Author Topic: 1GHz clock source  (Read 19412 times)

0 Members and 1 Guest are viewing this topic.

Offline phenolTopic starter

  • Regular Contributor
  • *
  • Posts: 146
1GHz clock source
« on: April 06, 2015, 08:37:21 pm »
Hi
I needed a 1GHz clock source for the AD9910 eval board as I didn't feel like using the on-board 25MHz xtal+pll multiplier.
I used a 100MHz OCXO feeding an AHC nand gate (used as a squarer), which in turn feeds a 515MHz (tuned down to 500MHz) toko helical filter, followed by a GALI74 amp. The amplified 500MHz signal is fed to a schottky doubler, a 1GHz helical filter and finally another GALI74. It puts out 10.5dBm at 1gig and that's with attenuation pads between stages... I couldn't be bothered with actually making a proper pcb, so i just used double sided copper clad and cut the islands with a sharp blade.

While it seems to work fine, i'd like to gain some inspiration from other designs of PLL-free multipliers

cheers
 

Offline dannyf

  • Super Contributor
  • ***
  • Posts: 8221
  • Country: 00
Re: 1GHz clock source
« Reply #1 on: April 06, 2015, 09:03:07 pm »
At that kind of speed, a wire is no longer a wire and you would need fairly specialized skills to do any design work.

I think a non-PLL precision oscillator is unlikely to exist at that kind of grequencies.
================================
https://dannyelectronics.wordpress.com/
 

Offline Howardlong

  • Super Contributor
  • ***
  • Posts: 5315
  • Country: gb
Re: 1GHz clock source
« Reply #2 on: April 06, 2015, 09:27:47 pm »
While you can use traditional techniques of discrete multipliers and filters to do this, these days the integrated PLL+VCO route is almost always the way, as there are fewer parts in a much smaller area, and there's nothing to tweak, which in a production scenario makes a lot of sense.

The mutipliers used to be a series of transistor (or mmic) gain stages followed by an LC tank usually tuned to extract the second or third harmonic at each stage.

Any reason you don't feel like the integrated PLL+VCO method?
 

Offline paulie

  • Frequent Contributor
  • **
  • !
  • Posts: 849
  • Country: us
Re: 1GHz clock source
« Reply #3 on: April 06, 2015, 09:46:31 pm »
maybe VCO/PLL gives OP the "jitters".
 

Offline Neganur

  • Supporter
  • ****
  • Posts: 1138
  • Country: fi
Re: 1GHz clock source
« Reply #4 on: April 06, 2015, 09:49:29 pm »
if you want to go completely analogue, have a look at the TEK 284 pulse gen, it has a 1 GHz sine wave gen with a 2N3478 RF transistor

I used an ADF4351 to generate a 3.5 GHz clock (for the AD9914), it's quite good, but I guess there's a limit to what kind of phase noise you can expect.
 

Offline joeqsmith

  • Super Contributor
  • ***
  • Posts: 11630
  • Country: us
Re: 1GHz clock source
« Reply #5 on: April 07, 2015, 02:26:13 am »
Hi
I needed a 1GHz clock source for the AD9910 eval board as I didn't feel like using the on-board 25MHz xtal+pll multiplier.
I used a 100MHz OCXO feeding an AHC nand gate (used as a squarer), which in turn feeds a 515MHz (tuned down to 500MHz) toko helical filter, followed by a GALI74 amp. The amplified 500MHz signal is fed to a schottky doubler, a 1GHz helical filter and finally another GALI74. It puts out 10.5dBm at 1gig and that's with attenuation pads between stages... I couldn't be bothered with actually making a proper pcb, so i just used double sided copper clad and cut the islands with a sharp blade.

While it seems to work fine, i'd like to gain some inspiration from other designs of PLL-free multipliers

cheers

PLL should do a good job.   I was looking at a few oscillators several years ago in the GHz + range.  One was quartz to comb to SAW to buffer.    Another was quartz to comb to LC to buffer.   

Offline phenolTopic starter

  • Regular Contributor
  • *
  • Posts: 146
Re: 1GHz clock source
« Reply #6 on: April 07, 2015, 04:07:11 am »
I preferred the xtal multiplier route because of phase noise. What I'm interested in are efficient comb generation techniques . Cheap single AHC/LVC gates seem to do the job reasonably well way above 500MHz.
Snap diodes is another option, but those are unobtanium. I read an Avago app note where they use some inexpensive pin diodes in passive x3/x5 arrangement. I have yet to see which configuration works better-the basic cmos squarer or the pin diode x5
 

Offline Rupunzell

  • Frequent Contributor
  • **
  • Posts: 349
Re: 1GHz clock source
« Reply #7 on: April 07, 2015, 06:26:26 am »
Contact the folks at Aeroflex:
http://ams.aeroflex.com/metelics/micro-metelics-prods-SRDs-chips.cfm

Snap diodes are still used in comb generators and frequency multipliers today.

