0-70V, 0-5A Lab Power Supply Design

[H713]

For testing audio amplifier designs, it is useful to have a fairly beefy bench power supply. My current supply is capable of 10-34V at about 3 amps, and it has proven to be insufficient. I have decided that it is time to build a bigger and better linear bench power supply. Here are my minimum requirements. As a side note, this is by no means my first power supply design project.

1) 0-70V
2) 0-5A
3) Extremely now noise/ripple. I'd like to see less than 2 mV peak-to-peak.
4) Linear, which complicates #5
5) Dual power supply in a 2U rackmount enclosure.
6) Fan-cooled is acceptable
7) Adjustable crowbar circuit to avoid damage to low-voltage circuits (like mixing console channel strips) in the event of a catastrophic failure. This will be a manual control that will likely see seldom use, however, it is essential to have for this application.
Obviously it is impractical to dissipate 700W in the series pass elements when operating at low voltages and high currents. As such, some form of "pre-regulator", which will probably come in the form of a Triac "dimmer" circuit on the primary of the power transformer, will be necessary.

Even still, the pass elements present an issue because when the output is shorted while it is set to 70V, the main filter caps will take some time to bleed down after the pre-regulator adjusts for the reduced output voltage. Traditionally these would be a BJT in a TO-3 package. I have an alternative idea, however.

An application note from International Rectifier suggests that for linear operation, older high-voltage MOSFETs with a high R_dson will be less likely to suffer from secondary breakdown. Testing of the FQA8N90C and the 2SK3675 verified this. A single 2SK3675 or FQA8N90C on a substantial heatsink was successfully able to handle 40V at 6A in its linear region. None of the BJTs I tested came close. I was not successful in killing either of the MOSFETs that were tested. I do intend to test with the full 70 volts, however, I need to dig a larger transformer out of storage in order to do so.

I haven't seen any linear power supplies using MOSFETs in this way, so I'm curious about what I'm neglecting to consider here. I was planning to use six of them in parallel for each of the two power supplies, and they will not be seeing the full 70 volts dropped across them for more than a second.

Here's the basic idea I've been simulating for the regulator circuit. Nothing complicated, and I'm using the IRFP240 and 2N3019 SPICE models because they're fairly generic. This is a generic schematic, obviously there will need to be changes made so that it is stable into capacitive loads, etc. R7/R11 and R15/R16 will be replaced with pots. Yes, there will be limits imposed to ensure you can't set the thing to current limit at 20 amps.

[moffy]

Best way to do the variable voltage while keeping noise down and power dissipation, is to have a multitap transformer, that changes taps in say 10v or 20v increments. You need a controller for the relays but it's fairly straight forward.

[H713]

I thought about doing tap switching, but that would require a custom transformer. I agree that it would be the simple and robust way to do it and I may still see if it can be an option without having to wind a custom toroidal transformer.

[moffy]

Yeah, a bit of a pain unless you use multiple say 5A/20v (secondary) transformers in series. But then packaging is a pain.

[Kleinstein]

It is very common to get transformers with dual output windings, like 2 x 24 V. So a single transformer would already give 2 voltages.
With only 2 U hight one does not have the hight for a single large transformer anyway. Keep in mind that continuous 5 A DC need something like a 8 A AC rating, unless extra PFC is used.
With 2 transformers for each half one would have 4 taps (possibly more with different voltages). Even just 2 taps nearly cuts the worst case loss to halve. In addition it reduces the voltage seen by the transistors which helps with the SOA.

For the tap switching it is still the question if one wants relays on the AC side or an electronic version at the DC side (e.g. 2 transistors in series). Both have there pos and cons.

Normally an audio amplifier has to work with a normal unregulated supply, so there is no real need to have super low noise for the high power part. One may not need / want normal lab supply grade current limiting. There can be extra peak current, and once the voltage drops too much chances are one would prefer a full off than holding the peak current. This can reduce the worst case power loss quite a bit.  I would consider the regulation more for getting adjustable voltages and to have some protection.

There are some HP supplies that use MOSFETs in linear supplies. It is not that unusual.

