8 bit PIC with RTC chip.

You need to show us the circuit diagram and ideally some pictures of the hardware.

The kind of problem you describe is often caused by bad layout, inadequate power supply decoupling or ciruit errors.

Without more information it is not possible to tell which, if any, is the cause here.

MK

Do your board follow general recommendations for crystals like following?

• No component under crystal, only ground plane
• Groud plane around crystal from all sides on the same layer
• Traces to chip as short as possible and preferably both the same size

Michael, thanks for the fast response. Well, the project was quite simple with fast building, so there is only a handmade schematic and the circuit is soldered to a prototype PCB. But I try by words... Power supply, step by step the components: and old NOKIA cell phone charger with 6...8V DC output, 1000uF/16V electrolit + 100nF ceramic capacitor, L4941 low drop 5V regulator, 1uF tantalium capacitor on output. PIC: 100nF ceramic closest to its pin1/pin20 power connections. PIC works with internal osc, 32MHz. RTC: 100nF ceramic closest to its pin4/pin8 connections. RTC osc: 32.768 crystal on pin1 and pin2 (closest), 2x22pF ceramic. Remark: the MCP79411 datasheet offers 6pF capacitors, but trying these, not working. The rest of the circuit (2 LEDs and 2 deprez solenoids from 4 PWM output) has all together 15mA current consumption with 100 duty. Thats all. Maybe the problem is caused around the RTC crystal/capacitor fail fitting. But it is clear, only at activated PWM signals detected. Anyway, similar issue was detected at my earlier project, a vintage Nixie-tube clock. When I tried to regulate the light of the nixies by switching PWM the 170V DC nixie power, the RTC went to fail timing. With different PIC, but with the same RTC.

You will have far more success if you learn to plan and draw schematics. Even if freehand in your notebook they can be copied.

What is a "deprez solenoids"

MK

Thanks, the recommendations are clear, but to easy to realize on a prototype board, check the attached photo. So it is not the best one, additionally the RTC chip is soldered into a SOIC/DIP PCB. But similar layout worked well in other boards. Only if a PWM is running, I have the problem. Can be the root cause the interference between the frequency of the RTC osc (32.768 Hz) and the used PWM period=32msec (31.25kHz)? May I try with totally different PWM period?

Is your PWM output on pin number 3?

The PWM signals are generated by the PIC, as all together 4 PWMs I need: meter for hours, meter for minutes, LED backlight for hour meter, LED backlight for minutes meter (these are bright regulated depends on the lighting of the environment), I use the PWM3, PWM4, PWM5, PWM6 capability of the PIC16F15345, pin7, pin6, pin5, pin8. All PWM channels are running from TMR2, with same base period, but separatelly programmable duty%. This part works well, I have wondered that this small PIC has 6 PWM outputs.

It sounds like the solenoids are messing with the power supply. Do they have snubber diodes? Can you isolate the power to the RTC? - supply power through a small resistor so you can put a capacitor on the power pins at the RTC. 

Please explain how the backup battery and the USB power interacts. Beside the hardware, the I2C commands can mess with the timer. For example, if you have code that stops the timer then restarts, it can explain loosing the 30 second you’ve mentioned.

Main power at normal operation, 5V: supplied the PIC VDD pin, supplied the MCP79411 RTC on pin8 VDD, supplied the anods of the LEDs. The deprez coil are drived from PWM pins of the PIC VIA NPN transistors, the common other ends of the coils are on the 5V of main power. RTC backup (when the main power is switched off):   simple a 3V coin battery is connected to the RTC pin3, as the manual describes. The i2c communication must not disturb the time management of the RTC. The program: the RTC alarm is programmed in sec=0 match, so once per minute. The alarm pin7 of the RTC is connected to the INT pin of the PIC, it is prorammed for falling edge interrupt. When the once per minute interrupt occurs, the interrupt rutin reads the hour and minute timekeeping register of the RTC via i2c and clear the alarm flag. I think this is the regular way. Finally here is my nice schematic...