Jaerder's Personal Blog : From the Electron to the framebuffer

And all was good. The latest version of the nixie clock survived it’s smoke test, and, as usual, I’ve let it powered for a while and then something quite awkward started to happen : The clock always got consistently late. Its was quite odd, since most of the PCB remained the same, except for the inclusion of the Nixie Power Supply and the re-positioning of several traces and components.

The fact that the failure is consistent leads me to believe that it should not be related to the Nixie’s HV PSU presence on the same PCB as the RTC. And the fact that the seconds are twice as slow (a consistent behavior) is clearly measurable.

After some digging, I’m inclined to believe that the root cause for the clock not behaving as expected was due to the pcb track re-routing.  When re-routing the PCB traces, some traces passed very close the the RTC’s crystal pins.  And yes, this is noted on Maxim’s RTC’s aplication notes.

Well, back to PCB etching!

Fixed design : GND guard ring around the RTC, and the traces were re-routed to avoid being close to the cristal.

The 2nd version of the nixie clock was done. The PCB’s were good, the EMI issue traced and fixed, therefore the clock met all specifications I’ve set for it.

Nixie Clock Version 2

Or not quite.
At First the ideia of having 3 diferent PCB’s made sense, since each board had a specific function (Nixie and Driver PCB, µC Pcb, User IO PCB), but this aproach led to a awkward  configuration to make or find a proper housing for. And there is still the issue of the PSU shielding, should the enclosure not allow for a bigger space between the PSU and the µC Pcb.

Not to mention the fact that I don’t have proper tools to cut the PCB’s, leading to some work to cut the PCB in 3 pieces for the Driver, µC and user IO PCB’s.

Therefore I’ve decided to make a incremental change to the Schematic and PCB design of the 2nd version of the clock.

HV Nixie PSU Breadboard Prototype

The goals I had in mind for the last revision of the clock were to :

• Have the Nixie PSU on the µC PCB, instead of a separate, pre-build PSU.
• Have the clock built in such way that it would be pleasing to look at, even when using a transparent enclosure.

On the PSU side, instead of designing one from scratch (and I seriously doubt that it would be trivial….) I’ve decided to use this schematic instead. The main reason was the lack of exotic components
I was not able to use the exact value for the inductor, but once the prototype worked, the PCB design was changed to accommodate the power supply . Previous testing with the power supply suggested that, as long as the PSU is far enough from the RTC and the I2C bus, I would have no EMI problems (before the testing I was considering using metal shielding on the PSU to avoid EMI issues).

Nixie Clock Version 2 PCB alignment check

On the “Pleasing to look” at side, since the PSU and the µC are on the same PCB, the user IO board, with a PCF8574 IO expander and some pull up resistors, was also put onto the same PCB. Therefore, the clock was condensed to only 2 PCB’s.

The connector between the Nixie board and the µC board was , along with the screw holes, aligned in order to be possible to assemble them as shown on the left photograph.

Nixie Clock Version 2 PCB (µC, PSU and User IO)

When the PSU and User IO boards were combined, I was able to better use the PCB’s surface area, since I could put the components as close as I could , regardless of their function.

Any new Hardware features may still be added without modifying the PCB layout, since the I2C bus is exposed by a 5 Pin Header.

Detail of Nixie Clock during assembly

Smoke Test of the PCB. The Nixie Diver board is the same from version 2 (they are compatible).

At this time, the nixie clock has passed trough 2 revisions.

The 1st version

The 1st one was a clock with a purely transistor based nixie driver.
The I2C bus was able to handle the multiplexing fast enough for the clock to be usable, but there were random Micro controller crashes. Those crashes were caused by noise induced on the SDA and SCL lines on the I2C bus (these lines were passing under the wiring of the nixes driver, including the HV PSU. This PSU is nothing more that a boost converter, so electrical noise should be expected).

Although the prototype reached the working stage, it never got past the test type (PCB). The clock failed promptly at its first power up test (aka smoke test).
The root cause was on the PCB design.

Due to the fact that EMI interference from the boost converter could bring the micro controller to a crash, I though it would be a good ideia to leave as much grounded copper on the PCB as possible.
In theory its a good ideia, but it did not take into account that the method used to build PCB can’t achieve the needed precision.

PCB issues, lacquer between tracks.

It was observed that lacquer residue was present between tracks, and that this residue was not always visible.
I believed that this could be fixed, by slightly over-etching the board, but this proved  to be ineffective.

And, during soldering, the solder would easily stick to the grounded copper that surrounds the tracks and pads, shorting them. Therefore, this added another point of failure to the test type unit.
Re-work proved quite tedious.
And to make thing even worse, the Micro controller source code was lost, due to the fact that the Arduino IDE saved the sketch the a folder that was not backed up.

Therefore, I’ve chosen to make a simpler nixie driver design.

The 2nd revision of the nixe clock.

PCB, 2nd version

The 2nd version of the nixie clock used anode drivers similar to the 1st version, and used a K155ID1 (equivalent tho the 74141) IC as the cathode driver. The interesting thing about this IC is that it driver 10 outputs from a 4 pin BCD input.
Other diferences include :

• The Micro controler drives the nixies trough its own IO’s instead of the I2C bus
• The I2C in only used to get the RTC data, user IO and this bus is exposed trough a expansion header, allowing the addition of expansion modules.
• The nixies themselves are on the PCB instead of a separate socket.
• Although the 2nd version of the clock uses the same PCB area that the 1st version used, the PCB is separated into 3 boards, the nixie board, that holds the nixies and the nixie drivers, a board the has the micro controller and the board the handles the user IO.

Nixie Clock with the PSU detached.

Nixie driver board

While I was looking for some information on VFD displays on wikipedia, I’ve came across this very old display technology used in the 50′s and 60′s, called Nixie tubes. After doing some homework, I’ve decided to build a nixie clock.

Nixie Tube IN-12

The main focus of the Nixie Clock  was not to build a discrete time keeping  circuit ,  but instead to build a transistor based discrete Nixie driver, handing over the time keeping tasks to a dedicated RTC and a µC, in this case an ATMEGA 168.  Unlike my previous endeavours, this time I’ve decided to put all other hobby related projects on hold, and start focusing in one project at a time.

Besides building the clock, the project also became a way to re-learn techniques and procedures, like the PCB positive preparation, CAD PCB design  and acquire (or build)  the means necessary to build other projects, such as  the UV box.

Nixie Clock Prototype

At this time, the clock is on a prototype stage, where the HW and SW are stable enough to work for several weeks without a HW reset.

I’m currently working on the test stage, were the clock will be assembled on a properly etched PCB, with a design being made on KICAD, a Electronics oriented EDA tool.