[CHANDRASEKARAN.V.T]

I am not sure if I can post my requirement which is as follows:
 I need to design an ISOLATED DC-DC CONVERETER WITH 1KV ISOLATION:
INPUT:12VDC+/-1V
OUTPUT 1:12VDC---30ma
OUTPUT 2:12VDC---30ma
Efficiency is not a serious consideration.
Wish to use LM3524D as the control IC  with feed back.
But designing this transformer with a toroid is my problem and can I get some support.
Regards
VTC
9980522856
BENGALURU,INDIA

[TheHolyHorse]

I am not sure if I can post my requirement which is as follows:
 I need to design an ISOLATED DC-DC CONVERETER WITH 1KV ISOLATION:
INPUT:12VDC+/-1V
OUTPUT 1:12VDC---30ma
OUTPUT 2:12VDC---30ma
Efficiency is not a serious consideration.
Wish to use LM3524D as the control IC  with feed back.
But designing this transformer with a toroid is my problem and can I get some support.
Regards
VTC
9980522856
BENGALURU,INDIA

Instead of hijacking this thread, start another one related to your project.

[Zero999]

2mVp-p ripple on 70V is an insanely crazy low specification of under 0.006%. Is it even possible? Just use batteries.

I don't see why the original poster needs all of the above. A decent audio amplifier should be fine with an unregulated power supply consisting of a large transformer and filter capacitor. A variac, with a suitable mains transformer, rectifier and smoothing capacitor should be sufficient for testing most audio amplifier designs.

Any pre-regulator circuit will introduce a significant amount of noise and ripple on the secondary side. Worse still it will be high frequency ripple, which is less easily to get rid of, as the supply rejection of op-amps tends to decrease, with frequency.

[capt bullshot]

Obviously it is impractical to dissipate 700W in the series pass elements when operating at low voltages and high currents. As such, some form of "pre-regulator", which will probably come in the form of a Triac "dimmer" circuit on the primary of the power transformer, will be necessary.

Even still, the pass elements present an issue because when the output is shorted while it is set to 70V, the main filter caps will take some time to bleed down after the pre-regulator adjusts for the reduced output voltage. Traditionally these would be a BJT in a TO-3 package. I have an alternative idea, however.

I'd recommend against a triac or SCR based pre-regulator if you want to achieve low noise etc. Your main regulator should be able to cope with the ripple produced by the pre-regulator, as it would have to do so anyways. But a "dimmer-like" circuit can produce large stray magnetic fields at line frequency repetition rate having high frequency content (depending on your transformer / choke / bulk cap) that might couple into sensitive audio circuits. So you'd better go for a tapped transformer.

[coromonadalix]

Why dont you simply buy like : an sorensen xantrex kukusui  or smps based psu ??/

It will be simpler and some of them already have or are near the needed specs ??

Sorensen DCS80 - 13E

Kikusui PAD70-5

you have this one selling right now for an incredibly low price ?? untested an weight a ton  loll
Sorensen DCR 80-20B
https://www.ebay.com/itm/Sorensen-DCR-80-20B-Variable-Regulated-DC-Power-Supply-0-80VDC-0-20A-1800W/113890939218?

My kikusui pad70-5 give 1 -13 mv of ripple under a full load,  sure the psu has been re-capped, on par with the manual,   but i've paid 50$ usd to have it minus the shipping
Just push a button and adjust the curent limitter, i  love this.
https://manual.kikusui.co.jp/P/PAD_L_TYPE1_2_E.pdf

My Kepco ATE 75-8  is slightly noisier,  have to re-cap it. It has an scr crowbar protection

But a 2mv ripple    use batteries as primary source.  And sure many of theses psu are kinda smps or a mix of smps and phase controlled psu  since you're going over 500w and up.

An linear one wil give tons of heat and power losses ....

[Miyuki]

I have a lab supply what have classic output range 2x 36V 2A with specified 1mVpp ripple output and some fixed outputs +-15V with 2mV ripple at full load
It uses thyristor preregulator with linear post regulator
But it is huge beast and as new cost a fortune and there is not much what can be done smaller when want do it linear with mains transformer and reliable   

[schmitt trigger]

You don't need a custom transformer to achieve several raw DC voltages, just a dual secondary transformer and a bunch of relays:

Low range: secondaries in parallel, FW bridge rectifier
Mid range: secondaries in series, FW bridge rectifier
High range: secondaries in series, voltage doubler rectifier

I'll let you figure out the relay switching logic. Probably best achieved with a small microcontroller.

[H713]

Ripple isn't a huge deal up at 70V, and I should have mentioned that it is much more of a concern in the 24V range where it will likely be used to test class-A mixing console circuits. Given that the 0-500V linear supply I built had ripple below 10mV under load, I feel that I should be able to get this significantly lower.

