50 kHz VCO w/sine output

Same thing, isn't it?

-- Les Cargill

for

The first mention of the AD9850 I can find here dates back to 2000, so it has been around for more than ten years already - always a good sign - and I think that it had been around for a while then. IIRR it was the first of Analog Devices DDS chips, and it seems to have been designed into a lot of applications, so it may be around for a while yet.

-- Bill Sloman, Nijmegen

e:

he

e with

the

e.

ne

ted.

I was thinking about the mixer approach for a simple network analyser gismo. In essense I need a amplitude/phase plot developed from two signals, operating from say 1kHz to 10MHz.

Use a DDS to generate a local frequency (the one that is investigated), feed that to the input of the back box. Feed a mixer with the output of the black box and another DDS with a slightly lower frequency (50kHz). Feed the output of the mixer to a high resolution low frequency aquisition device (low sample rate ADC). For each local frequency, derive the amplitude and phase difference.

Never done mixer designs, is mixers available which has predictable and stable gain? (from inputs to output). Or would a pre-calibration procedure be better?

Regards

Klaus

ote:

the

th

one with

(the

ice.

sine

ou

anted.

Hi Klaus,

Re Mixers: Typically mixers come spec=92d for conversion gain, IIP3, drive levels, impedance data and isolation from port to port. All as a function of frequency. When used as a component in an acquisition system such as a SA typically the system has some mechanism for completing self cal. That requirement is typically determined by your product accuracy specification.

Having said that, I=92m confused by your application.

You have a DUT, the black box. Driving it with a source. Take the output of the DUT and drive a mixer that uses another source that is a few KHz below the original source?

Are the souces phased locked?

Why are using the mixer? Do you have a low frequency A/D that you're doing frequency translation to?

What about just using characterized switched paths to drive a higher frequency A/D. Terminate the black box into a load, stimulate it with the DDS and look at the input and output.

I=92m probably just missing something.

Regards

Back in the hay day, we used double balanced mixers. You could generate a vary wide dynamic signal that was clean. Just use a low pass filter to remove the upper end.

They made great audio scramblers too! :)

Jamie

Or if you can, mix down a little lower so the sound card of a computer can pick it up. Some go up to 25-30kHz. But all this depends on your signal source. If it is by itself and no other noise is present, ok. But if there are other spectra then keep in mind that if, for example, you mix 1.45MHz with 1.4MHz you get a 50kHz output signal as a result. Plus

All the commercial and also the discrete mixers I have ever used are very true to their specs and stable. Ok, cheap single-transistor versions can be iffy. If you want top performance use a DBM from Mini-Circuits or similar. We have used them in ultrasound machines and there the whole algorithm would fall off the cliff if the amplitude error was larger than 0.5dB or the phase error was larger than 1% (often much less).

Best comment I ever received when I "transistorized" a contemplated massive army of ICs: "You mean that's IT?" ... "Yeah, that's it" ... "Amazing. So what's it gonna cost?" ... "About three fifty" ... "Threehundredfifty?" ... "No, three Dollars fifty" ... "WHAT?"

You can use that VCO and then mix it with 15MHz or whatever. But I don't know how you control the VCO. If via PLL, ok. If via hand and potmeter then definitely not because it'll be way too sensitive.

er.

nh to

we

at

h,

A VCO feeding a counter =3D=3D> simple decoder =3D=3D> a few resistors maki= ng a sine-weighted DAC might be a convenient way of making a sweepable break-pointed sine. A homemade DDS, as it were.

You could clock it at 256x clock rate or whatever, and get the harmonics super low with just a few breakpoints.

It has the virtues of purity, simplicity, stability, and repeatability. Easy to design. Oh, and it's cheap too. Joerg likes that.

-- Cheers, James Arthur

That chip has been around for 10 years. If you are designing something for long product life you can always make the DDS a module, and the software updatable to use any new part. Once you know how they work, they are fairly simple.

I used to rent the sound system to a high school for their football games. I would sit in my service truck & watch Wonder Woman on a 12" B&W TV I had modified to run on 12 volts. I had a member of their band walk out to my truck and start to tell me off because I wasn't watching the game. Then he saw what was on TV. A couple minutes later, most of the band was crowded around my truck to watch TV. :)

The contract specified that I had to be at the games, it didn't say that I had to watch them. :)

ble

ted

's the

with

done with

he

em (the

spice.

z sine

you

wanted.

if

...

ual

ome

s 6

n't

ll

nd

In essense I want to measure the gain/phase of a black box. It could be a filter, or could be a amplier, PSU etc

To measure the gain/phase plot, I will measure at a lot of discrete frequencies, up to about 10MHz (what makes sense for my application)

But, to measure phase in a simple way, I need to have the ADC capture the signal with a sample rate of at least 20-50 times higher than my signal, so that would be a very high speed ADC.

Instead, I purpose to shift the signal down instead, so I can use a cheap and LF ADC. The local oscillator is the exersitation for the filter, the second DDS is just used to shift the signal down. I am not sure if the need to be phase aligned.

Regards

Klaus

Oh, right, the conservation of energy thing.

John

Hi MeO

Hmm, this is a small part of a new project. I'm still trying to get a handle on the specs I need.

e, not

er your

lot

Well, an RF amplifier design ~20 years ago. (would have been nice to have a spectrum analyzer at the time, I had some ring down issues in one of the amp stages.) The only RF recently is a few watt oscillator for a Rubidium discharge lamp. (Where frequency stability is not much of an issue.) I've got a 'new' SRS770 spectrum analyzer, (DC-100kHz) which should tell me all I need to know about what's happening after the mixer.

to

when

Hmm a bit of padding should take care of that... I'll just burn power!

Thanks, I'd like to do a DDS design sometime, but right now that looks like a longer learning curve/ time sink. (strange as that may sound.)

Thanks again. At the moment I'm just going to order some VCO's from minicirciuts... (throw some money at the problem.) and see what the output looks like. I'll also push forward on the triangle wave/ sine shaper idea.

This is a nice app note,

George H.

ide

e for

be

Thanks Bill, that's good to know.

George H.

I wouldn't be so sure. A client was trying to find a noise/instability issue with one of those SRS analyzers. Couldn't, so they called me in. I couldn't find it either so I whipped out my trusty Durabook laptop. It has a fairly low-key sound chip, IIRC only 18 bits. Lo and behold, _that_ found it. Some jaws dropped.

Decouple the power rails well, that usually takes care of unwanted injection locking effects.

Only if you don't have to adjust the VCO by hand. Because that would get iffy.

George,

You may have already achieved a useful solution, but I did run across a numnber of potential models using the following search string:

Varactor SPICE Models for RF VCO Applications - Skyworks

Spice Models for Aeroflex / Metelics GaAs Hyperabrupt Varactor

and Infineon has some spice models for (e.g.) its BB535 varactor. Got to:

http://

Hope one of these helps.

Frank McKenney

Note that if you _do_ have mild injection locking problems they'll show up as harmonic distortion in your sine wave long before they show up as your VCO snapping to the reference frequency. Good decoupling will help lots, dividing your VCO frequency down by some convenient factor may help a lot more (and if you're careful about your transformers, you can get

I think that me0 is overstating the case -- but that depends a lot on just what you need, and you say you don't know the specifications exactly, so...

It's arguable that, for a VCO, you want to know if/how it'll be phase-locked (i.e. it's the sensitivity of the intended use, not the energy in the total harmonics, that will matter). The little corner of the remnant of the triangle wave might be a lot less important than other measures of sine compliance, or might be MORE important.

tside

it

ces

ble for

l be

George,

What kind of tuning resolution do you need?