OCXO for HP 53131A – A New Design

[Hexley]

The availability of $10 OCXO modules on eBay recently prompted me to upgrade the reference oscillator in my HP53131A counter.

I looked at HP’s original design, and also the excellent work done by Gerry Sweeney to extend that design to support multiple types of OCXOs. But I decided to go in a slightly different direction.

So I designed a new version of the OCXO board, using a Microchip DAC to replace the discontinued part from Analog Devices that was used originally.
The DAC that I chose supports an external reference voltage, which made it possible to use the OCXO’s own reference voltage in place of the separate chip that supplied that voltage in the original design. This simplifies the design. And since the reference voltage source is now inside the ovenized oscillator, the overall temperature stability of the system should be improved a bit as compared to a system with an external reference chip.

The Microchip DAC has a simple interface that is similar to, but not identical to, the original DAC’s interface. Both DACs accept a 16-bit SPI transfer and regard the last 12 bits as the data for the converter. The new part uses the first 4 bits for control signals, however; furthermore, it uses a different edge of SCLK to sample the incoming data line. So some glue logic is needed.

Referring to the schematic, the glue logic amounts to three chips: a counter (U4), a multiplexor (U3), and an inverter package (U5). The counter/multiplexor select hard-wired values for the first 4 bit cells in the incoming SPI data stream, then switch to that data stream for the remaining bits. The inverter corrects the sense of the clock and chip select signals, and also stops the counter when it reaches the fifth bit cell.

The timing diagram shows more detail about the interface. The “MOSI DAC” data sent to the DAC has the first bit held low (selects Channel A of the DAC), as well as the second bit (disables the internal reference voltage buffer); then the third and fourth bits are held high to select unity gain in the output buffer and to hold the DAC in active mode. After that, the bits are simply the input data sent by the counter (“MOSI”).

The schematic shows the circuit as built. I used the dual channel MCP4922 DAC because I had some on hand; the single channel MCP4921 should work as well. I also added three LEDs to show the presence of supply voltages, just to remind me that the board is always live when the counter is plugged in.

The OCXO used in the prototype was a CTS 970-2187-46; I have another unit that uses a Bliley NV47A1282 OCXO. Both are 5-volt, “Euro package” units that were available on eBay for $10. These appear to be telecom grade units, but work fine for the purpose at hand.

After installing the prototype in the counter I was able to perform auto calibration against an external Rubidium standard, and wind up measuring 10 MHz as 10 000 000.00 Hz, as expected.

[Hexley]

Here are some details of the prototype construction, for those who might be interested.

Total cost was about $20, with liberal use of the junkbox.

[jiarui]

Thanks for sharing.  I just purchased a 53132A from Ebay. I have ordered a 3Ghz prescaler, and an on OCXO oven. I'm planning on building the time base myself also.  Have you considered using just a PIC controller to do the protocol matching?  I would reduce the chip count a little.

[Hexley]

Have you considered using just a PIC controller to do the protocol matching?  I would reduce the chip count a little.

Yes, a PIC seems well-suited to do the protocol matching. You would receive the 16-bit SPI data from the counter, flip the proper bits, then transmit the modified data to the DAC. Probably could be done with one hardware SPI channel (for the counter data), and with bit-banging for the DAC. So even something as small as a 16F628A might work.

Good luck with your project, and congrats on the new 53132A.

[HighVoltage]

Very nice project.

[bitseeker]

Welcome to the forum, Hexley. What a great first post. Thanks for sharing.

[texaspyro]

The 12-bit DAC limits the accuracy to which the HP auto-cal routine can tune the OCXO (typically to around a milli-Hz).   

I considered building what amounted to a GPSDO based OCXO board for the HP531xx.  You would provide it with a 1PPS signal from a GPS receiver (which could be on the OCXO board) or a GPSDO and if it saw the 1PPS input it would tune the counter OCXO to the 1PPS signal using a 16-24 bit DAC (chip or PWM based).  If the 1PPS was disconnected, it would use the last DAC setting which was kept in EEPROM.

Another option is to avoid auto-cal circuitry and use a manually tuned EFC voltage.  This requires the tuning pot/range limiting resistors and reference voltage source to use very low tempco parts.  You also need an accurate frequency reference (like a GPSDO) when setting the EFC voltage.  Also it is best to mount the tuning pot such that it can be accessed through a hole drilled in the case.

[Zucca]

But if I feed that beast with an external 10MHz reference I don't need an OCXO, right?

[HighVoltage]

But if I feed that beast with an external 10MHz reference I don't need an OCXO, right?

That is correct.
As soon as the 53131A detects an external 10MHz, it will use this as a reference and override any internal reference.