I should note that I have decided to build this rather than purchase a Sorensen or Xantrex PSU because going onto eBay and clicking "Buy" isn't much of a design exercise, is it?

Regarding noise from the Triac/SRC "dimmer" pre-regulator: I was going to use an LC network to try and filter out the worst of the high-frequency noise before the primary of the main transformer. That said, I will be looking more seriously into tap switching since it is a much simpler solution.

I can try a doubler circuit for the high range, but I suspect that the voltage sag under load will be problematic unless I use excessively large filter capacitors. I need to do a few calculations to see if it's going to be realistic. If it is, that solves the problem.

[David Hess]

Since I try not to require regulated supplies for my audio power amplifiers, I use a conventional unregulated transformer/rectifier/capacitor power supply preceded by a variac for testing.

[H713]

I should note that while the primary use for this will be audio power amplifiers, it will see use for testing other more sensitive circuits. As an example, class A mixing console circuits operating on +45V rails. It's still probably overkill, but I think it will be an interesting challenge to design.

[Kleinstein]

Filtering the extra frequencies from a scr regulator is not so easy, as the frequencies are relatively low. It is possible but still not so simple.
Large filter caps can have the disadvantage of getting a even lower power factor, especially if there is no series inductor to improve the power factor.

A switched mode converter has the advantage that passive filtering is easier with the higher frequency noise. So the choice is not that clear.

For the current rating one could have peak current limits higher than the long time continuous rating. Especially transformers are relatively slow to react and can deliver more than nominal power for quite some time. This is especially sufficient for the audio output peak currents.

[H713]

Alright, here's where I'm currently at:
I believe I have a way to do transformer tap switching without having to design and order a custom transformer. The Antek AN-3236 has a pair of windings on both the primary and the secondary. Assuming that I use one transformer for each power supply, that gives me three steps, which should be workable.

Yes, this PS will probably be digitally controlled using a microcontroller. In addition to the tap switching, it will also manage the speed of the cooling fans. This brings me to a question regarding microcontrollers:
If I want to digitally control both transformer tap switching and speed of the cooling fans (dependent upon heatsink temperature), would this require two separate microcontrollers or could one handle both processes?

[Zero999]

My advice would be to not use a microcontroller to control the tap switching. Use comparator(s) on the output of the power supply to control the tap switching circuit. That way it will work irrespective of the microcontroller crashing and the same goes for the fans.

[Neomys Sapiens]

I would refrain from digital control other than providing the set values via D/A converters.

I think that those PS that use multiple MosFets in a linear mode must have had the benefit of a builder who could ask everything from a supplier, like very very tight selections. The only appliance which I have personal experience with was a custom unit which was used to weld hermetic cases for hybrids. If one of the Mosfets did break ranks even a bit, mass extinction ensued. While it is certainly feasible, you would have to do the selection yourself, which could be costly. If you want to try anyway, APT, IH (now Renesas) and IR had appnotes on the topic.

Where I see problems is the 2HE requirement. Even assuming that you use fans with tunnel-type heatsinks, the heatsink has to have some capacity for heat which means mass, and therefore size.
I would also recommend to give that brute a line switch with holding coil/undervoltage trigger and wiring the failure alarm contacts from the fans into this circuit alung with the thermal switches in the transformer and those at the heatsink.

Concerning the preregulator, careful consideration should go into the control law. It is quite easy to produce instability here. Also, there are some alternatives to switching windings by relay or using phase control. Switching by triacs would be instantaneous and could be fast enough to make it work as a cycle-to-cycle switch. I have to concede that varying phase angle between the line voltage and current in the secondary could make this tricky. 
Also, there is the magnetic amplifier (Transduktor), which introduces some noise by saturation, but certainly less than a phase control.

As you are doing audio power amplifiers, you might be already familiar with the McPherson circuit, where dissipation is limited by split rails. An example (using single transistors for a rather small and compact PS) was once published by Ian Hickman in the Electronis&Wireless World. My research led me to the original publication by McPherson, which was aimed at AF amplifiers. I unfortunately do not have it myself (IEEE transactions),

[splin]

If it is one off (or only a few required) it should be fairly easy to add additional windings to a standard toroidal transformers (provided the centre isn't potted!).

