Topic: Upsampling to edit 16/44.1 files

[nassau73]

Recently I've been going back to some old mini disc recordings that are 16/44.1 and remastering now that RX has come into my life.

Searching around various audio sites, I've come across articles that say when professionals are working with 16/44.1, they will upsample for editing and then downsample back to the format required by their client.

One comment in a forum at audiogon.com caught my attention and I thought it might be of interest to some here:

Upsampling is not just adding a bunch of zeros. Going from 16 bit to 24 bit (at the same sample rate) is just adding a bunch of zeros. However going from 44 KHz to 176 KHz adds data points that are calculated from the existing points. The interpolation is not just a linear one, but one based on analyzing the data points before and after the original points. How well that interpolation is done depends on the algorithm used. Which means that different DACs and different software upsamplers can produce different results, although the differences are usually pretty small.

Upsampling to higher sample rates is a controversial subject. Some claim that it cannot make a difference based on the mathematics of digital sampling. Others claim it can, because DACs are not perfect in how they do digital to analog conversion. Some handle 96 KHz, for example, better than 44 KHz.

The upsampling in a DAC can be bettor or worse than a software upsample. It really depends on the DAC. One reason J River added upsampling was to provide a better upsampling routine than some DACs used.

The big difference between a $500 DAC and a $5000 one is not the upsampling routines but the digital to analog conversion. Inexpensive DACs use off the shelf chips whereas at least some of the expensive ones are based on the companies own propitiatory algorithms.

The best thing to do is to try the software upsampling and see what you think. Note however that if the DAC is upsampling and you cannot turn that off, you will need to send it a software upsampled signal that is at least as high as its upsampled rate, or you will be using both the software and hardware upsampling. Also, you should upsample as an integer multiple of the original sampling rate to reduce computer usage and to also maintain the original data points as anchors in the upsampled data stream. For example, 44.1 KHz is better upsampled to 176.4 than to 192 KHz.

[lpmaskman]

Modern DAWs are already count at 32bit or 64bit float rates. So if you thow your audios into any DAW it's will processed upsampled anyway. So you don't need to upsample it separately.

[Gutbucket]

^ No.

Sample rate is not bit depth.  Bit depth is not sample rate.

Yes, most DAWs do their processing in a high bit depth floating point workspace, regardless of the actual bit depth of the original file.  However they will do such processing at the original sample rate of the file.  One could upsample/convert the original file to a higher sample rate before working on it.  However, there is an easier (and generally better, less potentially problematic*) way of achieving essentially the same thing, which is to use the up-sampling option available within whatever plugin you are using (when available). That upsamples within the plugin, allowing it's internal calculations to be more precise without rounding errors, then converts back again.  This is often selectable at 2X, 4X, 8X, etc, of the native sample rate of the file, with higher rates requiring more processing overhead.  I don't have Izotope RX, but pretty sure it provides such an option.  If its actually audible or not it's easy enough to use that, within the processing resource limits of the of the computer you are using.

*There is a Dave Rat video which is close to getting at this directly, yet has a slightly different focus, and essentially reaches this same conclusion.

[rocksuitcase]

WernerO has a fabulous 3 part description of their workflow which discusses upsampling, it's pros and cons thoroughly in part II:
https://www.tnt-audio.com/sorgenti/rip_it_2_e.html

For the beginning link: https://www.tnt-audio.com/sorgenti/rip_it_1_e.html

an excerpt on this topic:

What sample rate to choose? While psycho-acoustics suggests that 44.1kHz is almost enough, and while there seldom is real signal content above 20kHz on LPs, it still is wise to record at a higher rate.

How high? Quadruple rates (i.e. 4 x 44.1 = 176.4kHz and 4 x 48 = 192kHz) may appeal to the numbers brigade, but are often quite pointless. One reason is the lack of content in the source signal, the other main reason is that many quad-capable ADC chips in reality aren't. Audio ADCs these days are almost invariably of the sigma-delta type, i.e. massively-oversampled low-bit convertors running at several MHz, followed with digital anti-alias filtering, decimation, and large amounts of noise shaping to move the gross quantisation error of the low-bit convertor out of the audio passband.

For a base 44.1kHz ADC, the noise shaper must clean up the sub-22kHz range. For a 96kHz convertor this is the sub-48kHz range, and for a 192kHz device this would be the sub-96kHz range. But what we see is that many ADCs, operated at 192kHz output sample rate, exhibit a lot of shaped noise above 48kHz, still in the bandwidth of interest (which extends to 192/2=96kHz). At the same time their datasheets reveal that their anti-aliasing filter has a poor suppression and in-band flatness. The underlying reason is that 4 x ADC chips often are 2 x designs that were stretched to accommodate another octave. But this octave isn't captured cleanly, and thus quite useless

His decision became a choice depending on the clients desired output media of: 88.2k or 96k for recording.

[Gutbucket]

Except that's all focused on the capture side of things, thus analogous to what sample rate we use for the original recording, in contrast to up-sampling an already existing digital recording for processing which is the OP's question.  The bit about over-sampling converters is related, yet different from up-sampling and existing digital file.

Dan Worrall goes into upsampling the entire project verses retaining the native rate and up-sampling within the plugins in this FabFilter vid.  The whole video is helpful, but you can skip to 21:27 for examples.
https://youtu.be/-jCwIsT0X8M

[rocksuitcase]

Except that's all focused on the capture side of things, thus analogous to what sample rate we use for the original recording, in contrast to up-sampling an already existing digital recording for processing which is the OP's question.  The bit about over-sampling converters is related, yet different from up-sampling and existing digital file.

Dan Worrall goes into upsampling the entire project verses retaining the native rate and up-sampling within the plugins in this FabFilter vid.  The whole video is helpful, but you can skip to 21:27 for examples.
https://youtu.be/-jCwIsT0X8M

WernerO's main focus is recording from Vinyl to digital using a Tascam DA-3000. (read post #1) So I was figuring it as the vinyl is akin to one of our recordings of a certain bit rate/sample depth and he is upsampling it to the digital realm. But, I see your distinction for sure.

[wforwumbo]

I've spoken about high bit depths and sample rates at length on this website. Feel free to peruse my comment history for a good overview.

Upsampling from a native format will always introduce error. The error can be minimized, and arguments can be made regarding how perceptible the error is. It's always better to record at a native sample rate higher than what you need, and downsample later when releasing.

Re: the discussion of noise shaping and sigma-delta converters... some of this is slightly out-dated compared to the modern state of the art. It isn't entirely untrue, and the high sample rate 1-bit decimation filters are still widely used, but the noise shaping was more for quantization errors when integer data types were the common format on the block. 32- and 64-bit floating point precision have changed the game a little bit. It ALL depends on the data format, and understanding that fundamental truth keeps me gainfully employed.

Reinforcing that bit depth is not sample rate, and sample rate is not bit depth. Additionally, you can never rely on DAWs, plugins, or systems to do sample rate upconversions cleanly unless you understand why you are using one algorithm over another.