ee books

I found it - right in front of my nose. I was searching for "SRD", and you wrote "step recovery diode". My search will always fail.

OK that part finally sunk in. When you change frequency, the loop breaks lock. Your trick to use positive feedback forces the loop to scan through the entire range of the vco until it finds the lock again. It settles on a quadrature relationship at the new frequency and you have satisfied the requirement. The process is fully automatic and only requires a bit of patience.

Neat trick!

The rate multiplier pulses steal 1/100 cycle each, so the lock point only moves 3.6 degrees per time. The RMS error is only a degree or two, so the sinusoidal phase detector works fine.

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

Funny you didn't tell him to come onto this news group. It's easy - you pose a question, get 10 answers, and choose the one you like best! j

Thanks for the explanation. I guess I don't have enough pieces to understand the system completely, but it is a pretty impressive accomplishment for the time.

Here's a short list.

So, maybe a recommendation for _Electrons_and_Holes_in_Semiconductors_ (by William Shockley) is in order here... gonna be expensive to find that book, though.

There's a lot of good stuff in the older literature: _Radiotron_Designer's_ Handbook_ (4ed) is the only place I've found T-pad and L-pad attenuators discussed. For classic 'tricky' things (review Foster-Seely detectors lately?) you gotta have a few dusty old-school books.

Thanks. The 10/11 chip and the two decade counters work exactly like an N/N+1 prescaler, with N ~ 135 or so. When the carry output is active, it divides by 134 instead of 135, so the period of the output is slightly shorter, causing a few-degree phase transient but no extra pulses.

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

$800.00

BSTJ has what appears to be an earlier version that discusses the theory of PN junctions, 24MB:

Another way to do that is the old slidewire: get a meter stick and a spool of music wire. Or, thermocouple wire (type T and K use resistance-wire Manganin and Chromel). It's available prebuilt, too

but that doesn't look like any fun.

If on one else has already suggested it, how about a Dilbert Annual?

Thanks - that helps a lot. Now that I know what the output is supposed to be, I can go back and fill in the missing pieces so it will work.

The problem with Fractional_N synthesis is the dreaded sidebands that are so hard to remove once they have been created. There are a number of solutions:

1. Divide and Mix: Rick Karlquist illustrates this method in "A Narrow- Band High Resolution Synthesizer Using a Direct Digital Synthesizer Followed By Repeated Mixing and Dividing", at

1. Use the DDS as integer divider: The SRS SG-380 series signal generators offset the reference a small amount so the DDS operates in integer mode which eliminates the spurs. This is called "Rational Approximation":

The signal generator displays sixteen digits, which is pure marketing fluff:

1. Multiple PLL Loops: The HP Triple loop is an example of using multiple loops to give fine frequency resolution:

1. Diophantine Equations: Diophantine synthesis is an interesting new approach to synthesis that eliminates the sidebands caused by DDS and Fractional-N and gives fine frequency resolution. An example is by Sotiriadis, "Diophantine Frequency Synthesizer Design for Timekeeping Systems":

He has written many other articles that are worth checking if you are interested.

Now all I need is to get someone as smart as you interested in this technique to show me how to implement it!

That's the traditional direct-synthesis method, where each decade is an identical module. AIUI it tends to have a high white noise floor.

That's sort of what my gizmo did--it relied pretty heavily on the VCXO to get low phase noise outside the loop bandwidth. You could look at the whole fractional-N thing as a way of moving the reference frequency around.

Multiple-loop synthesizers tend to have the same sort of spurious problems as double-conversion receivers, though, don't they?

That's an interesting one, 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

Yes, big problems with various harmonics mixing together, intermodulation products, spurs due to stray signal leakage from other modules, increase in the noise floor, etc. I suppose in some limited cases where only the frequency difference was needed, a simple digital mixer might be applicable. Any issues due to metastability could be eliminated by adding a second stage as a shift register, although in my experience it is virtually impossible to get the Motorola Eclinps d-flops to show metastability. I did extended work with the 100EP52 and never once saw a metastable state. But the Motorola 10H ECL series, and just about any 74X74 would go metastable very easily.

Mixers again. He mentions getting better than 100dB SFDR, but that's not much improvement over the AD9910/9912 DDS.

I kind of like the idea of running an AD9912 into an ADF4108 8GHz integer PLL to control a YIG. The nice thing about a YIG is the Q increases with frequency so the noise floor tends to remain constant with frequency, and the bandwidth is quite narrow if you only use the main rf coil.

If close-in spurs are a problem at some frequency, I can easily change the ADF4108 division to use a different reference and change the DDS to suit.

I got three 4GHz to 8GHz YIGs on eBay and am working on a wide range synthesizer for low level phase noise measurements with cross- correlation, vector network analysis, spectrum analysis and general rf work. I am especially interested in Leif Asbrink's approach to low phase noise oscillators to replace the Wenzel ULN series. They are expensive, with very slow delivery, and have extremely poor rejection of power supply noise and physical stress:

Like you, I have a HP8566 22 Ghz spectrum analyzer, but that thing is so heavy I hate to have to lug it around. Also it uses harmonic mixing which ruins the noise floor as you go up in frequency. With the AD8300 series log converters from ADI, a bit of DSP from TI and help from Hittite, I think I can do better.

