Final Output Spectrum
Observing the frequency spectrum of the final output reveals that you can not get something for nothing and helps us to understand when the frequency aligned mode should and should not be used. I have used the infinite persistence display again and this time it was even more useful to do so. As before, plots on the left cover up to up to 50MHz and on plots on the right show ±5MHz centred on 12.5MHz.
3rd
12.5MHz Harmonic Fundamental
12.4125MHz Fundamental
These plots show that the 12.5MHz signal is actually not as good as that generated directly at the output of the phase accumulator. Although clearly centred at 12.5MHz the spectrum shows that there is an increased bandwidth. This reflects that the DCM is tracking the input frequency even though it doesn’t really need to do anything. It is rather like balancing on a wall; we know the wall isn’t moving but we still wobble a bit to stay balanced because we are unable to freeze completely due to other influences on us and the need to breath etc!
5MHz/division |
| 1MHz/division |
10dB/division |
| 10dB/division |
|
|
|
The full spectrum of the 12.4125MHz case shows how the previous ‘family of spectral components’ associated with the 5ns of cycle jitter have been removed and that the noise floor has been returned to normal levels. The zoomed plot now shows a fundamental with what looks like modulation sidebands rather than fixed spectral components at ±0.74MHz. This again reflects average frequency tracking as well as the removal of the 5ns cycle jitter. Note that an agile frequency component has a lower energy (W/Hz) than a static component.
12.5MHz Fundamental
12.4125MHz Fundamental
±1.3MHz
Hint – If you can synthesize a perfect waveform with a phase accumulator or other direct clock division circuit then it does NOT make sense to use a DCM in frequency aligned mode. In all non integer division cases the DCM will dramatically help jitter performance but some frequency ‘tracking’ must be accepted.
Frequency Generator for the