steve wrote:Interestingly, linear units are used a great deal in digital signal processing. When I first started using Audacity I was also very critical of the +/- 1.0 default scale. However, I have softened my opinion over the years. I agree that novice users will benefit from "education", but we need to balance that against "raising the bar" for entry level.
I know, Steve. dB is relative, linear should be absolute. Only, most of the time, my equipment isn't that well calibrated that I can compare measurements with others
Going a bit deeper, "dB" does not provide "absolute measurement" - "dB is not a unit" - "dB" only gives a "ratio" that describes the relationship of one thing relative to another. On the other hand, the linear scale does provide "absolute measurement". A signal that has a maximum of +1.0 and a minimum of -1.0 is full scale - this is not an arbitrary assignment, it is the actual numeric values of the samples. For 16 bit integer, sample values are determined by dividing the binary value of the sample by 2^16. which normalizes the value to a range of +/- 1.0. Similarly for 24-bit integer, sample values are determined by dividing the binary value by 2^24. Thus, regardless of the number of bits, all integer formats have the same range - that is, full scale for 8-bit produces the same analog amplitude as full scale for any other integer format. For floating point formats (usually 32-bit float or 64-bit float), no normalization is used, so a sample value of 1.000000000000 has a linear amplitude of 1.000000000. If you want to experiment with this, you can capture individual sample values using the Nyquist command SND-FETCH
Legacy syntax (mono track):
New (version 4) syntax (mono track):
Again, thanks for all the effort.
I'm trying to concentrate on the project at hand: deploying a recording tool to about 20 users. To start with. A bit different from my own use for Audacity, which is mainly editing, shortening and occasionally, recording. Whenever I have to record "professionally", I use Boom Recorder. Because it has a number of things Audacity doesn't have. Like take administration. Or a running record buffer for these times I'm just a second slow. And time code. But BR doesn't do editing, it doesn't even do playback of it's own recordings once the file is closed. That's a big advantage, but makes it useless for this project.
As I'm easily distracted and Win 10 and Reaper 5 were released, I'm juggling a couple of other things too. So it's not a good time to start playing with nyquist code.
And for that project, I think most of these inexperienced users will do better with a realistic view. Even in general, I still think it's better to do away with the scientific observations. Both methods are a good as their references. And I'll admit dB has a reference problem. But so does linear. Let's take the following example:
You have an amp with 10 times gain. You know that, because you've measured the voltage on the in- and output. However, if you don't know the impedance, you don't know the amplification. Because there is also current amplification. Once you set that in dB's, the problem is gone. At least if you know the different dB's...
Even if inexperienced users are new to dB's, I can assure you explaining impedance to them is much, much harder. And you have to start somewhere to get anywhere.