I'm a big fan of YIG-tuned oscillators as well. For phase noise, the new SDR-style analyzers can't compete at all. A pity they're taking over the universe because they're so cheap.

What I mostly wish the 8566B had is a tracking generator. Sitting there accumulating sweeper measurements in MAX HOLD mode is a bore.

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

They're great if that's all you need. For more serious work, a YIG is hard to beat. Theoretically they have no spurs. Better phase noise than LC oscillators. No phase noise shelf from the pll loop. Octave bandwidth so you can divide down to any frequency and gain the advantage of lower phase noise.

About the only disadvantages I can see is they are slow (~100 Hz) if you only use the main coil. But the FM coil can give much higher bandwidth, perhaps up to 200 KHz or more. They are also sensitive to stray magnetic fields and vibration from fans so they need a clean environment.

That's why I need three YIGs. Two for the local oscillator in VNA applications, the third for the stimulus signal.

For spectrum analysis and VNA applications, the ADI AD603 and AD8307 can give a logarithmic amplifier with 130 dB dynamic range. Not quite as good as a Rohde & Schwarz, but still usable. The parts are old but still available in most distributors.

Unfortunately, the article by Barrie Gilbert that shows how to do it seems to be wiped off the web, but I can email you a copy if you want.

Some black ESD bags are also fairly conductive (I mean the black ones, not the shiny grey ones). The ones I have looked at are made of three layers: two layers of resistive plastic with an insulating layer in between. Therefore be aware of the possibility of shorting the two resistive layers together at the edges where you cut it, and who knows what happens when they heat-seal the edges to make it into a bag. The big advantage of these is that I keep getting sent them whenever I order semiconductors.

Chris

The best thing you (John) can do is use the DDS as an offset synthesizer. Any scheme that involves multiplying a DDS output is never going to be genu inely "low-noise."

Look at the block diagram of your 8566, and think about using the AD9912 in place of the 20/30 loop. Use a comb generator with a 100 MHz or higher cr ystal as the multiplied source ("M/N loop" in 8566 parlance.)

The second best strategy, if you insist on multiplying the DDS with an ADF- series chip, is to feed it through a narrow crystal filter first. I wouldn 't do it that way again because the N factor is necessarily very high. If you use a narrow filter to kill the DDS artifacts, your N factor will need to be large to accommodate the tuning range between integer steps. If you use a wide filter at a higher frequency, it won't be as effective at cleani ng up the spurs. Using a fractional-N chip could help with that compromise , of course, but now your design is more complex...

If you can find an old HP 8620/86222 sweeper outfit, you can turn it into a TG without much trouble (see

Only for the 2.5 GHz band, of course. :(

-- john, KE5FX

Cool. I actually have two of those, so I might well do that when I have a spare moment. (BTW thanks for the GPIB tools as well.)

Thanks

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

Hi John, Thanks for the post. I want you to know I'm a great fan of yours. I appreciate the comments you make on the Time-Nuts forum, and your web site is full of valuable information. Your GPIB Toolbox and TimeLab are extremely valuable contributions, and your TimePod is simply a masterpiece. I'd like to ask two favors. Do you have a description of the file format used in the HP E5500 Phase Noise Test Set? I'd like to be able to send data to your TimeLab for analysis, but I can't find anything that describes the format and any headers.

The second favor is do you have any information on how sweep generators that go to 50GHz work? I'm finding that YIGs stop around 40GHz, and I really need to go higher, perhaps up to 64 or 75GHz. A Gunn diode will provide spot frequencies, but sweepers such as the HP VNAs can operate continuously. Another example is the Gigatronix 2500B series shown here:

For the HP 8566, that is a real hairball. Rather than slogging though the maintenance manuals, there is an excellent description of the synthesizer in the August, 1979 issue of the HP Journal:

The M/N loop is described on Page 14. It takes 7 pll loops, which is really a lot. There is an article titled "Phase Noise Performance of the HP 89400 Series Vector Signal Analyzers"

Graph 1 on Page 2 shows the 8566 has the worst phase noise in a group of spectrum analyzers at 1 GHz, so I'd be more interested in looking at the other synthesizers to see how they do it.

As far as adding a crystal filter to the DDS output, the SRS SG-380 uses a crystal oscillator to clean the DDS. They end up with a bunch of low noise oscillators which would be impossible to duplicate. The circuit descripton starts on Page 135 of

The plan for now is to build the equipment in modules to get it up and running, and make several copies. Then change one module at a time and record the improvement. This will allow investigating Fractional-N synthesizers, offset loops, and various other methods. I plan on shielding and isolating everything to minimize signal leakage, and hopefully this will give a good idea where the critical points are that need to be addressed when the system is assembled in the finished configuration.

Thanks for your help.