Major D is good, but we prefer the French Roast by a small margin.
-- John Larkin Highland Technology Inc www.highlandtechnology.com jlarkin at highlandtechnology dot com Precision electronic instrumentation
Is that where they use a wood fire to cook Joanie dee zap?
So what do you think went wrong? Motivation, maybe, or is it weed?
I read Gardner like a novel when I had to design my first PLL, whose residual FM specification was (unbeknownst to me) insanely hard. (5 Hz RMS in 5-100 Hz BW around a 14 GHz carrier, after being multiplied by
120x, and tunable in 1-MHz increments at 14 GHz (8.333... kHz at my output frequency.) Talk about being chucked in the deep end, me and my astronomy degree. ;) Dr. Johnson said something like "Nothing concentrates a man's mind so wonderfully as knowing he is to be hanged in the morning."That's one place where I honestly can never see the difficulty. "Centroid" is a perfectly well defined, simple concept, easily computed, and "common centroid" means that the centroids of two or more sets of regions coincide. Do people get confused by other sorts of symmetry, or something like that?
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
Sounds like you might be pushing the state of the art. How did you measure the resulting phase noise? What reference oscillator did you use? What is the best way to multiply by 120? 5*2*3*2*2? Any chance you could release the schematics?
One way is to build two of them and mix the outputs... assuming they don't lock to one another or share anything that makes phase noise.
-- John Larkin Highland Technology, Inc jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
There are schemes for "higher order" compensation of gradients, i.e. more than a linear gradient.
This paper has a pile of layout schemes:
You can get a lot of interconnect capacitance with some of these fancy schemes, so you may compensate for a DC effect and create AC problems.
I've seen schemes where they try to use different orientations of the devices to compensate for some mask creation issues.
There are also knock down drag-'em fights over where to put the most critical diff pair regarding the least physical stress on the lattice.
You would have to work for a company with money to burn to really prove every old wives tale regarding fancy layout compensation schemes.
You don't get much arguments over dummy devices, other than the added capacitance.
Yes, the HP E5500B, 3048A and others use this method, as well as most of the industry. The problem is the specs Phil was talking about pretty much require cross-correlation to measure, which requires three identical oscillators and takes a long time to measure the close-in phase noise.
The problem with long measurements is small drift in the setup can degrade the results. Also, as you mentioned, the oscillators have to be extremely well shielded and isolated from each other to prevent injection locking.
For an example of measurements with the E5500B, Tom van Baak at Leapsecond has posted some results with fairly good reference oscillators at
You can see none of them would meet Phil's requirements after multiplying to 14GHz, and after you go through a PLL the results would be even worse.
Which led to my questions.
Well, this was 1981-83. ;) I do have the schematics someplace, but not here in the lab, I don't think. It was called the Timing and Frequency Unit (TFU) of the AEL Microtel "Spacetel" system, which was the first civilian direct-broadcast satellite system. You could get voice and data anyplace in the world where you could see a geosync satellite.
The reference was an HP 10811A 10-MHz OCXO, designed by the inimitable Rick Karlquist, whom I know slightly from a side gig.
We measured the phase noise by building two of them, running them at slightly different frequencies to avoid any injection locking funnies, mixing down to a few megahertz, and looking on a spectrum analyzer.
ISTR that the multiplier was a tripler and a quadrupler in cascade, followed by a step-recovery multiplier to get the last factor of 10. (That part was done by another outfit whose name I forget.)
After spending quite a long time convincing myself that there was no way of doing it with an 8.333... kHz comparison frequency, especially with the CMOS LSI synthesizers of the day (e.g. I tried the MC145152), I built a strange sort of fractional-N loop that had an 833 kHz comparison frequency.
IIRC it used a Colpitts crystal oscillator made of a Motorola MM8006 NPN and a 2N5109 buffer, a Motorola MC12013 10/11 prescaler controlled by two cascaded 74S190 synchronous presettable decade counters,a 74S85 magnitude comparator, and a 74F74 DFF for resynchronizing.
That got me the three top decades, with an output frequency of 833.333 kHz. The bottom two decades, I got by running two CD4527 CMOS BCD rate multipliers in cascade, which produced 0-99 pulses per 100 input pulses, and using those to control the CARRY IN pin of the lower-order 74S190. That way I could nudge the (average) comparison frequency around without having to divide down to 8.33 KHz and suffer a 40 dB phase noise increase.
The phase detector was a MiniCircuits RPD-1, which gave me about 2V p-p output iirc, i.e. K_phi = 1 V/radian, which was great. I came up with a pretty slick way of acquiring lock, too--I just put a bit of positive feedback around the integrating loop filter, so that when the loop was unlocked, it swept slowly back and forth across its whole tuning range. It turned out that this had been invented by somebody else a few years earlier, but I still think it's a neat hack.
The rate multiplier jitter was much worse than what you'd get from a phase accumulator system such as a DDS or a Digiphase, but it was mostly outside the loop bandwidth, so it didn't matter too much. Especially, it was all at multiples of 8.33 kHz, so there was none whatsoever below
100 Hz offset.A nice feature was that you just set the desired output frequency (at 14 GHz) on the rotary DIP switches, plugged in the right crystal, and off you went.