Here is a 1964 hp Journal article about snap diodes and frequency multipliers. App notes alone will not yield good results. To design and used these diodes properly requires knowledge of microwave design, practices and construction techniques.
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1964-12.pdf

EH Research made pulse generators using snap diodes back in the 1960's. These had sub nanosecond rise time with volts of peak to peak output into 50 ohms.


Bernice
 



Snap diodes is another option, but those are unobtanium.
« Last Edit: April 07, 2015, 06:30:55 am by Rupunzell »
 

Offline Neganur

  • Supporter
  • ****
  • Posts: 1138
  • Country: fi
Re: 1GHz clock source
« Reply #8 on: April 07, 2015, 06:33:03 am »
I'm confused, did you ask for an oscillator source for the AD9910 or comb generator?
there's also 1 GHz oscillators on digikey
 

Offline awallin

  • Frequent Contributor
  • **
  • Posts: 694
Re: 1GHz clock source
« Reply #9 on: April 07, 2015, 06:40:36 am »
FWIW with the AD9912 DDS we used an ADF4350 eval-board (PLL+VCO in one chip) and the DDS-output does look better with an external 1 GHz clock compared to the internal PLL+VCO on the DDS itself.

Let us know if you do a comparison between DDS output with an ADF4350 PLL+VCO like clock-source and a PLL-free multiplier!

AW
 

Offline phenolTopic starter

  • Regular Contributor
  • *
  • Posts: 146
Re: 1GHz clock source
« Reply #10 on: April 07, 2015, 10:33:30 am »
Like I said, i have a 100-MHz OCXO and i wanted to multiply its output to 1GHz. The 1Ghz oscillators on digikey may be a viable option, but they lack the stability of the ocxo. There seems to be a thermally compensated option there, but its eur70 price tag is prohibitive.
I am not going to mass-produce any of those things, so the added complexity is not an issue. Low phase noise is all need.

As for snap diodes, i came across this document:
 http://cp.literature.agilent.com/litweb/pdf/5966-4998E.pdf
They use the widely available HSMP3822 pin diodes as odd order multipliers.
I may attempt to duplicate their x5 multiplier instead of using SN74AHC1G00. It has some schmitt trigger action on its inputs and I am not sure if and how that would affect its residual phase noise.

As far as PLL sources go, i am not going to use one.
 

Offline Neganur

  • Supporter
  • ****
  • Posts: 1138
  • Country: fi
Re: 1GHz clock source
« Reply #11 on: April 07, 2015, 10:47:47 am »
but its eur70 price tag is prohibitive.

fair enough :) from what I can see the phase noise of those things is not uber good either (-91 dBc at 1 kHz)
 

Offline phenolTopic starter

  • Regular Contributor
  • *
  • Posts: 146
Re: 1GHz clock source
« Reply #12 on: April 07, 2015, 11:00:47 am »
well, they are probably pll-based.

but its eur70 price tag is prohibitive.

fair enough :) from what I can see the phase noise of those things is not uber good either (-91 dBc at 1 kHz)
 

Offline Howardlong

  • Super Contributor
  • ***
  • Posts: 5315
  • Country: gb
Re: 1GHz clock source
« Reply #13 on: April 07, 2015, 03:00:36 pm »
As a matter of interest, what are your phase noise/jitter requirements?
 

Offline LaserSteve

  • Super Contributor
  • ***
  • Posts: 1270
  • Country: us
Re: 1GHz clock source
« Reply #14 on: April 07, 2015, 03:35:38 pm »
If your looking for an alternative multiplier structure, take a look at W1GHZ PCB multipliers..

The LO Board for the 1296 transverter runs  at 1152 (I'm building a pair now...)

http://www.w1ghz.org/small_proj/small_proj.htm

Otherwise you could use the sapphire oscillator in a microwave "brick" and one of the many phaselocking schemes used for ham radio microwave.

The driver stage in the brick usually runs around 900 Mhz to 1 Ghz....

http://www.ke5fx.com/brick/brick.htm


Steve

« Last Edit: April 07, 2015, 03:57:18 pm by LaserSteve »
"What the devil kind of Engineer are thou, that canst not slay a hedgehog with your naked arse?"
 

Offline phenolTopic starter

  • Regular Contributor
  • *
  • Posts: 146
Re: 1GHz clock source
« Reply #15 on: April 07, 2015, 04:33:46 pm »
As a matter of interest, what are your phase noise/jitter requirements?

better than -135dBc/Hz at 1k offset and 100MHz
 

Offline phenolTopic starter

  • Regular Contributor
  • *
  • Posts: 146
Re: 1GHz clock source
« Reply #16 on: April 07, 2015, 04:56:58 pm »
Thanks for the links.

w1ghz has an  interesting doubler design built around a dedicated MCL doubler brick. I was looking into such an option, but I finally did it with a dual schottky diode and a small balun at a tiny fraction of the price of that doubler. The helical filter immediately after the doubler does a good job rejecting the out of band garbage, especially the 500MHz fundamental.

