Forums

DSP Implementation Woes

Started by Unknown January 30, 2008
brendan_online@hotmail.com wrote:
> Hi Ben, > > I was *told* that this chip supports several discrete clock > frequencies....namely 8,16,24,32,44.1,96 KHz, and it doesn't use a Low- > pass interpolation/reconstruction filter at the lower clock > frequencies. My problems are occurring at 8KHz. The 'modulation' > envelope is having an effect when trying to reconstruct a sinusoid at > about 3.4KHz. It's not too dissatisfactory at this point, but by 3.6+ > KHz there is alot of harmonic distortion from what I desire.
Add your own reconstruction filter. You will probably want sinc compensation to compensate for signal droop caused by the zero-order hold.
> In fact, at 8KHz the reconstruction looks of zero-order-hold type.
Without the ZOH, there would be practically no energy out. Almost all DACs do that. A reconstruction filter is in addition.
> So when I said 'push the chip', I simply meant to clock it up to 96KHz > and see if the signal fidelity of a discrete sinusoid at a discrete > frequency of about 0.8pi is any better.
Of course it will be, because of the mysteries previously revealed. (It will be the filter, not the sample rate.) Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
brendan_online@hotmail.com wrote:
> Okay let me state what I know about the hardware now. > > There is a fixed LPF of bandwidth 48KHz, as the maximum sampling > frequency is 96KHz. This is fixed, even though the DSP chip support > the clock frequencies that I previously listed. > > As such, the effects of the ZOH at the DAC are seldom mitigated, and > one closes in the vicinity of the Nyquist frequency this 'beating' > effect dominates. > > I think this explains it, or so.
Not exactly. You see the effect of insufficiently attenuated images, some of them inverted. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Of course the ZOH will attenuate so much of the infinite spectrum of
images of the baseband but of course without the tightly bound LPF
then as you said, the effect of the attenuated images plays its part.

Thanks for your input!! Any further comments are very welcome.

Regards....Brendan
On Jan 31, 8:36�am, brendan_onl...@hotmail.com wrote:
> Hi Ben, > > I was *told* that this chip supports several discrete clock > frequencies....namely 8,16,24,32,44.1,96 KHz, and it doesn't use a Low- > pass interpolation/reconstruction filter at the lower clock > frequencies.
I think the chip must be using some form of low pass interpolation/ reconstruction filter at the lower clock frequencies. Do you mean 'in addition to a ZOH'? A ZOH can be thought of as a rudimentary low pass reconstruction filter.
>My problems are occurring at 8KHz. The 'modulation' > envelope is having an effect when trying to reconstruct a sinusoid at > about 3.4KHz. It's not too dissatisfactory at this point, but by 3.6+ > KHz there is alot of harmonic distortion from what I desire.
If the chip was using simply a ZOH with a 'steplength' of 1/8000 seconds, the output would look distinctly un-sinusoidal at anything much above 1kHz.
> > In fact, at 8KHz the reconstruction looks of zero-order-hold type. >
At 8kHz are you seeing 8000 ZOH steps per second or are you, by any chance, seeing 8000n ZOH steps per second, where n is equal to (perhaps) 8 or 10? As pointed out earlier, the reconstruction filter must have a sharp cut-off and should have significant attenuation at half the sampling rate if frequencies just below half the sampling rate are to be generated. The reconstruction filter in your DAC may well be implemented as an FIR operating at n-times the sample rate you are concerned with and with a ZOH output (at n-times your sample rate). Input to the FIR (at n-times your sample rate) could be your samples padded with (n-1) zero values between each one. Try writing a data sequence in which only every (say) 100th value is non-zero to the DAC and you will get an idea of the impulse response of the reconstruction filter - it may well be the sort of truncated (windowed) sinc shape that is typical of a low pass FIR. And it may be made up of ZOH steps at n-times your sampling rate. Such an FIR low pass reconstruction filter _could_ be designed to have very significant attenuation at half the sampling rate but would then also attenuate frequencies just below half the sampling rate - in your example that would mean that by 3.6+ kHz there might be very little output at all from the DAC. Some audio codecs are designed so that the FIR reconstruction filter response is down only 6dB at half the sampling rate. (Does the datasheet for your DAC give the magnitude frequency response of the reconstruction filter?) Running at a sampling rate of 44.1kHz, the effect you describe will be present for frequencies of 20+ kHz but in practical situations may not matter because other components in a 'CD-quality' audio system may have low pass characteristics that mask the aliasing close to 22.05kHz.
> So when I said 'push the chip', I simply meant to clock it up to 96KHz > and see if the signal fidelity of a discrete sinusoid at a discrete > frequency of about 0.8pi is any better. >
What did you find? Regards, Donald
> Thanks again, > Brendan