1
\$\begingroup\$

I'm in preliminary stages of designing a PCB for my project which will have a PIC18F4680, an PCF8583 RTC plus some logical circuits and transistor arrays (exact components are to be decided) for driving 210 LEDs (six 5x7 matrices), all in through-hole technology. The PCB will be home-made which at this time limits me to a single sided 10 cm by 16 cm board plus jumper wires.

To keep the noise as low as possible and to deal with small board size (since the LED displays are going to be 52 mm tall), I'm planning on having the LED displays (and perhaps some of the logical components driving them) placed on perf-boards connected via jumper wires to one edge of the PCB and place the PIC close to that edge.

I've read in the datasheet and manual that in order to get as much accuracy as possible, I should place the RTC in a quiet part of the board (so opposite side of the PIC) and pour Vcc under the RTC, its crystal and trimmer capacitor and solder the crystal can and one side of the capacitor to Vcc (yes, I know that the usual procedure is to use GND for this, but the datasheet explicitly calls for Vcc with this model).

On a two sided board I expect that this would be done by having the top, component-side, layer be Vcc and have it go under the chip and in case of a PTH components, have a plated-through hole connected to the Vcc for RTC's power supply. Unfortunately, I don't have the luxury of having plated-through holes and multilayer boards, so I'm considering my options.

The simplest option would be to just pour a Vcc layer under the RTC, capacitor and crystal with the two of them on the component side and give up on soldering the crystal can to Vcc layer. I'm not sure how good (or bad) results would this produce.

The second option would be to put the crystal can under the RTC and solder it to the Vcc pour and RTC pins with RTC being on the component side. That would look something like this (the location of decoupling capacitor for this question isn't important): enter image description here

Right now, I don't see any problems having the crystal on the bottom side of the board and it would be very easy to solder it to the RTC. This, I think, should provide me with greater immunity from noise than the first option.

The third option would be to place both the RTC and the crystal on the bottom side of the board and maybe only leave the trimmer capacitor on the top side (depending on the type of enclosure I manage to find). The problem here is that I'd have the pins of RTC sticking out of the board on the component side and that I'm not sure how I'd actually solder the RTC. I could perhaps solve the problem by using an SMD version of the RTC, but I'm not sure that I could make PCB which would be precise enough to fit the component on it.

So what I'm asking is: Are there any negative side effects that I missed and in particular are there any negative side effects if I decide to go with option 3?

Also, in case anybody here worked with this particular RTC, what can I expect if I decide to go with option 1? Am I over-thinking this whole issue? I'd like, if possible, to keep the whole thing within ± 5 minutes a year, as far as accuracy is concerned.

\$\endgroup\$
  • \$\begingroup\$ So, what did you do in the end ? \$\endgroup\$ – Rocketmagnet Jun 16 '12 at 20:14
  • \$\begingroup\$ @Rocketmagnet Well if I'm lucky, the PCB will be done tomorrow (I'm still fixing my UV lamps so that they can cover the entire 300 mm x 200 mm PCB nicely) so I can still change my mind, but the I'll most likely use SMD RTC with Vcc pour under it and I'll solder the crystal body to the Vcc pour which will be under the crystal. The crystal will be on the bottom side of the PCB, next to the RTC. I've decided to go with a 33 pF fixed capacitor instead of the trimmer. In the end the design changed a bit and I'll be getting time from GPS, so in total the RTC's accuracy doesn't need to be that great. \$\endgroup\$ – AndrejaKo Jun 16 '12 at 20:32
  • \$\begingroup\$ @OK, well GPS makes everything good. Thanks for the tick. \$\endgroup\$ – Rocketmagnet Jun 16 '12 at 21:23
1
\$\begingroup\$

To be honest, I don't see any problems, though I have read the question twice now.

I don't even see any problems soldering the RTC, as an SO8 package really isn't that small, even for a hand made PCB. What's your PCB production method?

To maintain good accuracy, you might consider keeping the circuit at a controlled temperature. You could use a Peltier element plus a thermistor to regulate the temperature.

There are such things as temperature compensated oscillators (TCXO) and Oven Controlled Oscillators (OCXO), but these don't seem to be available at 32.768kHz.

\$\endgroup\$
  • \$\begingroup\$ I'll most likely be using one of the toner transfer methods such as shiny magazine paper method or should it fail, blue press and peel transfer foil method. I had pretty horrible results with both in the past and at one point gave up on making my own PCBs, but in this case, I think that I should use a PCB instead of a perfboard. I don't think that for this type of project it would be worth the trouble to keep the crystal at certain temperature. \$\endgroup\$ – AndrejaKo Apr 16 '12 at 17:43
  • \$\begingroup\$ If you have a good printer, I can highly recommend the tracing paper method. Print the PCB onto tracing paper if it's an inkjet, or onto translucent acetate if it's a laser printer. Tape it ink side onto a photo-resist PCB, and leave it out in the sun for an hour (under a sheet of glass to hold the paper flat onto the PCB. Then develop, wash it in water and rub it well with a wet paper towel, and etch it. \$\endgroup\$ – Rocketmagnet Apr 16 '12 at 18:08
  • \$\begingroup\$ Seems interesting! I'll try that one out. \$\endgroup\$ – AndrejaKo Apr 16 '12 at 18:11
0
\$\begingroup\$

You need to additionally care about PCB inductance and capacitance components when you designing Microwave/RF circuits. Here I don't think you need to take a special care.

Unlike RC the temperature coefficient of Crystal is very low in Physics. Crystal watches usually skip a one second per 5 years. Since you using a RTC, you don't need to care about it.

from wikipedia:

A common parabolic coefficient for a 32 kHz tuning fork crystal is −0.04 ppm/°C

\$\endgroup\$
  • 4
    \$\begingroup\$ A typical watch crystal will have a frequency tolerance of 20ppm (a good value!), which means that it may be off by more than 10 minutes a year. That's a lot more than 1 second in 5 years, which is simply wrong. Also the graph in the datasheet shows a deviation of 20 ppm at a 25°C temperature difference, that's almost 1 ppm. The value of 0.04 ppm may be correct for temperature changes less than 0.1°C (top of the curve), but in practice this is irrelevant. 0.04 ppm/°C is wrong as well. \$\endgroup\$ – stevenvh Apr 16 '12 at 18:04
  • \$\begingroup\$ I have read it in a manual, "1 second per 5 years" ! BTW I don't think it have any mechanism to synchronize with UTC broadcast. Some RTC boards have. \$\endgroup\$ – Standard Sandun Apr 16 '12 at 18:13
  • 1
    \$\begingroup\$ Do you have a wrist watch? These are pretty good temperature controlled through the contact with your skin. Don't you ever to adjust it? Mine is off by about 15 seconds a month, which falls neatly in the 20 ppm. One second in 5 years is 6 ppb. That's right: parts per billion! \$\endgroup\$ – stevenvh Apr 17 '12 at 6:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.