Back in the '80s, at least at AEL Microtel (where this was) it was SOP to switch crystals to switch channels, so nobody minded this too much (except me).
So that's why I was inhaling Gardner's book at such length. I should also credit Don Lancaster, because it was in his TTL and CMOS Cookbooks that I found stuff like rate multipliers, which I would never have bothered reading about from databooks.
(I also did the Pilot Tone Generator, which transmitted the frequency reference to the remote sites, and learned in great detail why you don't use D flipflops as mixers if you care about jitter. The metastability was a nightmare.)
That was one of my most fun jobs ever, but I left after a couple of years to go to grad school.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
The "small drift" is also phase noise, so measure that too.
The telecom industry separates "noise" from "wander", the cut point being 0.1 second.
-- John Larkin Highland Technology, Inc jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
I think this is the SRS clone...
Pretty good, SC-cut rock, weird thermal design.
-- John Larkin Highland Technology, Inc jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Yes, I have several. Phase noise is not too good. I wouldn't use them to multiply to 14GHz.
[snip interesting description]OK, your description was so good there's no need for schematics.
Thanks very much.
Interesting, thanks.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
I'm not sure what they used in production--something purchased, anyway. For our purposes they were fine, because with two nearly identical synthesizers, almost all the reference phase noise contribution went away when they were mixed down.
Thanks for asking--kicking around old projects is fun.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
I think the patents are 5,708,394 for the bridge oscillator and 5,729,181 for the thermal design. The SC-10 uses a standard Colpitts oscillator and what appear to be a conventional thermal design, so it doesn't seem to be any kind of clone.
You can see Rick's patent list at
Sorry, I was thinking of Rick's hockey puck design which has excellent thermal performance, but the bridge oscillator is not as good as some conventional OCXO's. I tried to find the HP part number for the hockey puck but it may not have been used in any HP products. The ones I have are in a filing cabinet behind two 10 ton hydrauic presses so they are kind of hard to get at, otherwise I would post the HP part number. Too bad he didn't use a conventional Colpitts, which would have made an outstanding OCXO.
The HP 10811A and 10544A are pretty good oscillators and should perform adequately when multiplied to 14GHz. For a teardown of the 10811A, see
Sorry, your thermal reference led me to believe you were thinking of Rick's hockey puck design. It definitely is an interesting approach and may be worth studying for other applications.
But the HP 10811 and 10544A are fairly conventional OCXO's and the SC-10 may well be a clone as you state.
Just for reference, the hockey puck OCXO is the HP E1938A. Tom van Baak states, "The HP E1938A is a very modern ovenized SC-cut quartz oscillator of radical design that had little or no product release due to project cancellation. As far as I know it was the first new oscillator designed by HP since the venerable 10811A. The performance of the E1938A is truly outstanding."
Further down on his web page, he states, "The performance of this E1938A is ok but not stellar...", so it's not clear if he got a bad one or the design is not very good. Other references say the phase noise is poor, which is understandable from the bridge design compared to a standard Colpitts.
They are still available on eBay:
Bridge oscillators ideally null out the phase noise contribution of the active device, or so I'm told. (That's one of those pebbles in my shoes that I have to figure out some day soon.)
The amplitude limiting mechanism is one of the huge contributors to oscillator performance--AGC oscillators such as (JT's) MC1648 are a lot better than normal Colpitts ones like mine in the TFU, which rely on transistor cutoff for amplitude stability.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
Take a look at "Getting Started in Electronics" by Forrest Mims. Rat Shock used to sell it, but now you have to order it online. It has very little math beyond Ohm's Law. Lots of 1N4148, 2N2222, 2N3904, caps, resistors, LEDs, relays, real simple 7400 TTL and 4000 CMOS circuits, things like that. (This book is for your EfS course, not for your new- hire MSEE.)
Matt Roberds
You can control the amplitude by controlling the current into the oscillator. Rohde shows this in Fig. 13, "Recent Advances In Linear Vco Calculations...",
"
and numerous other articles. Bruce Griffith shows another approach using a PNP to control the bias such as Q2 in "Development of a low noise high reverse isolation low distortion RF Amplifier",
Another approach is to simply drive the emitter of a Colpitts using a current source or a suitable value of resistance. The resistor is the most common approach, but it can lower the Q of the tank and has no feedback to reduce flicker noise.
The 1648 is a Butler, which switches at the zero crossings. This is the worst place to switch for phase noise. Hajimiri illustrates this in Fig.
4 of "A General Theory of Phase Noise in Electrical Oscillators", inThe Colpitts switches at the peak of the sine wave, which is the best place for phase noise. This is why the best oscillators for phase noise are invariably Colpitts (or Clapp to be technical.) The Butler configuration rarely or never appears in a high performance oscillator.
Leif Asbrink, SM5BSZ, shows a different approach that gives phase noise similar to the best oscillators by Wenzel Associates, down to -180dBc/Hz at 20 kHz using the cheapest possible standard crystals at 14 MHz. He uses back-to-back 1N4148's as limiters:
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.