[time-nuts] Triangle Waves

Bruce Griffiths bruce.griffiths at xtra.co.nz
Tue Feb 2 02:48:34 UTC 2010

Bob Camp wrote:
> Hi
> Ok, next up on the dual mixer stuff is checking the limiter chain. To do that with any chance of the results meaning anything you need a good triangle wave. You certainly can build some pretty complex gizmos to make them. There also appears to be a fairly simple approach.
> If I take a fairly good 16 bit DAC that will accept a clock a bit above 1 MHz, I can feed a simple count up / count down into it. That should give me a triangle wave at (clock rate) / 2^32. Simply put, 1.3 MHz data gives me a 10 Hz triangle wave. The digital crud should be almost entirely up around the clock rate or higher and>  90 db down. That assumes that the DAC is a low clock feed through version and that it's got good linearity.
> A reasonable dual mixer or heterodyne system should have some kind of low pass filter built into it. Even a 150 Hz lowpass should knock the digital stuff down into a -160 noise floor.
> The gotcha seems to be flicker noise out of the DAC. There's no guarantee that the gizmo will have a 1nV/Hz class noise floor. The same sort of audio spectrum analyzers used for phase noise should be able to measure the noise coming out under various conditions.
> The nice thing about this gizmo is that it does not have to *only* put out a triangle wave. If you drive it with a micro, you can tell it to do all sorts of things. You might try a number of DC levels as you check for noise. You might also try various triangle wave levels to see how everything matches up. Slew rate limited square waves also sound interesting.
> There are a couple of other details like DC level shifting and driving it all with a decent clock. Both need to be done properly, but they don't appear to be the limiting factors in this kind of setup.
> I suspect this approach has been tried before. Any record of it out there?
> Bob
> _______________________________________________

You arent going to find a DAC with a 1nV/rtHz noise floor off the shelf 
due to the reference noise.
Heroic filtering measures will be necessary to reduce the reference 
noise, then you will have to deal with the DAC component noise which 
will almost invariably be greater than 1nV/rtHz.
If it has an external reference capability you could try using a series 
stack of leds as in this application the reference tempco shouldnt be 
too important.
A Josephson junction stack would work as a DAC well with very low noise.
NIST uses such JJ stacks as sigma delta DACs to calibrate the Johnson 
noise thermometers.

Why can't you just use a sinewave test source?
Only the part near the zero crossings is of any importance.

Another effect to consider with diode mixers/phase detectors is that at 
10MHz the amplitude sensitivity may be as high as 256ps/dB with both 
inputs ports saturated.
Reducing the input port VSWR with a series resistor and attenuator pad 
can reduce this effect by a factor of 10 or more.


More information about the time-nuts mailing list