LR8 high volltage linear regulators (2023 Update)

Are you aware of the inherent difficulty in stabilizing high- side current-amplification circuits? The regulator has an internal high-gain feedback loop, whose gain drops 20dB/decade with frequency, with an associated 90-degrees phase shift.

. ,--- S D -----------, . | G _______ | . | | | | | . ---+-/\/\-+---| |--+-+----+---- . |_______| | _|_ . adjust | | --- Cout . gnd -- pot --/\/\--+--/\/\--' | . gnd

The external amplifier, a P-channel MOSFET or PNP transistor (although a PNP is not preferred because it suffers from severe SOA limitations), adds voltage gain inside the feedback loop. This gain also drops by 20dB/decade with frequency, and has an associated phase shift, due to the high capacitance load on the output. Recall that Gain = gm * load In this case the load is a reactance that drops with frequency and has a 90-degree phase lag.

The high gain and combined phase shifts in a feedback loop with two poles, as we call it, can lead to high-frequency oscillation. In the case of an LM317, this possibility is moderated by the high speed of its error amplifier and by a fortunate assist from the internal self-resistance of the large electrolytic output capacitor. This resistance, called esr, adds a pole-canceling zero essential to the safe operation of this type of circuit.

Evaluating this scene for an LR8, we can suppose that its error amplifier does not have the speed possessed by the LM317, and may therefore be more susceptible to oscillation. We observe the LR8's transient response is considerably slower than the LM317.

You can add a little series resistance to isolate the capacitor and move the zero to a lower frequency, solving the oscillation problem. This comes at the expense of degraded load regulation, with a voltage drop outside the feedback loop, but that may be acceptable in some high-voltage regulators.

. ,--- S D -----------, . | G _______ | . | | | | | . ---+-/\/\-+---| |--+-+--/\/\--+---- . |_______| | _|_ . adjust | | --- Cout . gnd -- pot --/\/\--+--/\/\--' | . gnd

BTW, when testing circuits like this for stability, be sure to test with the fill range of load currents, that's because gm and therefore gain, increases nearly proportional to current. It's useful to use the regulator's response to load-current steps to evaluate its loop stability, looking for ringing, etc. One other comment, Spice modeling will not work for such a circuit, unless you correct your power MOSFET model for sub-threshold operation.

One attractive possibility is to beef up the regulator's current capability with an N-channel power MOSFET source follower, rather than a P-channel high-side amplifier. First, this greatly reduces the danger of oscillation, also very important, you can get bigger N-type than P-type high-voltage power MOSFETs,

. ,---/\/\---- D S ----, . | _______ G | . | | | | 680 | . ---+--| LR8 |---++-/\/\-+--+-- . |_______| | _|_ . | | --- Cout . ,--/\/\--+--/\/\--' | . | gnd . '--- pot --- gnd

Although the source follower lowers the output voltage by an uncertain load-dependent amount, typically three to four volts, this isn't so painful for high-voltage regulators, where you will no doubt have an output voltage-monitor meter anyway.

A power resistor in the MOSFET's drain can help reduce its heating at high currents, and provide a crude current limit. You can also easily add a true current limit,

. . ,---/\/\---- D S -+-/\/--, . | G | | . | | B | . | _______ +- C E ---+ . | | | | | . ---+--| LR8 |---+-+--/\/\---+--+-- . |_______| | _|_ . | | --- Cout . ,--/\/\--+--/\/\---' | . | gnd . '--- pot --- gnd

This is best done by adding two resistors to make a "foldback" current limit, which reduces the maximum current during a short circuit, where the highest power dissipation would occur. At high voltages, with Vout up near Vin, Icl = Vbe/Rs primarily determines the current limit, but at lower Vout, R4 and R5 subtracts from Vbe in the equation. You can choose R4/R5 to reduce the current limit for a short to half or even 1/3 the full level. This protects Q1, substantially reduces the heat sink size, and may eliminate the requirement for a fan.

. High-voltage high-current regulator . Q1 . ,- Rp -+----- D S ----------, . | | G | . | '- R5 -- | ---+-- R4 --+ . | | | | . | | B Q2 | . | _______ +- C E --, Rs . | | | | | | . --+---| LR8 |---+-+-- R3 --+--+--+--- Vout . Vin |_______| | _|_ . | | --- Cout . ,-- R1 --+-- R2 ---' | . | gnd . '--- pot --- gnd

One other thing, be safe and add a protective 10V zener across Q1's gate to source.

I was a fan of Intersil's HIP5600, a bipolar linear IC, which sadly was discontinued some time ago. But I still use it in my designs, drawing sparingly from a lifetime-supply purchase.

--
 Thanks,
    - Win

(snip comprehensive reply)

Great answer (kinda restores faith in usenet), thanks Win. cheers M (Looking forward to the next edition of "The Art..")

