---+ DAN Transfer Function

Motivation

The current implementation of Digital Active Nulling (DAN) uses an integral control loop to null, i.e. cancel out the current through the amplifier. This should leave very little residual signal at the ADC.

Here, I will use G as an arbitrary gain factor.

Results

The 195 kHz DAN build seems to operate as expected. The signal is nulled out at low frequencies, but not at high frequencies (far away from the demodulator). When the gain is increased, the residual signal drops until an instability is reached, increasing the residual quickly.

Transfer Function at G = 1

Transfer Function at G = 10

Transfer Function at G = 100

Transfer Function at G = 1000

Transfer Function at G = 10000

Below here are old results from the 12 kHz DAN build


Results

I set a small carrier some frequency away from the nulling/demodulation frequency. This produces a sine wave which is to be nulled away by DAN. The following plots show the nuller amplitude, as well as the residual signal at the ADC given a signal between 3 Hz and 305 Hz (80% of Nyquist at FIR4) for various gain settings.

Transfer Function at G = 2

Transfer Function at G = 20

Digitization appears to be an issue for G = 2 and G = 20. The gain shows step-like behaviour due to the discretization of the signal in the accumulator.

Transfer Function at G = 200

Transfer Function at G = 600

At G = 200 and G = 600, DAN seems to be well behaved, nulling well at low signal frequencies and decreasing in effective gain with frequency.

Transfer Function at G = 2000

At G = 2000, DAN is unstable, which can be seen from the large residual signal at the ADC.

Conclusions/Issues

  • G should be chosen to be high enough to null over the required signal bandwidth, but low enough to avoid instabilities.
  • I don't know why the quadrature sum of the residual signal and the nuller amplitude is not constant.
  • I don't know why the measured signal increases at high frequency.

-- TijmenDeHaan - 27 May 2010

Topic attachments
I Attachment Action Size Date Who Comment
PNGpng dan_10.png manage 31.3 K 2010-09-02 - 16:36 TijmenDeHaan  
PNGpng dan_1000.png manage 23.8 K 2010-09-02 - 16:30 TijmenDeHaan  
PNGpng dan_10000.png manage 33.1 K 2010-09-02 - 16:30 TijmenDeHaan  
PNGpng dan_100png.png manage 25.8 K 2010-09-02 - 16:30 TijmenDeHaan  
PNGpng dan_1png.png manage 27.7 K 2010-09-02 - 16:36 TijmenDeHaan  
PNGpng tf_gain2000png.png manage 28.8 K 2010-05-27 - 22:00 TijmenDeHaan  
PNGpng tf_gain200png.png manage 36.8 K 2010-05-27 - 21:59 TijmenDeHaan  
PNGpng tf_gain20png.png manage 38.3 K 2010-05-27 - 21:59 TijmenDeHaan  
PNGpng tf_gain2png.png manage 32.9 K 2010-05-27 - 21:59 TijmenDeHaan  
PNGpng tf_gain600png.png manage 32.8 K 2010-05-27 - 22:12 TijmenDeHaan  

This topic: ColdFeedback > WebHome > DanTransferFunction Topic revision: r3 - 2010-09-02 - TijmenDeHaan
© 2020 Winterland Cosmology Lab, McGill University, Montréal, Québec, Canada