For example, I have a 2x50V Vigortronix 625VA, VTX-146-625-250, which requires only 1.68 turns/volt. So an extra 12V secondary would only need 20 turns. The centre has a diameter of around 38mm so plenty of room to accomodate extra windings. Starting with a 2 x 15V transformer you would get the advantage of lower copper losses in the secondary at lower voltages, whilst adding the extra windings with progressively finer wire to make it a bit easier to wind.

Alternatively, you could carefully cut the existing secondary to add your own taps. In the case of my 625VA, the existing secondary is 2 x 84 turns of 1.25mm dia wire in one layer. This wouldn't be feasible for much smaller transformers of course with secondary windings having many more turns of finer wire.

A further example; my 225VA 2 x 50V Vigortronics VTX-146-225-255 has 2 x 150 turns of .82mm so 3 turns/volt (with 30mm inner diameter).

Or remove the existing secondaries and rewind. Some transformers have the primary winding on top of the secondary(s) but that isn't the case for my Vigortronix. Jerry Walker has some really well presented videos on transformers and has done lots of rewinds. This one is good where he designs a linear supply to power an RD6000 switching supply and rewinds a pair of toroids:




[EDIT] He actually rewinds the transformers in part 4. Search for 'RD6006 Linear Part4' (without the quotes).

[duak]

The hp 6002A is a supply of similar rating that uses thyristors to switch the transformer secondary windings.  The general design is written up in this hp Journal issue: https://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1977-06.pdf  The authors go into the algorithm of how the windings are selected.  Some older hp supplies have two voltage ranges that the user can manually select, kind of like a transmission in a vehicle.  Automatic range switching is nice, but can sometimes be fooled close to the transition point(s).

I've come to appreciate the design of hp and Lambda lab power supplies.  Most use a floating regulator where everything is referenced to the positive output terminal.  This makes the voltage and current feedback loops much easier to stabilize as the error amplifier does not need to swing over the full range of the raw DC input.  There are a lot of bells and whistles in the 6002A that aren't needed here.

Foldback Current limiting can protect the pass transistors by reducing the current limit as the voltage across them increases.  Not much could be worse than to have a pass transistor in the supply fail shorted while powering a test circuit.  You could easily have two corpses afterwards.

[H713]

I would refrain from digital control other than providing the set values via D/A converters.

I think that those PS that use multiple MosFets in a linear mode must have had the benefit of a builder who could ask everything from a supplier, like very very tight selections. The only appliance which I have personal experience with was a custom unit which was used to weld hermetic cases for hybrids. If one of the Mosfets did break ranks even a bit, mass extinction ensued. While it is certainly feasible, you would have to do the selection yourself, which could be costly. If you want to try anyway, APT, IH (now Renesas) and IR had appnotes on the topic.

Where I see problems is the 2HE requirement. Even assuming that you use fans with tunnel-type heatsinks, the heatsink has to have some capacity for heat which means mass, and therefore size.

I'm hoping to mitigate the issues with the FETs in a few ways. For one, I'll be using 900V rated parts with a fairly high R_dson. Additionally, I will be using quite a few of them in parallel to have as much overhead as possible. The idea is that there will be enough beef in the series pass transistors that they will survive even if the tap-switching fails. They will also have some rather generous source resistors in order to help with the current sharing.

Heatsinks are not a problem either- I currently have an Ashly FTX-1000 on my bench that is beyond repair (No seriously... it's cooked. https://www.diyaudio.com/forums/solid-state/352468-ashly-ftx-1000-failure-analysis.html#post6154245). The heatsinks and chassis may end up being used for this project, unless I come up with something better.

Also, I have plans of running the heatsinks live and not trying to isolate the MOSFETs from the heatsink, so the thermal interface should be quite good.

[magic]

Maybe you really need a fancy high-spec 24V supply and another more relaxed one for 70V?

As for MOSFETs, whole threads are written on DIYAudio on how to use them for amplifier output stages. It seems the most "interesting" problem is their very high fT and resulting proclivity to crazy fast oscillations.

I have actually had a surprising encounter with this problem when a simple ON/OFF rail switching FET went into tens-of-MHz oscillation during transition and coupled noise into the control circuitry which promptly turned it off. The rail never came up even though everything seemed right

[H713]

When I was testing the FETs for this project, I initially had about 1 meter of 20 AWG wire connecting it to the filter caps, leading it it to oscillate so violently that it was picked up on an oscilloscope almost 10 feet away. 1 single MOSFET, 1 meter of wire = violent oscillator. