LCD timer, 2nd gen.

This project has been ongoing since summer last year and by any measure has been drawn out and troubled. False starts, design error & fabrication flaws that rendered boards salvageable, and various outside factors that delayed getting things done. As a result this article is not a complete telling of the story behind the PCB.

Background to making LCD timers

Experimentation

The first stage was understanding how LCD displays worked and the specifications of the driver chip that a display needed, which was the end result of the effort that went into LCD display & driver. I realised that rather than being able to wire up an LCD panel the same way I could test an LED module, LCD displays require AC signals that came with their own nomenclature. However this was in parallel with the task of finding an LCD panel that had the segment layout suitable for time display, which was just one of many challenges.

Packaging

The next stage was turning the experimental circuit into a self-contained PCB that would be used as a daughter-board with minimal connections. Difficulty with routing all the tracks made me use it as an excuse to see what designing a four-layer PCB would be like, which I have no regrets about even though I now know that two layers were sufficient. Main realisation is that laying out tracks is more of an art than a science.

Practical use

After a long delay I finally got round to building a timer that used the LCD display daughter-board, which in turn was based on a prior LED-based timer — the latter separated timing and display control which communicated via I²C, so only the content of the I²C update commands needed updating. Initially prototyped using perf-board it was later converted into a PCB, but unfortunately it was botched. After some rework I did finally build the enclosed timer I had long-intended.

Circuit design & implementation

LCD panel & driver sub-circuit

Looking back at my original experimental try-out article the circuit wiring table was very useful as it mostly negated the need to trawl through data-sheets, aside from a few spot-checks in places I felt a mistake was possible, but it was from the perspective of the LCD display which made checking driver pin connections a bit difficult. The table below gives the pin connections for the PCF85162 itself, together with an overview of what is connected to them. The only pin I would consider changing is SA0 which chooses between one of two I₂C addresses. The pins A0 thru A2 use some contrived mechanism that allows multiple drivers to use the same I₂C address, which to me seems close to abuse of the I₂C specification.

When I was designing the stand-alone LCD time-display PCB I had trouble with the number of tracks that needed to cross, and I ended up making it a four-layer PCB rather than the usual two-layer. In itself I have no regrets about making it a four-layer but having laid out the connections again from scratch, it is now clear that a fundamental mistake I made was having the driver chip horizontal rather than vertical. Rather than crossing the wires needed to be attached to the driver pads underneath the chip, they are routed out the ends and around the wires that are attached from the outside — this is most obvious just off the bottom-left corner of the driver chip in the cross-section rendering below.

Previously I would have avoided putting tracks between mounting holes and for space reasons I wanted all tracks to be between the two rows, but this would have just meant using a lot more vias. This time round I specifically tried to minimise vias, as this was intended to be part of a much larger PCB, and I think getting it down to six was good going — the extra four to the right are place-holders, since the rendering was made before the rest of the circuit was laid out. I later managed to remove the three vias that are below the driver chip.

Timer control circuit

I originally decided to rely on a programming clip to flash the PIC16F1823 and reuse the adapter I built previously, but when I had to redesign the circuit due to faults I decided to add a dedicated flashing header to the second revision. A problem with programming clips is that sometimes it is hard to get a good connection.

Sounder circuit

I eventually built a proof-of-concept driving circuit that used two 555 oscillators — one to control the pitch, and one to control the on and off periods of sound. This approach was used as-is and is shown in the schematic below — the output of the second 555 drives a Pro-Signal (now part of Multicomp) ABT-414-RC sounder via a Sziklai pair of transistors, although the latter is not shown. While it is possible that one or both 555s could be omitted and the task instead done in firmware, the hardware solution provides more intuitive control over how the alarm sounds, and allows the same firmware from the previous timer to be re-used unmodified.

For some reason at the time I designing the circuit the NE555P was only easily found in DIP form, and I usually avoid through-hole components for PCBs due to them taking up space on both sides of the board. Because the NE555P at least in theory requires at least 4.5 volts I switched to the LMC555 which is a pin-compatible variant that needs only 2 volts. For the second board revision I also added provision for a grounding capacitor to the CV pins rather than leaving them floating, as well as bypass capacitors.

Power supply

Circuit design faults

Incorrect PCB markings

For some reason the PCB markings had “ARM” even though this is really PIC-based, and I can only speculate as to why I put in this incorrect text. A somewhat careless cosmetic error that may be due to having designed and ordered for fabrication other ARM PCBs around the same time.

Incorrect Sziklai pair wiring

The Sziklai pair that powers the sounder had the collector and emitter of the PNP transistor the wrong way round in the schematic. It is possible this was taking account of a pin-flip on the PCB foot-print, but due to differences between pin numbers in KiCad versus the transistor data-sheets I ended up deleting and redoing this part of the circuit from scratch.

Missing I²C pull-ups

During development I switched between having a single PCB and a two-board design, and I suspect somewhere along the line the I²C pull-up resistors got omitted when I recombined the schematics into a single circuit. However omitting pull-ups is a mistake I have made more times that I want to admit, not helped because for testing purposes I often in the past used I²C devices with a master that had them built-in.

Components too close together

The proximity of the microcontroller to one of the 555 trimmers meant that I was not able to securely attach the programming clip. At first I thought it was due to excess solder causing a poor grip, but after many cleaning attempts I notice the clip banging into the trimmer. This is the fault that made me abandon the PCB and do a redesign.

555 oscillators wrong way round

The two 555 chips in the sounder circuit had the wrong capacitor footprints, which was a problem as 220μF is not available in SMD size 0805. I only noticed this one later when recycling the board for lead-free solder testing and it is just as well I held back ordering in the revised PCB.

Circuit production

PCB fabrication

As it turned out when Seedstudio fabricated the second revision of the board there was a serious manufacturing flaw. Several pins were bridged together that should not have been, as shown in the picture below, and this rendered the PCB useless. Had I noticed this back in March when I received the order I would have asked them to do the job properly, but since I did not unpack the PCB until July it was a bit late to demand that. I downloaded the Gerber files attached to the order in my Seeedstudio account and they looked fine — I actually submitted this zipfile unchanged to Aisler who went on the fabricate the board properly.

I suspect this error is deliberate sabotage by a disgruntled employee, because it does not look like the sort of thing that would be caused by a simple slip, and in the past Seeedstudio have spotted and corrected mistakes on my part. Either way I was obviously disinclined to trust Seeedstudio with the respin, especially with a 3-5 week turnaround time.

Component sourcing

The LCD display and drivers were from an existing stock I bought over from the US via Digi-Key back in early-2018 but I found out they are also available from Mouser without having to pay the import duties — I am guessing it is cheaper for them to pay the duties than to actually have a warehouse in Europe. I am not sure if Mouser have the same range as Digi-Key but Mouser is now my go-to vendor for stuff that Farnell do not stock.

Component soldering

Post-reflow rework

I am not sure how the flux paste compares to the flux pen I have used a lot in the past, but it does at least have the advantage in that it does not start flowing away until a soldering iron is applied. Bridge-removing rework is a bit of an art, sometimes with bridges lifting up easily and at other times remaining stubborn. The soldering iron was brand new so I suspect being very clean was a factor.

Remarks