Need ideas to derive 10 MHz standard from 26 MHz OCXO

[rstofer]

Finding a 4 bit counter with synchronous reset is no problem.  Just decode state = 12 and use it as the reset signal.  Dividing by 13 is pretty easy.  It's the multiplying by 10 that's the problem.  Multiply by 10, divide by 13 and then divide by 2 - nice clean square wave.

http://www.nxp.com/documents/data_sheet/74LVC163.pdf

[chris_leyson]

If you want to preserve the OCXO phase noise, I wouldn't bother trying to derive 10MHz from 26MHz unless of course the phase noise is really bad and then you might have a chance. Maybe divide by 13 and multiply by 5. Theoretically the divider will reduce the phase noise but that all depends on the logic family you use, faster is better as a general rule. Then you have to consider the buffer amplifier between the OCXO and divider, how much pase noise does that introduce and the 5x frequency multiplier will increase the phase noise. You have to have really clean power supplies to avoid the AM to FM conversion in anything non linear and the best I could do with minimal component count a small AT cut crystal and a divide by 4 divider was -150dBC/rootHz at 1kHz offset. A CT cut crystal would have been better but bonkers because of the cost. What is the phase noise spec for your 26MHz OCXO ?

[chris_leyson]

It's the multiplying by 10 that's the problem.

Agreed, 3x, 5x, 7x and 9x is easier, Fourier, but a theoretically perfect 5x multiplier will increase the phase noise by at least 5x. If I needed a low phase noise 10MHz source I wouldn't even consider PLLs, multipliers or mixers, I would just buy a 10MHz oscillator with the required phase noise spec or build one, a Schottky clamped Pierce is a good place to start. Why is phase noise so important anyway it's not like you've got cosited transmitters and receivers, if it's just a standard for frequency measurement then phase noise doesn't matter, it will average out.

[katzohki]

If you don't mind telling us, what is the end use for this 10MHz? Perhaps that would help narrow down what you really need? My guess is radio, judging by your user name.

Sorry to jump on the eBay bandwagon, but here is the results of mine in the image. Also tested with spectrum analyzer. There are a lot of decent shops (sure, avoid China if you're nervous) in the US and more recently Israel. It was an Oscilloquartz. They look worse for wear, but the OCXOs tend to hold up to some abuse pretty well from my experience.

Edit:
To add, many of these units have tuning inputs available and so even if it is a little off on receipt, you should be able to finely tune it in. Depending on the accuracy of your equipment, of course.

[bingo600]

I'd try a STM32 as mentioned earlier in this thread , a cheap tssop20 F0'

PLL: Divide by 13 , Mul by 10 --> SysClock 20Mhz
Set up a timer for divide by 2 , and you got a 10Mhz square.




You would get some jitter , and maybe temperature variations.
But it's cheap , and maybe even fun.

You'd have to adjust the 5v out of the pletronics to 3.3v before feeding it to the STM.

The real way us a 10Mhz OCXO as suggested by others here.

/Bingo

Ps: I have 5 of those Pletronics 26Mhz , and they have nice specs (even efc) - see J.Beales notes attached

[dannyf]

I recently looked at a MCU that has a fractional pll. I don't recall what it is now, unfortunately.

It was PIC32MX1xx/2xx's refernce oscillator. It allows a divider with a M/1024 fraction -> not exactly useful here but may be for other frequencies.

[German_EE]

Feed your OXCO output into a two way power divider:

Output 1 is divided by 2 to give 13 MHz

Output 2 is divided by 13 to give 2 MHz then by 2 to give 1 MHz

The 13 MHz signal is mixed with the 2 MHz signal to give 11 MHz and then passed through a 11 MHz band-pass filter. This signal is then mixed with the 1 MHz signal to give 10 MHz and passed through a second band pass filter to give the signal you want.

However........................

The output will be of very low quality even if you use the best filters available.

A DDS is not an option since 10 MHz is very close to the 13 MHz Nyquist limit when using a 26 MHz clock

[rstofer]

ICS525R-02 but I leave the calculation to others.  There used to be online calculators
http://www.mouser.com/ProductDetail/IDT/525R-02ILF/?qs=SmUuHNCnblrPufgFV3bBXw%3D%3D

I'm not sure this will work but i remember using it on a Spartan II FPGA to derive the baud rate  clock.

