10MHz Rubidium Standard

Well, actually, this has been buttoned up for a while, but life (and sadly also death) intervened in the past two months. I've been enjoying a nice stable 10 MHz reference source for my experiments. I used it recently to calibrate OCXO replacement project for my HP 5328 frequency counter. The last bits I finished up this weekend were to fix some software bugs - the most annoying was the too short a waiting time to allow a cold LPRO-101 to achieve warmup. I also added the LPRO case temperature sensor and worked out the ADC scaling to report in degrees C.

I still have the runt pulses in the comb generator to work out, but I'll probably debug that using one a spare PCB and, once I find out what the issue is, apply the fixes to the completed unit. So, not quite "complete" yet, but really close enough for the summertime!

I used Front Panels Express to fabricate the front panel for my standard. I received the panel on Friday. Wow! Very nice indeed, much better than my normal acrylic panel attempts.

It took a couple hours to install all the switches and LED and connectors to the panel, but when it was done, it's all working.

...well, sort of...

The comb generator, which I hadn't tested much because it's sort of the "low priority" item, has an issue with some of the comb rates. I suspect it might be in need of a bit more bulk supply bypassing in the 74HCT151 mux chip, but that's just a guess. There are some "runt pulses" into the 74AC74 pulse generator FF that cause it to double-trigger. weird.

I started on the firmware for the PIC that I chose for this project - the PIC16F887. I had several of these laying around, so it seemed appropriate to use one of these for the job. I also decided to take a try at the Microchip XC8 free C compiler, rather than Sourceboost C I had been using for many years now.

I used a state machine approach for the control code, because that's normally how I think. In this case, it's really simple, so the state machine is a breeze.

One of the goals for this project was also to use a WS2811 based T-1 3/4 LED for the status indicator. These have a pretty annoying PWM-type interface, that runs a bit fast. Unfortunately, so fast that the PIC 16F877 can't generate the proper bit timing when running from the internal 8 MHz clock. I even tried bodging in a 20 MHz xtal oscillator clock, but it was still too slow (why? not sure. wouldn't have expected that).

So, I'm swapping out the PIC 16F887 for a PIC 18F45K22, which has an internal 16 MHz oscillator and a 4x PLL, so I can boost the clock up to 64 MHz. That will most certainly be able to generate the WS2811 bitstream.

So, yes, pretty dumb that I ended up going to a whole new micro just to accommodate the status LED but, hey, that's the way things go sometimes. I'm also eager to work with the PIC 18 series again - it's been a long while since I've done so.

I've been incrementally stuffing and testing the controller board. First was the power supply section; it works fine. The 10V intermediate rail came right up, and from that the 5V rail was trivial. The 8V rail required that I force the 8V rail enable transistor (doh - should have had a test point there) Q5 on. I trimmed that rail to 8V without issue. Not tested is the 24V high-side switch; need to find the chip in my messy shop.

Next, I stuffed the 10 MHz distribution circuitry. This is a 2-way splitter right from the 10 MHz Rb oscillator. One branch (the "analog branch") gets amplified up and then fed into a 4-way splitter. The other branch (the "digital branch") gets Z matched to about 1.5K ohms and fed into a 74AC08 gate to convert from sine to square. I used my HP 8660 sweeper to feed in a 10 MHz test signal at -20dBm, and used my HP 8557 spectrum analyzer to check levels along the analog path. Then increased the power to +13 dBm (about the same as the Rb oscillator) and saw that I was getting the expected levels at all four outputs of the final passive splitter. The digital branch is working, also, but I think I'd like to play with the DC level to get the duty cycle on the 10 MHz square output closer to 50%.

It's hard to test the comb generator without the microcontroller up and running, so I'll focus on the micro next. The comb generator needs an 8:1 multiplexor forced to select the comb rate (100kHz ... 10 MHz) so that's the holdup there. So, next steps are getting the control program written to allow switches to turn on/off outputs, etc.

I received the PCB's today. They look pretty good - this is my second project to use the "dirty PCB" service, and I'm satisfied. I'll start the build in a week or so - this week there's too many other things going on to get the uninterrupted time that I'd like.

Waiting on boards...I submitted this and the DerbyLink order to DirtyPCB service, so looking forward to seeing what comes back!