If your looking for an alternative multiplier structure, take a look at W1GHZ PCB multipliers..

The LO Board for the 1296 transverter runs  at 1152 (I'm building a pair now...)

http://www.w1ghz.org/small_proj/small_proj.htm

Otherwise you could use the sapphire oscillator in a microwave "brick" and one of the many phaselocking schemes used for ham radio microwave.

The driver stage in the brick usually runs around 900 Mhz to 1 Ghz....

http://www.ke5fx.com/brick/brick.htm


Steve
 

Offline phenolTopic starter

  • Regular Contributor
  • *
  • Posts: 146
Re: 1GHz clock source
« Reply #17 on: September 11, 2016, 12:07:12 pm »
the topic is rather old, but I'm going to keep it active for the time being, as I found some cheap step recovery diodes (Macom) with which I'm planning to build another x10 multiplier, but this time a single stage. The one i have now is x5x2 with interstage helical filters. direct x10 means that I'd have to use more selective filtering at the output, probably copper tubing coax cavity filter. Let's see what happens...

 

Offline eb4fbz

  • Regular Contributor
  • *
  • Posts: 178
  • Country: es
Re: 1GHz clock source
« Reply #18 on: September 11, 2016, 01:48:49 pm »
There are some narrow band low noise VCOs for 1GHz sources, for example Crystek CVCO55CX-1000-1000 or CVCSO-914-1000. You can build a synthesizer with a low noise integer PLL from Analog and phase noise at 1GHz could be very low if you design the loop filter correctly.
 
The following users thanked this post: evb149

Offline phenolTopic starter

  • Regular Contributor
  • *
  • Posts: 146
Re: 1GHz clock source
« Reply #19 on: September 11, 2016, 04:41:32 pm »
Those look promising, eliminating the need for bulky filters,but.... they're worth their weight in gold.
I've been looking for SAW alternatives to 1GHz cavity or helical filters without any success.
 

Offline phenolTopic starter

  • Regular Contributor
  • *
  • Posts: 146
Re: 1GHz clock source
« Reply #20 on: September 11, 2016, 05:18:15 pm »
ma144769-287T, got it from Mouser.
I guess alternative HP or USSR (2D524A,2D528A...) parts can still be found here and there. I do also have the PIN diode mentioned in that Agilent app note, but I haven't done anything with it yet.
« Last Edit: September 11, 2016, 05:23:16 pm by phenol »
 
The following users thanked this post: evb149

Offline cncjerry

  • Supporter
  • ****
  • Posts: 1269
Re: 1GHz clock source
« Reply #21 on: September 11, 2016, 07:32:02 pm »
I needed a Ghz source to set the master oscillator on a Spectrum Analyzer.  My approach is to use the harmonic of a 100Mhz oven controlled oscillator filtered with a cavity filter I already have and then amplified.

I was also thinking of just using the harmonics of my GPSDO, filtering and amplifying as well.
 

Online Berni

  • Super Contributor
  • ***
  • Posts: 4922
  • Country: si
Re: 1GHz clock source
« Reply #22 on: September 12, 2016, 05:39:27 am »
I would still love to see some photos of your dead bug prototype 1GHz source. Don't worry if it looks messy.
 

Offline phenolTopic starter

  • Regular Contributor
  • *
  • Posts: 146
Re: 1GHz clock source
« Reply #23 on: September 12, 2016, 02:38:04 pm »
it's not exactly dead bug style, i carved the islands with a knife, drilled many holes and did via stitching, literally, with thin silvered wire. In addition to vias, there is also copper tape running around the perimeter of the board connecting top and bottom ground.
The helical filters are on the opposite side. The single shield plate is there to keep the amplified 500-MHz signal from coupling onto the input of the final amplifier. Ideally, each stage should be confined into its own shielded compartment.
The doubler xfrm uses a binocular ferrite and only a single loop of braided copper wire (3 interwoven enameled wires). I am not sure how symmetric this arrangement really is and how well it suppresses the fundamental. There are small attenuation pads here and there for better matching (helical filter response) and amplifier stability.

Because I only found 515MHz triple helical filters, I had to tweak them down somewhat using the tracking gen of the specan.



I would still love to see some photos of your dead bug prototype 1GHz source. Don't worry if it looks messy.
 

Offline Howardlong

  • Super Contributor
  • ***
  • Posts: 5315
  • Country: gb
Re: 1GHz clock source
« Reply #24 on: September 12, 2016, 06:10:57 pm »
Beauty is in the eye of the beholder!

Thanks for the description and annotationed photo.

How well does it perform compared to the earlier specs?

Nice job.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf