<div dir="ltr"><div dir="ltr"><br></div><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><br>
So I'm having a hard time seeing where these images could come from, or how you'd ever detect them. You definitely couldn't hear them.<br><br></blockquote><div>This is highly context-specific. Depends on 1) your sample rate or more particularly, 2) the ratio of your sample rate to signal bandwidth 3) as Roman said, the presence of any nonlinearity in the path<br><br>You are right that there is no such thing as an instantaneous step in the real world, but for audio purposes I think it's reasonable to consider the continuous time impulse response to be a "boxcar" considering the rise time in relation to the sample period.<br><br>We can think of the DAC output as our discrete signal passed through a 0-order hold of period 1/Fs; the discrete signal contains images at integer multiples of the sample rate. These images are the <b>same</b> level as the baseband signal in the discrete domain - I think this is what Tom was getting at. Only when we pass the discrete signal through the DAC (0-order hold) do we shape the output spectrum(including those images), which you can picture as a train of scaled pulses convolved with the frequency response of the ZOH. <br><br>The ZOH is basically a crappy lowpass filter. Attenuation at FS/2 is like 4db. for a high bandwidth signal, this is *not acceptable* generally, so use reconstruction filters to 1) attenuate the images and 2) possibly correct for passband droop due to the SIN(X)/X attenuation of the ZOH. the second one is probably done more in non-audio applications. Basically, we try to filter our steppy signal back into the perfect band limited version we intended, but there is no causal brick wall filter, so this is theoretically impossible. The best you can hope for is to push those images far down enough that they're below the noise floor/threshold of human hearing.<br><br>Re: being able to hear those images: if you sample high enough, it likely won't be a problem, in a perfect world. but since we're talking about synthesizers here, nonlinearities will undoubtedly exist, as roman references. All it takes is mild intermodulation to bring those images back into baseband. <br><br>As always, this is all dependent on context/application.</div></div></div>