Discrete Regulators +/-15V with good performance

First.
To feed the regulators you need a 2x15VAC transformer and rectifiers.
See first image.

The discrete regulators for positive/negative uses 5 transistors each.
The negative regulator is a mirrored circuit.
Test run shows very good data and precision thanks to 2 current sources.
 

Attachments

  • Transformer 2x15VAC.jpg
    Transformer 2x15VAC.jpg
    99.7 KB · Views: 492
  • Regulator Positive.jpg
    Regulator Positive.jpg
    152.3 KB · Views: 502
  • Regulator Negative.jpg
    Regulator Negative.jpg
    151.9 KB · Views: 496
  • Like
Reactions: 1 users

TNT

Member
Joined 2003
Paid Member
Impedance traces, line regulation, load regulation, noise, current capacity and PSSR - we want to know the characteristics if the 2 (+ & -) devices please.

And the lingering question.... is it better then LT3045? ;)

//
 
  • Like
Reactions: 1 users
As this is a lowdrop Regulator it will work very well even at 15.2V input.

Test of Line Regulation.

Line Regulation with 100mA load.
25.0 Vin: 15.0001
22.0 Vin: 15.0000
20.0 Vin: 15.00000
18.0 Vin: 15.0000
16.0 Vin: 14.9999
15.5 Vin: 14.9999
15.2 Vin: 14.9999

Line Regulation with 1A load.
25.0 Vin: 15.0001
22.0 Vin: 15.0000
20.0 Vin: 15.00000
18.0 Vin: 15.0000
16.0 Vin: 14.9999
15.5 Vin: 14.9999
15.2 Vin: 14.9997
 
Test Load Regulation. How well regulation is with different, varying load.
With 20V input it works excellent. In all loads.
Also with 15.5V input it is good. With one exception: 5A load.
5A load needs like 16V input.

Load Regulation at 20V input.
0.01 A Load: 15.0000
0.02 A Load: 15.0000
0.05 A Load: 15.0000
0.10 A Load: 15.00000
0.20 A Load: 15.0000
0.50 A Load: 14.9999
1.00 A Load: 14.9999
2.00 A Load: 14.9998
5.00 A Load: 14.9996

Load Regulation at 15.5V input.
0.01 A Load: 15.0000
0.02 A Load: 15.0000
0.05 A Load: 15.0000
0.10 A Load: 15.00000
0.20 A Load: 15.0000
0.50 A Load: 14.9999
1.00 A Load: 14.9999
2.00 A Load: 14.9997
5.00 A Load: 14.8584

Load Regulation at 16V input.
5.00 A Load: 14.9995
 

TNT

Member
Joined 2003
Paid Member
Line (load) regulation "should" also be tested with a step in the load i.e. not a staty state test. It's like a transient test.

How much noise do these regulators generate?

What is the output impedance? ohm vs frequency please?

//
 
Member
Joined 2011
Paid Member
Mouser sells 2200uF 25V electrolytic capacitors for less than 1.00 USD. You could pair that with a 10 ohm resistor to make a lowpass pre-filter before the regulator, giving you an additional 20dB of line rejection ("PSRR") at 100 Hertz. It would also reduce the VDS across your output MOSFET by (10 * 0.1) = 1.0 volts.

Or if you are willing to spend even more money, for additional 0.50 USD you could connect a Bourns RLB0914-102KL inductor in series with the resistor, now giving yourself an R-L-C lowpass filter with good line rejection at 100 Hz and tremendous line rejection at 1000 Hz. Since the inductor's series resistance is 2.1 ohms you'd drop the fixed resistor's value to approx 8.2 ohms, in order to keep the total resistance around 10 ohms and the voltage drop around 1 volt.

You can ask your circuit designer to explore these proposed modifications in a circuit simulation program like SPICE, to find out how much line rejection you'll achieve (without, and then with, the modifications). Then the two of you can decide whether it is a good use of 1.50 USD to add the changes.
 
Member
Joined 2011
Paid Member
The easiest way to reduce noise is to increase the RC timeconstant of the lowpass filter between the LM329 and the opamp. Or, for even greater noise reduction, subdivide it into an RCRC ladder plus shrewdly calculated partitioning. This will reduce higher frequency noise dramatically.
 
Last edited:
I took the liberty of doing some sims of a slightly modified circuit. Replaced the 1K5 resistor with a constant current source and the LM329 with back to back zeners.

Load-line regulation
1694900774034.png


PSRR
1694900854473.png


Output impedance
1694901062718.png

1694901089829.png


Noise
1694901266131.png
 

Attachments

  • 1694899670687.png
    1694899670687.png
    43.1 KB · Views: 57
  • 1694899679936.png
    1694899679936.png
    78.9 KB · Views: 49
  • 1694900201199.png
    1694900201199.png
    44.6 KB · Views: 49
  • 1694900258090.png
    1694900258090.png
    83.3 KB · Views: 41
  • 1694900421169.png
    1694900421169.png
    56.5 KB · Views: 49
  • 1694900439562.png
    1694900439562.png
    81.1 KB · Views: 77
  • Like
Reactions: 1 user
Member
Joined 2011
Paid Member
Walt Jung (and I) would recommend that you increase R2. Walt's recommendation is R2=5.6K, but he envisions implementations where R8 is fixed and unchangeable, and "Rpot" does not exist. Seeing that you do include "Rpot", I strongly recommend R2=22K. Which is even larger.

Also I would suggest that you think very hard about why you use an approximately-constant-current-source at the tail node of a diff amp whose input is a fixed, constant, unchanging, stable DC voltage. There is no common mode input signal hence no need for wonderful CMRR. In what way is the CCS implementation superior to a simple resistor between the emitter tail node, and ground? Why spend the parts cost and the board area? Are you simply copying someone else's design, imagining & hoping that they know more than you know?