LT1431 + DN2540N3 + D44H11/D45H11 Dual Rail Reg
Here is a dual rail regulator circuit idea if anyone would like to mess with a dual rail TL431-like regulator. I'm posting this as a follow up to a post I did before the holidays of a TL431A + FET regulator in the thread: http://www.diyaudio.com/forums/power...ml#post2405784 . In addition to noise, a problem with that circuit is 30mA through the TL431A pull down resistor, a by-product of switching from a DN2540 depletion mosfet (about a volt across the pulldown) to a FET (about 4 volts across the resistors).
After some attempts at increasing the values of the TL431A pulldowns to reduce the current, it turns out that current can't be reduced much below 20mA or the circit becomes unstable (high frequency oscillation). The Linear data sheet says that in addition to the capacitive loading on the output, the 3 terminal version needs fairly low voltage gain for compensation. So to solve that issue this circuit uses their LT1431 part which breaks out the open collector transistor as a pin and an external compensation network as a pin. Their 3 terminal "Z" version of the part is just the TL431/LM431.
Digikey and Mouser don't have the LT1431 last I looked, but Arrow's online site at arrownac.com has it for $1.50 plus shipping.
The simulation plots below are for 5Hz, 20Hz, 20kHz, and then 200kHz just to verify high frequency stability. I've built and tested 5Hz through 35kHz but haven't tested the 200 kHz end so don't know if that high end is real or not.
I've gone back to DN2540 depletion mosfets from the JFETs. The part is available in both a N3 small signal 120mA version and a N5 500mA version which is kind of cool. The regulators are self-starting since the depletion mosfets are on until told otherwise.
The positive rail is pretty straight forward. Linear Technology recommends a 150 ohm resistor in series with a 0.015uF capacitor for the compensation network in the data sheet. I found that I had to use at least a 0.02uF, and I have shown 0.033 to account for capacitor tolerances. The output tantalum capacitors are also necessary minimum values for compensation.
The negative rail is a little more tricky due to p-channel depletion mosfet parts not existing. Here the DN2540 returns to ground. The Sziklai transistor Q3 is not strictly necessary - the DN2540 is well withing bounds at the 12mA-or-so of maximum base current from the D45H11 at full load - but without it all 12mA gets wasted returning to ground. With Q3 that current contributes to feeding the load.
Another negative rail trick is that the LT1431 needs at least 2.5 volts from the open collector to ground, hence the voltage level shifter made up of the DN2540 and the string of white (blue should work too) LEDs. I used the LED string instead of a zener for presumably less noise. As far as the negative rail LT1431 is concerned the negative rail is "ground", which works just as long as everything is consistently referenced like this.
I haven't measured or tried to calculate noise. I don't expect it to be much - if any - better than a TL431A, so this is probably not a low-noise regulator. I'm using it in a couple of powered bookshelf speakers so for my application I can't even hear a difference. But probably not the best of supplies for a pre-amp.
Green = positive unregulated rail
Blue = negative unregulated rail
Red = positive regulated rail
Aqua = negative regulated rail
Magenta = voltage at positive test load (0.8A max load current)
Gray = voltage at negative test load (-0.8A max load current)
AC sweeps are of the positive and negative rails from 1Hz to 250kHz
LT1431 + MOSFET Dual Rail Voltage Regulator
Here is another circuit idea for experimenting with a dual TL431-like supply. This is a version of a LT1431 dual rail voltage regulator using enhancement power mosfets for the series pass devices rather than the BJTs.
In this one I've used a CRC PI filter for the unregulated voltage input, but really no necessity for that. The error amplifier loop is fast enough to follow the incoming sawtooth wave with no problem, which is why I just used the single capacitor input filter on the BJT version. But the PI filter can't hurt through, minus the small voltage drop across the "R", reducing the input ripple and rounding off the sawtooth a bit.
