Adding numbers is the math, my teacher told me.
Adding which numbers, exactly? Please specify what is being summed to achieve a simple repeating of a given sample?
Exact interpolation to an arbitrary error is a closed form math problem, approximations are only due to computational issues.
So what is gained , I don't get it? You could repeat them 8X to 352.8k but the analog output is the same staircase waveform with no benefit at all?
I tried to address it in a earlier post; unfortunately the term "upsampler/upsampling" is used inconsistently, as it sometimes denotes just the sampling rate increase and sometimes denotes the whole resampling process including the interpolation which means a digital low pass filter.
It gets a bit more "exciting" if the sample rate change factor is not an integer number.......
In the integer case - according to plasnu - its simply raising the sample rate by the integer factor N and repeating the respective sample N-1 times.
Last edited:
Scott, he's actually wrong when he indicates that the output waveform is the same in either case. What doubling the sample rate, simply by repeating successive samples, does is to trick a digital interpolation filter into not removing the lower half of the first image band, in the case of an 2x rate increase. So, instead of filtering the signal at it's original Nyquist frequency, it filters it at the new higher rate's Nyquist frequency, which then will leave parts of the lower image band(s) unfiltered.
The analog signal output from the DAC will not be correctly reconstructed. The analog signal will be somewhat similar to NOS, and somewhat similar to OS. An hybrid, which may combine the best subjective aspects of both NOS and OS, or maybe the worst subjective aspects of both, or maybe simply sound different from either without sounding better.
Last edited:
Yeah, i mentioned that feature in an earlier post; i have my doubts that the izotope resampling plugin really allows such a process (at least not without a warning) as it otherwise seems to combine the worst of both if sending the new data sequence to a DAC (which will react with an internal digital low pass filter setting according to the "new" sampling rate) while the additional samples would represent additional distortion.
Indeed. Are we talking about what happens when you feed those data streams to a DAC chip, or to a DAC in a box with a receiver and additional components?
I didn't say the output waveform. I was just answering his question comparing before and after upsampling without interpolation. The waveform is the same.
The OP's questions can be construed in multiple ways. As such, one answer is that there is no such thing as playback side over/upsampling which doesn't utilize some form of interpolation. Pseudo-over/upsampling via the simple repeating of successive samples changes nothing about the digital signal. It can, however, be utilized to improperly program the cut-off frequency of an following digital interpolation filter. In which case, the digital signal is still run through an digital filter.
Oversampling can be applied at the recording side of the chain. Meaning, the channel bandwidth is purposely made much wider than the signal bandwidth. In which sense, the signal bandwidth in being widely oversampled. Wide enough to enable the playback end to utilize only analog reconstruction filters without utilizing an digital interpolation filter.
DSD is an existing commercial example of an widely oversampled system requiring only relatively mild analog signal reconstruction filters. Such oversampling is also possible utilizing an PCM format, however, I know of no commercial standard involving PCM.
Pathetic!!
Here's some help ... pay attention. It's only 11 min long...
YouTube--A brief introduction to how Finite Impulse Response (FIR) filters work for digital signal processing.
Right... not even after 8 pages (71 posts). Pathetic!!!!!!!!
Here's some help ... pay attention. It's only 11 min long...
YouTube--A brief introduction to how Finite Impulse Response (FIR) filters work for digital signal processing.
Last edited:
And in all those years, you haven't come up with comparing the input to the output, as that's the way any audio equipment is being tested since the birth of the universe?
To quote Jesse Peterson: Amazin'
Wow indeed.
DACs are tested by comparing input to output? Certainly news to me.
Come to think of it, this statement of yours looks like a classic deflection. The guess I was referring to looked to be based on subjective ('only benefit') criteria rather than testing but happy for @plasnu to explain further to demonstrate the error of my assumption.
Last edited:
I think we have answered it, although perhaps in a slightly roundabout way due to the original question being unclear and the subsequent discussion introducing confusion.plasnu said:
The original question was
There are several sensible answers to this, all of which have been included somewhere in this thread:ygg-it said:
1. Proper oversampling requires a digital filter so to do oversampling without a digital filter is not possible.
2. You could do 'oversampling' by merely repeating samples; this has no effect - it cannot confuse a digital filter as we have already excluded a digital filter (which excludes some DAC chips).
3. You could do 'oversampling' by zero stuffing; this has the effect of reducing the HF rolloff inherent in simple NOS but at the same time it will increase image amplitude - which may be unwanted.
Of cause you can test DACs that way, it has been done forever.
A dac is part of a digital recording and reproduction chain.
You simply compare input to output.
Jeezz.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.