LT Dual Tracking 3A Regulated Power Supply 1.25V to 20V

[woody]

Okapi, what a cute little circuite. I bet it boils down to the fact
that that 5k pot is acting like a pair of 2.5k pots in paralle with
r2 and r4 as you saw but each of the imaginary 2.5k pot is not
returned to ground like r2 and r4 but to the + & - 1.25v above and below ground this just about solves where that 2 volts
went.

[okapi]

i had some time this weekend and i designed a pcb for the power supply as well as tested the tracking.

i'll post the pcb later, but i wanted to report that if i unload and load one output rail, the opposite rail moves in parallel, maintaining the voltage difference between rails - just as expected!

i also requested a couple of sample lm329b's from linear so that i can test the high stability circuit.

thanks for everyone's help.

[okapi]

This is the schematic i've come up with for the second stage of my voltage supply. It is essentially the tracking circuit with filter and bypass caps, the necessary protective diodes, and some optional voltage dropping resistors at the input.



Based on my reading about the tri way power supply in audioXpress as well as comments posted here on DIY audio (ex. 1.) it seems that the use of a star ground is preferable to using a ground plane.

Taking this idea to the extreme would require that the ground side of every component be run to a central ground. This would seem to be very difficult to implement.

What is the next best option? How do i decide what grounds to bring together before i run them to a central ground? Where should this central ground be located? Should it be closer to the rectifier bridge or at the load?

thank you for your help.

[woody]

Why did you use the combination of a resistor and the 1n4148 ?
wouldn't just a 1n4001 without a resistor be simpler and better ?

[okapi]

Hi Woody,

i copied this idea from the audioXpress article (Tri-Way Low Voltage Supply, Pt. 1, by Paul J. Stamler).

Here is a direct quote from the author: " I followed (Ben) Duncan's suggestion and used 1N4148s as regulator safety diodes, along with current-limiting resistors; these are less likely to conduct high-frequency crud around the regulators"

I think 4148s are much faster than the 4001 family of diodes but can't conduct as much current. I think the resistor acts to protect the diode while at the same time attenuates any "high frequency crud".

do you think this is a reasonable idea?

thanks for following the thread.

[Bakmeel]

Originally posted by okapi

.... do you think this is a reasonable idea?

You understand the diode is there for protecting the regulator when the input voltage is lower than the output. Limiting the current capability of this diode also reduces its capability to protect the regulator. It will only handle small voltage differences. Is that good enough for you? Would you be able to do without diode at all?

I think there's a typo in the value for C5 or C6.

[okapi]

by rereading the data sheets it seems like the necessity of the protection diodes is similar for both regulators.

For the LT1033 the data sheet says:

if the output voltage is greater than 6V and an output capacitor greater than 20uF has been used, it is possible to damage the regulator if the input voltage becomes shorted, due to the output capacitor discharging into the regulator. This can be prevented by using diode between the input and the output.

for the adjust pin diode

if the bypass capacitor is larger than 10uF and the output is larger than 25V, a diode should be added between the output and adjustment pins.

For the LT1085 the the data sheet says:

In normal operation, the LT1083 family does not need any protection diodes. Older adjustable regulators required protection diodes between the adjustment pin and the output and from the output to the input to prevent overstressing the die. The internal current paths on the LT1083 adjustment pin are limited by internal resistors. Therefore, even with capacitors on the adjustment pin, no protection diode is needed to ensure device safety under short circuit conditions.

Diodes between input and output are usually not needed. The internal diode between the input and the output pins of the LT1083 family can handle microsecond surge currents of 50A to 100A. Even with large output capacitances, it is very difficult to get those values of surge currents in normal operations. Only with a high value of output capacitors, such as 1000μF to 5000μF and with the input pin instantaneously shorted to ground, can damage occur. A crowbar circuit at the input of the LT1083 can generate those kinds of currents, and a diode from output to input is then recommended. Normal power supply cycling or even plugging and unplugging in the system will not generate current large enough to do any damage.

so it looks like i will need the diode between in and output on the LT1033 and LT1085. But i can drop the diode across the adjust pin on both regulators.

The next question is can i use a 4148 diode with a 24R resistor in series?

My back of the envelope calculation indicates that i can:

The max output voltage of this circuit is 20 volts. 20 Volts across a 24 ohm resistor give 0.83 amps. The 1N4148 IFSM = 1.0 A (Pulse Width = 1 sec), 4.0 A (Pulse Width = 1 uSec) (Non-Repetitive Peak Forward Surge Current).

It seems like the IN4148 can handle the current associated with the maximum possible load it might see with a bit of headroom left over.

[okapi]

Originally posted by okapi


What is the next best option? How do i decide what grounds to bring together before i run them to a central ground? Where should this central ground be located? Should it be closer to the rectifier bridge or at the load?

thank you for your help.

The data sheet also shed some light on the question i posted earlier. It indicates that R2 (the second resistor in the voltage divider) should be connected to the positive side of the load separately from the positive (ground) connection to the raw supply.

i have updated the schematic to reflect this change and correct for the error bakmeel pointed out.

it still remains to be decided where i run the grounds for the input and output capacitor banks?? Should they all meet at the load? That would require 3 or 5 (if i keep the + and - ground rails seperate) individual wires running to the a star ground located at the load. Will this solution lower the noise floor? What are some other good solutions?

[okapi]

I have come up with a preliminary circuit board design for the tracking regulator. It does not include the high stability circuit but the LM329BZ's arrived today. once i test a prototype circuit i may include the high stability regulator in the final PCB as a optional configuration.

i used expressPCB's proprietary software to make this. Express PCB charges approximately $60 dollars for a single board and each additional board is about $12. If i can shrink the length by about half an inch i can use their MiniBoard and get three boards for $51. But then the width would be about half an inch too large.

this is my first circuit board design so any helpful feedback would be greatly appreciated.

on a side note, from reading a more recent LT application note it seems that bypassing the output with a cap that has a relatively high ESR (the LT1033 application note recommends a tantalum cap) is only necessary when the output voltage is low (less than 5 V) so it looks like ceramics will be the way to go (as bakmeel already kindly pointed out).



red = top layer
green = bottom layer

[okapi]

1. i have updated my PCB design for the tracking regulator. The major change is the implementation of a star ground. i have also included the option of running the adjust pin ground to the load or jumping it to the ground on the board. The plan is to run the load ground back to the star ground on this board.

2. i have successfully prototyped the high stability version of the tracking regulator which uses the LM329B voltage reference. I will post the schematic later. Resistor values needed to be tweaked to get it to work correctly.

3. i would like to test the noise level of the regulator, however, my bench supply (agilent E3630A) has about 5 mV peak to peak noise. There may be higher amplitude high frequency noise but my scope only goes to 20 Mhz so i can't see it. The tracking regulator noise is lost in the supply noise. I was thinking of using a bank of 9V batteries instead of my bench supply to get the noise level down.

here is the latest pic of the PCB