[wb0gaz]

OK, I'm convinced that using 26 MHz OXCO (it is a part that is/was used in GSM cellular base stations which for whatever reason have 13 MHz as a core frequency) is not suitable for my application, given that I want to preserve spectral purity of the OCXO (it will be used as a non-disciplined timebase for some RF test gear). Complexity of the numerous alternatives seem inevitably likely to degrade phase noise significantly, which is not desirable for this application.

I'll pursue 10 MHz OXCO alternatives.

Thanks for all of the discussion!

Dave

[CJay]

A DDS is not an option since 10 MHz is very close to the 13 MHz Nyquist limit when using a 26 MHz clock

unless of course the DDS has an internal multiplier, the AD9851 for instances has a 6x multiplier which would make the Nyquist limit 78MHz.

But of course, it'd still need filtering and it'd not be anywhere near as clean as the original OCXO.

[German_EE]

Just figured out a way to do it:

1) A 10 MHz VCO is divided down to 1 Hz and used as the timebase for a frequency counter.

2) The input of the counter is the 26 MHz OCXO

3) If the VCO is on frequency it will count to 26,000,000 or 18CBA80 Hex, if the VCO is low in frequency the count will be higher and if the VCO is high then the count will be low.

4) Feed the counter output into a CPU that has a built in DAC to do the math, feed the DAC into the VCO

[hli]

TI has some nice PLL Clock Synthesizer / Multiplier / Dividers, for example the CDCE706. They run their internal PLL with a n:m divider from the input frequency, which is what you seem to need.

[alank2]

So I received the two 10 Mhz ocxo's I ordered from eBay and they are fantastic.  They were quite a bit larger than I thought they would be, I guess when they wrote "cent." on the diagram they mean center not centimeter.  All measurements are in inches!  Anyway, really nice part.  Starts out using 275mA @ 5V and eventually drops to 140-150mA.  The metal screw on it is just a cover - removing it exposes a tiny internal multiturn screw to adjust which is quite fine.  I connected my Rb output to the trigger input on my scope and was able to adjust it so the wave would barely move left or right.  It takes about 8 minutes for it to warm up, but once it does, it is very nice.  BTW, don't bother trying to offer anything less than $15 on it.  I tried $10 and was ignored for 2 days and then $12 and was ignored for 2 days even after sending a message.  I don't get the idea that they are too familiar with eBay and how to use it perhaps.  But...it will take an offer of $15 immediately which is half price.  I ordered 3 more today so I am good to go on 10 MHz ocxo's.

[edavid]

So I received the two 10 Mhz ocxo's I ordered from eBay and they are fantastic.  They were quite a bit larger than I thought they would be, I guess when they wrote "cent." on the diagram they mean center not centimeter.  All measurements are in inches!  Anyway, really nice part.  Starts out using 275mA @ 5V and eventually drops to 140-150mA.  The metal screw on it is just a cover - removing it exposes a tiny internal multiturn screw to adjust which is quite fine.  I connected my Rb output to the trigger input on my scope and was able to adjust it so the wave would barely move left or right.  It takes about 8 minutes for it to warm up, but once it does, it is very nice.  BTW, don't bother trying to offer anything less than $15 on it.  I tried $10 and was ignored for 2 days and then $12 and was ignored for 2 days even after sending a message.  I don't get the idea that they are too familiar with eBay and how to use it perhaps.  But...it will take an offer of $15 immediately which is half price.  I ordered 3 more today so I am good to go on 10 MHz ocxo's.

He is talking about these new OCXOs being sold by Isotemp Research:

http://www.ebay.com/itm/Isotemp-OCXO-23-11-10MHz-Sine-Wave-Oscillator-5V-/171831243622

Great deal for people who don't want to take a chance on used parts (but note that there is no warranty beyond eBay's buyer protection).

Also, it doesn't have a VFC input, so you can't use it for e.g. a GPSDO.

[alank2]

Also, it doesn't have a VFC input, so you can't use it for e.g. a GPSDO.

This is correct, there is no steering it other than the multiturn pot.

[CJay]

Just figured out a way to do it:

1) A 10 MHz VCO is divided down to 1 Hz and used as the timebase for a frequency counter.

2) The input of the counter is the 26 MHz OCXO

3) If the VCO is on frequency it will count to 26,000,000 or 18CBA80 Hex, if the VCO is low in frequency the count will be higher and if the VCO is high then the count will be low.

4) Feed the counter output into a CPU that has a built in DAC to do the math, feed the DAC into the VCO

Similar concept to this one:

http://g4oep.netii.net/gsppll/gpspll.html