DF1704

DF1704







Home page
News
Projects
Download
Questions
Gallery
Links
Article
LM317, understanding and listening - July 2007
Last revision on July 2007

LM317, understanding and listening - July 2007, by Eric Juaneda

   LM317 regulator is very popular device. It is present in most preamplifier or CD. In this article I try many decoupling method and compare sound quality.
With regulator, we think at first noise rejection. The LM317 datasheets from National semiconductor gives decoupling solution. What about sound and musicality?

For this tries, I use my model 3 preamplifier with following power supply.

Power supply
Note: 1N4007 is very bad choice. Prefer fast rectifier diodes to prevent reverse current.

It is not easy to describe sound quality. For a comprehensive comparison, I use 8 categories.
  • Softness,
  • Tone realism,
  • Dynamic,
  • Definition,
  • Bass level,
  • Bass depth,
  • Space clarity,
  • Voice position.
First listening test

R1 = 240 Ohm, R2 = 4.3K Okm

Test 1 For this first test, I follow solution given in LM317 Datasheet.
Cout = 1µF Wima MKS2
Cadj = 10µF Tantalum (while datasheet mention solid tantalum)

Test 2 Cout = no capacitor (risk of ringing!)
Cadj = 10µF Tantalum

Test 3 Cout = 1µF Wima MKS2
Cadj = No capacitor

Test 4 Cout = no capacitor
Cadj = No capacitor


To achieve comparison, I use switches to compare solutions.
Comprehensives diagrams. Only given to understand which solution sound better than another one. Diagrams are not drawing to scale. Do not compare between us. Only comparatives values, no absolute values.

Note about listening test
Softness

Realism

Dynamic

Definition

Bass level

Voice depth

Space clarity

Voice position

First remarks
  • Without capacitors LM317 sound good but still very noisy (16mV pp). see details
  • Datasheet recommendations greatly reduce noise, but sound is deadly boring!
  • Cout maintains stability while Cadj reduce noise.
  • Incorrect capacitor's values sound bad and can cause ringing.


Second listening test

R1 = 240 Ohm, R2 = 4.3K Okm

Test 5 Cout = no capacitor
Cadj = 10nF styroflex
R1 = R1/2 (120 Ohm), R2 = R2/2 (2.2K Ohm)

Test 6 Cout = 4.7µF MKT + 22 Ohm
Cadj = no capacitor
R1 = R1/2 (120 Ohm), R2 = R2/2 (2.2K Ohm)

Test 7 Cout = no capacitor
Cadj = No capacitor
R1 = R1/2 (120 Ohm), R2 = R2/2 (2.2K Ohm)

Test 4 (Reference point, same values as first test)
Cout = no capacitor
Cadj = no capacitor

Test 8 Cout = 4.7µF MKT + 22 Ohm
Cadj = no capacitor



Softness

Realism

Dynamic

Definition

Bass level

Voice depth

Space clarity

Voice position

Second remarks
  • Driving LM317’s reference point by current gives good result.
  • Decreasing Cout value, increase quality.
  • Qualities of capacitors greatly change sound.
Conclusion
Driving LM317 reference point by current gives good result.
Taking care with capacitors technology is very important. Even a little 10nF modify sound quality. Prefer polypropylene or polystyrene to ceramic or plastic capacitors. You can adjust tonal balance by using different capacitors technology, polypropylene or polystyrene for soft sound, ceramic or plastic for savage sound.
Use Cout as small as possible. Take care ringing with too small Cout value.


Here is an example of good subjective sounding solution.
Power supply

http://tech.juaneda.com NOTICE: The information provided on these pages is intended as guide for designing audio equipments. They are free to copy and or publish. To use any information from this WEB site, please make sure to refer to title source, author name and URL Link.