Frequency synthesis, and a variable attenuator for up to 200MHz

The Mota chips were okay, but pretty specialized. Back in the day I tried using them a few times, but they never quite fit the bill.

DDS will give you much lower phase noise, if you pick the right crystal oscillator. (Some packaged XOs, especially the programmable kind, are actually RC VCOs with an internal frequency synthesizer. Don't use one of those for a frequency reference.)

If you use a DDS with an internal DAC, which AFAIK they all have, you don't have a lot of harmonics to deal with at all, at least not compared to most oscillators.

That's not an antique method. It's by far the most accurate way to do it.

Or just hit the RF OFF relay while the attenuator settles.

MDACs are limited to audio or low MHz frequencies. Check out the attenuation vs. frequency plot in any MDAC datasheet you like.

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

Will not get you to 200 mhz, but have you seen the cheap dds modules on Ebay.

Ebay item no. 131120372446

Dan

You might also search on Ebay for AD9854

That and a frequency doubler might work.

Dan

Phil Hobbs wrote

There is an ADF4116 with an internal VCO, 50-400MHz, which looks a lot better. $3 or so...

I thought that if e.g. you use a DDS chip to generate say a 100MHz carrier, the DAC is fed with one data word every 10ns, so you will get loads of all the harmonics. The way to get the harmonics way down is to oversample e.g. drive the DAC at 1GHz. That makes filtering much easier. But maybe I misunderstand.

Yes; I can see why. I guess MOSFETs are the best way for VHF...

Yes - if the application can tolerate it.

DDS's do over-sample. The DAC is up-dated at the DDS clock frequency, and w hat come out is the spot-value of the sine-wave at that instant.

As you can imagine, the approximation to the sine-wave gets better an bette r as the as you cram more steps into the stair-case approximation.

If you want to get analytical about it, the error waveform is a sawtooth, s o there are no harmonics below the DDS clock rate (if the DAC is perfect, a nd 14-bit DACs are common - if you pay enough for the DDS) and the higher f requency harmonic amplitude goes down in proportion to the harmonic number.

t.

If the application can pay for it, have two synthesisers and switch between them as you fiddle with the attenuator of one or the other.

--
Bill Sloman, Sydney

Harry Lythall is very smart fellow. He did a pretty good job of coming up with what is a practical equivalent to the old Moto and CB synth chips, in CMOS...

tab over to his synth links, this is synthesis 80s and 90s style. His 1.2 Gig synth explains "Pulse Swallowing" style M+N synthesis.

While what he has on the page is not going to get you a good wide range, fine stepping PLL, it is interesting to read about the old way. Otherwise you need to start with DDS or one of the many fine books on the subject. Ulrich Rhode's book is good.

TI/National also has some fine PLL chips...

Steve

I sort of miss the dual-modulus days meself, but I sure wouldn't use one in a product now, even if I could get one.

You ought to be able to put a synthesizer into a CPLD, including the Motorola-style phase/frequency detector with two outputs (i.e. no dead zone).

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

It isn't "antique." It is the right way to do it.

It is pretty easy, see:

(Steps down to 0.25 dB.)

I would use the DSA's with a "bypass" for the negative voltage, because otherwise an internal charge pump causes video spurs. If you don't see the spurs, you aren't looking hard enough.

Simon S Aysdie wrote

That's brilliant - thank you. I can't find any pricing - do you have a ballpark idea?

Do you mean decouple it with capacitance?

A lot of commercial RF signal generators, expensive ones, have astonishing THD specs... like -20 dBc.

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

2?cur=USD

otherwise an internal charge pump causes video spurs. If you don't see the spurs, you aren't looking hard enough.

No. Read the data sheet: supply your own "clean" negative voltage.

Optional External Vss Control (VssEXT) For proper operation, the VssEXT control must be grounded or at the Vss vol tage specified in the Operating Ranges table. When the VssEXT control pin o n the package is grounded the switch FET's are biased with an internal low spur negative voltage generator. For applications that require the lowest p ossible spur performance, VssEXT can be applied to bypass the internal nega tive voltage generator to eliminate the spurs.

First you want to start with a spec on outputs such as phase noise and harm onic distortion to determine the best technology for obtaining the results.

For Crystal references, phase noise will be multiplied by the fractional N rates unless you have additional loops to reduce the phase noise. In the o ld days HP used high frequency stability sources such as Cesium oscillators which have high phase noise and certain Quartz resonator types for low pha se noise. THen use fractional N with pulse swallowing and inserting metho ds to achieve high resolution .

SC cut are about 1000x better than AT cut which are ten to hundred times be tter (Q)than MEMs type resonators. Don't confuse frequency stability with phase stability for each resonator type. Series mode resonators were used due to lower phase noise, but parallel mode used for lower power dissipatio n.

1ppm TCXO's are a good start and pretty cheap these days but watch out for phase noise specs.

But it all comes done to cost performance specs you need.

Put a little cover over them to keep off air currents. That will greatly improve close-in phase noise.

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation

John Larkin wrote

That THD must show up in the spectrum, surely.

-20dbc is 1% THD, if I have it right. That's atrocious. You will be radiating all over the place.

I am not yet sure what performance I need to achieve but it is something of the order of a Marconi 2020 signal generator when it comes to accuracy and stability over a few hours. This isn't going to be achieved with a 20p crystal connected across a 40106 :) I am going to have to buy a self contained oscillator.

But fortunately the frequency steps I need to achieve are about

200kHz, around the 160MHz region, and the PLL settling time is not important (100ms is OK), so I don't think there is any point in using DDS. An old fashioned PLL with an integer divider should work fine, and if I have a high quality VCO that ought to do it. Analog Devices do some nice PLLs with built-in VCOs. Or one can build a discrete VCO.

I have looked at the Mini Circuits parts and I can see they are very handy in that you get a small package but I don't immediately see an advantage against modern ICs. I used to read EDN in the 1970s/1980s and Mini Circuits used to have big adverts there all the time and back then they were a really good solution for non cost sensitive projects where you didn't want to mess around with RF in too much detail.

He did say they were astonishing :P

Bit of a shame if you want to work with 3.2V though, as I do.

Haven't got my VCO with AGC and microprocessor FLL working yet at

144MHz. I think I need faster transistors than LM3046, and suspect HFA3046 is too fast - might go discrete. It also uses an MSP430 to FLL from the 32KHz xtal, and at 144MHz that needs an MC12080 pre-scaler. I'll be sure to crow here when I get it working.

Clifford Heath.