Indeed. Methodological listening test procedures are needed. They don't always have to be the standard audio textbook type though; perceptual analysis is a deeper and more complex field than simply rote ABX DBT testing (which are sometimes done in an amateur/incompetent way by engineers who have no particular expertise in perceptual testing/analysis). Its hard, time consuming, and expensive to do to publication quality, so it rarely is done quite that formally.
Exactly. Which is why I am vastly more skeptical of "many people say" arguments than I am of measurements. Just as I view qualitative science with more skepticism than quantitative science. I'm not saying that's right or wrong; only that it is my preference.
Tom
Tom
Measurements with instruments are no doubt more precise, no question there. However its just as ESS and others have pointed out, standard measurements don't tell the whole story, so we are left in a quandary as to what to do, if anything, about missing information. Should we just ignore the subject, pretend it doesn't exist, or what? Its not an easy problem.
Search LM317 gives 10 pages of posts. Elvee D-Noizator has 2900 posts. There must be many successful projects using LM317. There ought to be a best practice thread with ratings of various designs. The minimalist ones would have a poor rating for audio work.
Here is a reference I like. https://www.edn.com/simple-circuits-reduce-regulator-noise-floor/
Here is a reference I like. https://www.edn.com/simple-circuits-reduce-regulator-noise-floor/
Of course. Its always one choice. Just saying low noise is not always the predominate thing that matters. Sometimes it matters a lot, and other times other things can matter a lot too. Depends.
Regarding noise floor, noise as seen on an FFT can be complex subject. Anyone interested in going there? Reason I ask is understanding the how and why of it can take some effort. Not everyone cares about noise enough to bother.
Regarding "best practice," there is no best practice for everything in a general sense. Does a regulator need low noise from 10Hz to 20kHz, or does it need wider bandwidth such as the low noise RF regulators like LT3042. What about output Z versus frequency? Is it okay if it becomes more inductive as frequency goes up, or should output inductance go down as frequency goes up (over the bandwidth of interest)?
What about stability, is a power supply stable with the intended load?
Only point is, focusing on low noise alone is not sensible from an engineering perspective. First it should be known what a particular load circuit needs. Is a moving coil preamp? Is is a dac? Is it a power supply for some relays that switch preamp inputs? Right?
Last edited:
The crucial thing here is the precision of the 6V supply. According to "some that know", this device needs 6.04V. You could go through a batch of 7806s and (possibly) find one that's at just the voltage you need, but I wouldn't bet on it! The '317 is a better solution, since the output voltage can be minutely tweaked, and if you use good quality multiturn preset pots, it'll stay put! I was also pleasantly surprised by how well the temperature compensation works in the '317. I've used them in a few designs to provide reference voltages (where absolute precision isn't called for) and the results have always been fine.
Incidentally. the constructor of this PSU might want to consider giving the temperature compensation a helping hand by fitting a small heatsink to the '317. It appears that he needs around 600 mA, and - depending on the input voltage - he might be warming the IC up a bit!
Incidentally. the constructor of this PSU might want to consider giving the temperature compensation a helping hand by fitting a small heatsink to the '317. It appears that he needs around 600 mA, and - depending on the input voltage - he might be warming the IC up a bit!