Good to know and agreed it is very good practise to filter the DC supply. If anything its way too noisy at that stage.
On the capacitors for Vout and Vadj it's a bit tricky for other reasons as well, LM317/337 Noise TNT Audio for more details.
I've done some spectrum measurements using Scarlett 3rd gen usb interface up to 96khz at the differents stages that I can post laster if anyone is interested.
On the capacitors for Vout and Vadj it's a bit tricky for other reasons as well, LM317/337 Noise TNT Audio for more details.
I've done some spectrum measurements using Scarlett 3rd gen usb interface up to 96khz at the differents stages that I can post laster if anyone is interested.
If you are handy with second order differential equations, and their quick solution using Laplace Transforms: you can install a high current inductor in series with the prefilter "R" and get a C - R - L - C prefilter upstream of the LM317 . . . . . As long as you use Laplace Transforms to calculate the damping ratio "Zeta" of the damped resonant circuit. If Zeta is sufficiently large, voila: no transfer function "peaking" , and tremendously better rejection of HF noise. Ten ampere inductors cost less than $3.00 in qty=1 ; and if you're running an LM317 you probably don't need ten amps; I imagine a four ampere inductor would suffice.
I can't say I am but as it were I don't do any power supply without a power "capable" inductor. I usually combine R and L depending on what the current needs are and the space I have to get a workable mix of R and L in the same inductor. And I always listen directly to the supply rails /check it with the scarlett and REW.
For this (125mA) I used a 2.2mH air core inductor at 2.5 ohms between rectifiers/10,000uF and 3,400uF at the LM's pcb which I am aware could do with higher resistance but it was what I had and it lowers the mains fundamental by about 20 dB to -60dB relative to 5V rms (measured between +- ie 84VDC) and it's harmonics by a lot more at that stage, the LM's almost get me down to the noise floor of the Scarlett G3 USB interface. Adding 0.25 ohms between the LM's and the amp plus additional filtering at the amp have gotten me down to the noise floor of the scarlett, at least up to 96khz.
Perception and sonics wise the 0.25 ohm resistor between the LM's and the VAS in the amp was a surprisingly big improvement.
I still proceeded to build an addition regulation stage based on the TL431 and implemented it with the LM's at 36.5V, a 13 ohm resistor and the the TL's set to 34.5V letting the TL's sink 20-30mA.
Subjectively the lower frequencies got a bit more powerful but also less defined and the top end a bit less forward but still very refined and present however the whole setup got super sensitive and to get the voltages right without over temping the TL's was difficult.
The differences was however very small, the Scarlett and REW could not tell them apart and neither could I listening directly to the amplified noise at the rails so the TL's are out and I am sticking to the LM317/337 as above for this application 😊
For this (125mA) I used a 2.2mH air core inductor at 2.5 ohms between rectifiers/10,000uF and 3,400uF at the LM's pcb which I am aware could do with higher resistance but it was what I had and it lowers the mains fundamental by about 20 dB to -60dB relative to 5V rms (measured between +- ie 84VDC) and it's harmonics by a lot more at that stage, the LM's almost get me down to the noise floor of the Scarlett G3 USB interface. Adding 0.25 ohms between the LM's and the amp plus additional filtering at the amp have gotten me down to the noise floor of the scarlett, at least up to 96khz.
Perception and sonics wise the 0.25 ohm resistor between the LM's and the VAS in the amp was a surprisingly big improvement.
I still proceeded to build an addition regulation stage based on the TL431 and implemented it with the LM's at 36.5V, a 13 ohm resistor and the the TL's set to 34.5V letting the TL's sink 20-30mA.
Subjectively the lower frequencies got a bit more powerful but also less defined and the top end a bit less forward but still very refined and present however the whole setup got super sensitive and to get the voltages right without over temping the TL's was difficult.
The differences was however very small, the Scarlett and REW could not tell them apart and neither could I listening directly to the amplified noise at the rails so the TL's are out and I am sticking to the LM317/337 as above for this application 😊