0
$\begingroup$

I'm using a Raspberry Pi 2 and an AD/DA Board, with no extra circuit. Reading on the internet though, I found that DACs need a low pass filter in their output, and I don't understand if the behaviour of my DAC is normal, and I wonder if I need the low pass filter. Its output goes down to 0V at every sample, is it normal? Is this the reason why low pass filters are used?

Thanks in advance,

Alberto


I tried with a better oscilloscope to analyze the signal, and it's all fine. I'm so sorry if I created a useless thread, but I still have learnt new terms I didn't know before.

$\endgroup$
  • $\begingroup$ Can you please clarify if you observe an output like 1 0 2 0 3 0 4 0 5 0 6 0... and so on (?) or is it that the DAC outputs a few values and then drops to zero and stays there? Also, is this simulation or an actual DAC in a circuit? $\endgroup$ – A_A Nov 23 '16 at 8:52
  • $\begingroup$ It goes to zero at every value. The DAC is not simulated, is a DAC8532 by Texas Instruments, controlled by a Raspberry Pi 2 $\endgroup$ – user25006 Nov 23 '16 at 10:29
  • 2
    $\begingroup$ Can I please ask you to add all this information to the body of the question? This is all valuable to answer the question. Also, if you have access to an oscilloscope, can you take a screenshot of the output and include that too? Are you sure it is not a problem with clock synchronisation or the way the DAC is driven by the Raspberry Pi code? Is it on GPIOs which are driven programmatically through the PI's module? The DAC should have provision for what is known as Anti-Imaging filter, have you configured that up? $\endgroup$ – A_A Nov 23 '16 at 10:31
  • 1
    $\begingroup$ If every value goes to zero,then how does the spikes come. Like A_A mentioned can you add a snap of output? $\endgroup$ – Navin Prashath Nov 23 '16 at 11:49
  • 1
    $\begingroup$ Eventhough being quite exciting, this is a hardware biased question. Whether your DAC's output voltage or current goes to zero or jumps to spikes depends on your electrical circuit topology and not on the way you process your signals. The reconstruction lowpass filter would only smooth out your DAC's steppy output voltage. That filter cannot be responsible for any weird electrical behaviour (assuming properly connected) $\endgroup$ – Fat32 Nov 23 '16 at 14:22
1
$\begingroup$

The TI DAC8532 uses a Kelvin resistor-string (string of resistors that divide the reference voltage, with individual switches from the nodes to the output), so it is guaranteed to be monotonic. What you see does not sound normal, you must be writing zeros to it over the serial interface somehow. How did you implement the serial protocol?

The low-pass filter you put on the output is the reconstruction filter. It is used to reduce imaged frequency content (repeated frequency content above the Nyquist frequency). It's not always needed. The DAC has a built-in zero-order hold element that attenuates the images somewhat. If you are driving a system with low-pass frequency response, like a DC motor with a heavy mass attached, then you don't need the reconstruction filter.

$\endgroup$
  • $\begingroup$ Could you explain to me what a Kelvin resistor-string is and what monotonic means? Thank you $\endgroup$ – user25006 Nov 24 '16 at 10:55
  • $\begingroup$ @user25006: I've added some links and a little explaining. $\endgroup$ – Arnfinn Nov 24 '16 at 11:10
0
$\begingroup$

The output of a DAC is a continuous serie of Square pulses with an intermediate zero hold between each update of the digital inputs. Normally, the output does not drop to Zero after an update. The output appears spectrally as a serie of rays starting from the fundamental of the digital modulation and all of its repeated images. Those rays are modulated by the Sinx/x or Sine-Cardinal x enveloppe. The process is similar to the one of sampling at the input of an ADC and follow the same Nyquist laws. To remove the images a low-pass filter is necessary. Such filters must preserve the phase coherence and Bessel or Paynter filters are recommended. If necessary with the addition of notches on the first images if a large output bandwidth is required. In some cases, glitches may also appear on a DAC output at the major carry transitions like at mid range from 0111 1111 1111 to 1000 0000 0000. These glitches depend of the DAC architectures and may require a special "deglitcher" circuitry.

$\endgroup$
  • $\begingroup$ I just thought I should mention that string-DACs do not have large glitches as each level is individually switched. The desired phase response is application specific. A Bessel-filter might be good for audio, but they are terrible in a control loop (due to the large phase delay)... $\endgroup$ – Arnfinn Nov 23 '16 at 22:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.