Since the TL1431 has an open collector transistor as a separate pin from the V+ for the error amplifier, the pulldown can actually be returned to a different voltage up to the 37V max for the part. I've taken advantage of that here to return the pulldown to the unregulated supply for the extra 4.5V the gate of the mosfet will need over the 18.0V regulated rail. Returning this to the unreg side doesn't add ripple since there is already a inverse ripple signal at the gate from the error amp, as the sims show.
The V+ for the error amp is then returned to the 18.0V regulated side. I've used a DN2540 depletion mosfet again to insure start-up of the positive rail, since the error amp V+ is fed from the regulated side. At t0+ current will flow through the normally-on depletion mosfet to the rail via R4, allowing the voltage to turn on the pass element and start the loop up. I've picked values to run about 1 mA through the depletion device at the operating point.
For the pulldown R6 I've picked a couple of mA. The TL431 specs are clear about 1mA minimum, but the LT1431 specs are less clear. They list 1 mA, but I'm assuming that's the error amp and not a collector cutoff current of the open collector transistor (sounds high for that). That resistor could likely be increased ten fold to 27k and the pulldown current reduced then to 250uA, since all that is being driven is the gate of the depletion mosfet, but I haven't tried it.
The reference voltage divider of R1 - R3 also acts as a default load on the postiive rail in case the driven load is disconnected. The error amp part of the TL1431 also apparently pulls around 1mA from the spec sheet. Same is true for those parts on the negative rail.
The test load uses the same power mosfet as the series pass element. I've dropped the (resistive) test load to 4 ohms to pull a maximum of about 3.8 amps. This leaves the mosfet still within the safe operating area from the data sheet, by my read, when using RMS values, but not by all that much. It would be easy enough to parallel mosfets for the series pass with appropriate small source resistors for more current.
For the negative rail I was able to use a single PNP transistor Q1 to insure start-up, and also to provide the voltage inversion needed (the pull down resistor needs to pull "up" here). Following the negative rail loop around, if voltage across R14 becomes larger (more negative) than -2.5V then the negative rail has gone lower than -18.0V. Since the negative rail LT1431 thinks the rail is ground, and it is inherently a shunt regulator, the part responds by pulling down R9 with the open collector transistor to "lower" the voltage (thinking in positive rail terms).
Q1 is then turned on harder, increasing the voltage drop across R12, making the gate of the p-channel enhancement mosfet M6 less negative and turning it off, in turn making the rail less negative and restoring regulation to -2.5V across R14.
In the plots the key is:
green = unregulated pos rail before PI filter
pink = unregulated neg rail before PI filter
blue = unregulated pos rail after PI filter
gray = unregulated neg rail after PI filter
red = regulated 18.0V rail at 4 amps
green = regulated -18.0V rail at -4 amps
aqua = pos rail load voltage at load frequency at mosfet drain
dark blue = neg rail load voltage at load frequency at mosfet drain
magenta = pos test load current through 4 ohm resistor, scale is 0 to 4A
orange = neg tes load current through 4 ohm resistor, scale is 0 to -4A
The plots are done at load frequencies of 5Hz and 20kHz.
Here is a revised LT1431-based dual rail mosfet regulator schematic. After staring at the print I posted a bit more I realized the DN2540 depletion mosfet on the positive rail is completely unneeded. :rolleyes: The gate and attached pulldown resistor can be connected directly to the open collector of the positive rail LT1431. Guess I was still thinking BJTs.
The resulting circuit isn't substantially different from a couple of regulator application circuits on the LT1431 data sheet, except the chip is on the regulated side here, the pass transistors are mosfets, and I've placed a transistor gain stage between the LT1431 open collector output and the mosfet gate - the extra gain improves load regulation at the likely expense of amplifying the chip's noise. Q2 and Q3 can be removed and the mosfet's gate connected directly to the LT1431 open collector, but the compensation network will have to be moved back to ground due to the larger voltage swing.
I've changed the mosfets to a better matched pair and replaced the capacitance on the load side with single 10uF 35 tantalums. I've also marked voltages and currents at various places on the print.
The plots for this circuit are done at 5Hz and 200kHz. The AC sweep is 1Hz to 250kHz.
|All times are GMT. The time now is 01:10 PM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio