High Voltage Regulators (Maida or zener)

[valveuser]

Thanks for the feedback.

Always get confused with these FET symbols (drain to right, source left) but SPICE agrees D3 is the right polarity. It doesn't really do anything except protect the gate when voltages get dizzy.

You're right, R6 is high. It just feeds the zener Z4 to make sure there's 15V between gate and output. Didn't want it getting warm on start up. 30K would be a better value.

If you have LT spice, the LM317 model can be found here;

How to use LM317 LTspice Model

Maybe it's from the tech. groups at Yahoo;

http://tech.groups.yahoo.com/group/L.../files/%20Lib/

In case you don't wish to dig or mess around, I've zipped the files below, and included the .asc file. Don't know if that's sufficient for it to work.

LM317.zip

soft start 317 regulator.asc

Save symbol file as LM317.asy, in lib\sym\misc
Save model file as LM317.sub, in lib\sub

Regards,

Bob

[FLT]

Originally Posted by valveuser

Thanks for the feedback.

Always get confused with these FET symbols (drain to right, source left) but SPICE agrees D3 is the right polarity.

Yes, i was wrong. It is late and i need some sleep...

regards

[tomchr]

Originally Posted by valveuser

I came here to find a ready made PCB for an LM317 HT regulator with soft start. Nothing on the web it seems. I'm thinking of a design more or less straight from the data sheets like this.

Can anyone see anything wrong with it. It looks like it should even withstand a shorted output. Haven't tried it yet.

With 200 V in and a beefy cascode, you may be able to get the regulator to survive an indefinite short circuit. But note that the "current limiter" of the Maida regulator is really ill-defined and soft. Basically, what happens is that the LM317 runs out of headroom as the voltage across the cascode and input resistance to the '317 gets too large for the '317 to maintain regulation.

I've learned that there are a few things to watch out for with these high voltage regulators.
1) They need to survive start-up.
2) They need to survive momentary connects/disconnects of the load.
3) They should regulate the output voltage and remove ripple.
That turns out to be a pretty tall order. I got 1) and 3) taken care of in my circuit and if you add an external cap on the output and perhaps a snubber across the pass device, you can nail 2) as well.

I'm currently out of stock on my regulator boards so I'm trying to decide where to go next. I may make another revision that has some of the suggestions for improvements I have received implemented on the board. But I'm also considering revisiting the Maida Regulator. It's a pretty rugged regulator and works quite well for low voltages, currents. It would be nice with a beefed up version.

~Tom

[valveuser]

But I'm also considering revisiting the Maida Regulator. It's a pretty rugged regulator and works quite well for low voltages, currents. It would be nice with a beefed up version.

If you do go there, Tom, it would be useful to make a versatile generic Maida PCB which makes provision for;


1 A soft start pnp such as MPSA92 or higher voltage equivalent. The nodes are already there in the Maida circuit. It just requires 3 extra holes. Simply don't use them if soft start isn't wanted.
2 A current limiting resistor before the 317 input.
3 AC input, i.e. full wave input rectifier diodes with optional series resistors to provide the low noise equivalent of a valve rectifier. This followed by a beefy smoothing capacitor.
4 Two optional RC output filters such that outputs can be directly wired to a valve amp or to each channel of a preamp.


I'd buy a couple of them to help spread the cost. I'm thinking of this, perhaps with other optional add-ons;

[valveuser]

Actually, R6 in the previous circuit is redundant and D5 should have been connected directly to gate anyway, so I've removed it and re-named the resistors.

Enough for now. It's best that I build and test the circuit to see if it's worth pursuing any further. I'll order parts tomorrow and, hopefully, report my findings over the weekend. I'll monitor;

The soft start performance.
Rapid on/off switching from a power supply.
The effect of stroking the load lead against the output terminal.
Crowbarring the output (with optional RC filters disconnected).

Bob

[valveuser]

I built the prototype today and tested it without a load. The attached images show the prototype and the actual component types and values used.


It appears to work well. The soft start behaves very much like a valve rectifier, gradually increasing output voltage (over about 10 seconds with values shown). It isn't a linear ramp as would be the case with a P channel MOSFET since the pnp base draws a little current, and the timing resistor value needed to be approximately halved to get a similar time constant. LTSPICE predicts this.


It gracefully changes voltage state on repeated switch-on/switch-off of the unregulated (but smoothed) DC supply used.


Enough for today. Tomorrow, I'll try and destroy it.


Regards,


Bob

[tomchr]

I probably wouldn't include the input rectifier and reservoir cap. Those are very application dependent. The reservoir cap is also huge - in my setup at least half the board area if not more would be reservoir cap. So including those caps would basically double the board cost. These components are easy to place elsewhere in the chassis and easy to mount on a piece of vector/vero board.

I hope you're not in a rush. Nothing around here seems to happen in a hurry... Between work (you know what pays my parts budget ) and hockey, there isn't much time for regulator development. That said, I've been known to suddenly get the bug and get stuff cranked out. Stay tuned.

My main challenge is that I need a regulator that'll provide 470 V, 200 mA without blowing up when loaded by a capacitive load. Accounting for mains variation, this means upward of 600 V in... Lots of energy to blow stuff up.

~Tom

[valveuser]

I've zipped and attached MPSA92.sub from the data provided in the Fairchild datasheet to paste into /lib/sub, though it might have been captured in the attached .asc files. Let me know if it's redundant.


The simulation shows that switching a 10K load in and out causes no problem and this was then confirmed with the prototype by stroking a 10K load terminal against the output. From this, it can be concluded that loose valve connections in a preamp wouldn't destroy the regulator.


However, the simulation predicts that shorting the output places extremely high current spikes through the protection diodes and zeners. To simulate, change R_load to 0.001 ohm. A practical test obviously wasn't carried out. This doesn't question the requirement for a current limiting resistor in the design because this is a crowbar scenario.


The workaround is to use an RC filter at the output as previously suggested as being optional in my earlier recommendations for a PCB layout. Obviously, an RC filter is mandatory if shorts are expected or where the quiescent current is high and interrupted by loose valve connections. In this example, with R_filter = 20R and C_filter = 100uF, current surges in the diodes and zeners are negligible when the output is shorted. An RC filter with f-3dB of around 100Hz would be a good design choice. A practical test with RC filter in place wasn't performed for obvious reasons.


I've attached another .asc file to simulate input switching. Normally, with the required smoothing capacitors providing a slow voltage ramp up, current surges through the protection diodes and zeners are negligible. This example shows what would happen if the time constant was ridiculously and impractically small. I only include it to demonstrate that contact breaking (unlike making) causes no problem.


In summary, a successful exercise and quite suitable for a preamp power supply, so I'll leave it there and let Tom adapt it for use in power amps.


... and I still haven't got the PCB I came here for.

I probably wouldn't include the input rectifier and reservoir cap. Those are very application dependent. The reservoir cap is also huge - in my setup at least half the board area if not more would be reservoir cap. So including those caps would basically double the board cost. These components are easy to place elsewhere in the chassis and easy to mount on a piece of vector/vero board.

Input rectifiers and a smoothing capacitor could be included on a preamp PCB, Tom.


Anyone else interested in making a soft start preamp PCB to my previous specification?


I've designed what promises to be a really good DAC amp and I'm keen to build it - two stage feedback design using ECC88, critically damped HF with subsidiary DC feedback. All capacitors inside the feedback loop. I'll post on a separate thread if it ever gets built.


Bob

[valveuser]

I've just re-read my last post and my conclusions were inadequate and one statement was wrong so I thought I'd correct that before I leave it.


I attached the files in the wrong order which may cause confusion. The switched input .asc file should have been attached last, rather than first. This scenario is unlikely unless there are bad contacts after the smoothing capacitors, or if input connections are flashed onto a live DC power supply. The point here was that an abnormally abrupt making, rather than breaking, of the circuit was the problem.


The soft start power supply with the protection devices shown is fairly bullet proof under normal conditions. The main problem is an unlikely short circuit as shown in the second attachment (which should have appeared first).

... or where the quiescent current is high and interrupted by loose valve connections.

The above statement I made is wrong. An open output circuit doesn't appear to be an issue. Neither does re-closing the circuit with the correct design load. This being the case, loose valve base connections wouldn't be an issue.


In the shorted output scenario, it seems that the quiescent current is largely irrelevant, whether it be for a preamp or a power amp. The high current surges through the protection diodes and zeners are a direct result of the LM317 output being rapidly pulled down by the short causing capacitors to discharge through them. An RC filter on the output would slow this process as seen up front which would probably protect the components, although it might damage the capacitor. A foldback current limiting circuit instead of the current limiting resistor, as suggested by Tom and others, can also eliminate the problem.


Irrespective of the quiescent current, where the output is not rapidly pulled below it's normal regulating voltage, there wouldn't be damaging high current surges through the protection diodes and zeners.


I hope this clarifies and corrects. Sorry for any confusion.

[valveuser]

I started to look at ways of bringing the regulator output line down gracefully in the event of a short circuit to prevent the predicted high current surges through protection diodes and zeners.


My simplest idea was to use an inductor, and LTSPICE showed that 50uH was sufficient (reactance 3.14 Kohms at 10MHz), the DC resistance not being too important. Eager to try it out, I robbed one from a scrap valve oscilloscope and soldered it to the output of the prototype I'd made.


I removed the 2.7R resistor in series with the 1uF capacitor from the prototype since it typically represents the electrolytic capacitor's series resistance anyway, and I set all capacitor series resistances to 3R in the LTSPICE model.


The model predicted a surge of 17A during a short so I switched the prototype on and braced myself for a good spark. It was disappointingly small. Many repeated shorts were carried out at different times during the voltage ramp without damaging effect.


An RC filter typically comprising 22R and 100uF could be placed before or after the choke, according to SPICE, or an (R+L)C filter could be used.


It's interesting that the circuit diagram for the scrapped oscilloscope shows that these variable 50uH inductors were used in series with 4.3K or 3.9K load resistors in ECC88 differential amplifiers and tuned to produce optimum square wave response. That noted, the protection inductance could be used to enhance the performance of the amplifier in many designs.


So there it is. A high voltage soft start regulator, tolerant to shorts and now looking very bullet proof.