Home  Audio Home Page 

Copyright 2010 by Wayne Stegall
Updated March 11, 2010. See Document History at end for details.

Line Level Class A Power Supply

This is an example only.  It is not intended to be final or complete design.

Figure 1:  Schematic

Parts List

T1P 20V transformer (or somewhat higher if 20V unavailable)
BR1P bridge rectifier, peak current specification
exceeding peak transformer current.
C1P-C4P 0.1F ceramic
C5P 4700F, 35V electrolytic
D1P,D2P 1N4002 diodes
10F, 35V electrolytic
240Ω, 5%
R2P 5kΩ or 10kΩ pot
100F, 35V electrolytic
Q2P,Q3P IRF510 n-ch mosfet
D3P-D6P 1N914 or 1N4148 diodes
R3P,R5P 4.7kΩ, 5%
R4P,R6P 1kΩ, 5%
C8P,C10P 220F, 35V electrolytic
C9P 10,000F, 35V electrolytic
C11P 3300F, 35V electrolytic
These specific calculations are for a power supply for the circuit of JFET Phono Preamp - Active Inductor Example.  You may have to change some values for other applications.  See text.for calculations.

Initial Design Decisions


Some of the component choices are not very critical, but I chose to calculate these.
vTRANSFORMER = 0.7071 x (vSUPPLY + vgs-max-Q2 + vDROPOUT-LM317 + vDIODE-DROP) = 0.7071 x (18V + 4V + 3V + 1V)  = 18.3846V
Round up to standard value of 20V.
Refer article: Power Supply Ripple Calculations and Capacitor Size, equation (3).
iLM317-INPUT = iDD1 + iDD2 + iR1P + iADJ-LM317 = 27.9778112mA + 19.4563897mA + 5mA + 100A = 52.5342009mA
C5P =
60Hz x 0.5V
 = 1.75114003mF
Choose instead common larger value of 4700F I chose these values a good while ago to meet this specification.  How I chose the first component in order to calculate the second, I do not remember.  Perhaps it was a noise consideration.  Calculate total noise from capacitor noise equation.  (The capacitor is not the source of the noise itself but interacts with the resistor in a way to become the sole determining factor.)


= sqrt
1.3806504e-23 x 298.15K
= 4.325615651nV
R3P = 1/C8P = 4.545454545kΩ, rounded up to nearest 5% standard value is 4.7kΩ.

As a matter of curiosity, the 1N914 specification of 5nA reverse current at 20V suggests an ac impedance of 800MΩ at very low ac voltages.  (This is due to leakage resistance.  Ideal calculated reverse current is very much lower.)  220F x 400MΩ gives a settled time constant of 88,000  seconds.  Because this is such a large value, I suspect it to be very imprecise. The common source circuit supplied by VDD1 is very sensitive to its power supply.  Indeed, the power supply is in the signal path for this stage.  Maximum output current from VDD1 is 41.4651312mA. The gfs of IRF510 is 1.3S @ 3.4A.  I want primarily capacitive power supply output impedance.  The transistor output impedance is non-linear (although with nice euphonic 2nd order harmonics!) and with the capacitor value determines the pole above which output impedance becomes capacitive.  Initially try for a pole at 2Hz.

IRF510 specification:  gfs = 1.3S @ 3.4A.

Calculate MOSFET constant from gfs@iD  specification.
kn = gfs2
= 124.2647059mA/V2

Recalculate gfs for specific iD.
gfs = 2 x sqrt(kniD) = 2 x sqrt(124.2647059mA/V2 x 27.9778112mA) = 117.9263241mS

Rout-Q2P =
2 x sqrt(kniD)
2 x sqrt(117.9263241mS x 27.9778112mA)
 = 8.70477913Ω

C9P =
2π x 2Hz x 8.70477913Ω
 = 9.14181398mF

Choose C9P = 10,000F and solve for the pole frequency.
fpole =
2π x 10mF x 8.70477913Ω
 = 1.828362796Hz

Figure 2:  Bode plot representing suppression of MOSFET
non-linearity by chosen bypass capacitor of 10,000F

A more common choice of 1000F gives a pole frequency of 18.28362796Hz.  This choice may give a slight increase in low bass distortion, and perhaps why some reviewers hear mid-bass bloom in some class A equipment.

VDD2 drives a circuit (source follower) which is relatively insensitive to power supply fluctuations.  Therefore you can choose a more arbitrary value here.  Save money here to buy a larger C9P.
I choose  C11P = 3300F.

Document History

January 5, 2010 Created
January 5, 2010 Replace vague load specifications with exact ones from article JFET Phono Preamp - Active Inductor Example and update calculations.
March 11, 2010.  Corrected for improper gfs presumptions.