a first-time builder, although it could be accomplished by a first timer with a little coaching, I think. It is time consuming, and there is potential for difficult-to-detect/correct mistakes in many places. However, if you’ve built a few things before, you’re patient, and are ready for a challenge, it’s not so bad. The boards were laid out with room to spare in most places, and most of the passives are 0805 dimension; not too small for manual placement. In the future, SMT is what we’ll be doing for most everything; learning to deal with it is just going to be a requirement.
More to come after I make some decisions, and finish a few other tasks. Until then…