As you can see, I compared two ways of enhancing high-voltage low-current three-terminal programmable regulators, one with excess gain inside the feedback loop, and a second without but with hopefully-modest output voltage errors. There's a third attractive approach using low-voltage high-current regulators, like the LM317 or LM317L, and cascoded series voltage-dropping MOSFET(s) that always present the low-voltage LM317 with about 7V more than its output, for safe cool operation. I, and others, have written about these in earlier s.e.d. posts, complete with ASCII drawings. Here's an example from January 7th, 2002,

From: W>

The ST VB408 etc., parts that Xenos suggests are nice, but they may be hard to get. You may be interested in a simple high-voltage regulator that uses standard LM317 or LM317L chips plus an easy-to-get reliable high-voltage MOSFET.

600V FET _____ LM317L 5 to 500V IN o----+----, ,-+-----+--| |--+---+----+---o OUT | | | | s | | | | | | to 100mA | _|_V_|_ | |_adj_| 1.2k | | 2.2M ----, | | | | | 0.47uF 1/2 W | '-||--+-----+ | === | | , 68pF | | | 630V '--------+---| (Looking forward to the next edition of "The Art..")

Groan.

--
 Thanks,
    - Win

Please can anyone explain to me how to read the ASCII schematics in this post? Thanks

Welcome to Usenet, which has been around since 1979. This group (sci.electronics.design) is older than Google. The short answer is, "You can't do it on the Google Groups web interface."

Unfortunately Google collapses multiple space characters into one, which destroys schematics and other ASCII art.

Ditch Google Groups and use a real newsreader such as Thunderbird or Seamonkey.

You can get a free Usenet account from eternal-september.com. Setting it up is pretty simple--you'll be on the air in half an hour.

Cheers

Phil Hobbs

From a post made in 2006? No, I can't. I don't have access to it, I am not using Google Groups as you do.

You need to display them in a terminal using fixed width fonts, as I do. Not a web browser. And download them from an Usenet server that contains that old thread, not from Google Groups. I don't know any that has that old content.

For those on Usenet, it is this thread:

Yes, but the other problem is getting that ancient post. Most Usenet servers probably don't.

Mine doesn't, or my client expired it.

On 2023-03-09 07:52, Carlos E.R. wrote:> On 2023-03-09 13:37, Phil Hobbs wrote: >> On 2023-03-09 07:25, Jeremy Morgan wrote: >>> On Monday, 31 July 2006 at 00:57:28 UTC+1, Winfield Hill wrote: > ......................**** >

You can get 20G for $10 of downloading from astraweb, which goes back to

2003. (I splurged and got 180G for $25, so I should be reading SED till about 2523 AD. ;)

Supernews also has stuff back to 2003.

The Internet Archive has mbox files of a number of groups going back about that far, including this one.

At the moment I have some Seamonkey config problem that I haven't bothered tracking down, that prevents me from _posting_ via Astraweb, but it works fine via Supernews or ES.

Cheers

Phil Hobbs

If you're prepared to faff-about highlight the ascii schematic lines and (if using firefox) right click and select view source. Copy to vim and replace the <br> with returns (%s/<br>/\r/g).

Thanks for your responses and suggestions.

@JohnMay Your suggestion re: View source worked a treat thanks. I was able to paste it into a text editor in Windows. Using a monospaced font (Courier) made everything line up OK and I was able to make sense of the schematics.

It would be readable in a fixed-pitch font, with some line wraps fixed.

That circuit is probably the one on page 609 of Win's book, AoE 3rd Edition Improved. Everyone should have that book, and the X-chapter supplement (to which Phil and I made modest contributions.)

Given your requirements, we could play with HV regulator ideas here.

Readable version of Win's followup:

<snip>

Close, but no cigar. Although Win apparently cooked up the above circuit with ideas presented on page 608 - 609. What! Wait. Win's circuit went to pot as they say. So it's no good, right John? LOL. Regardless, this thread illustrates the virtues of hosting your own usenet archive instead of outsourcing it to the slop served by Big Tech.

Danke,

English translation, please?

Wow :-o

You preach potentiometer perdition, no?

Danke,

+1

(It deserves more than a plus 1, maybe a plus 1 times the number of characters in Win's schematic. :-) ) Ed

Yikes. The poor pot is going to have to dissipate 500 mW and drop 500V. That's your unobtainium Allen Bradley carbon unit there.

Cheers

Phil Hobbs

<snip>

Nobody sane would use a carbon film to dissipate 500mW.

A cermet part would work fine. The trouble with carbon film is the negative temperature coefficient of resistance. Treat them wrong and you can form a low resistance hot channel in the film. I've seen it done with a 10k carbon film resistor, and - worryingly - the resistor still measured 10k afterwards, though there was a thin dark line in the paint over where the hot channel had been.

Right. I was giving the +1's to Don's post with the link that contained the schematic, not the circuit. But I sure didn't make that clear.

Ed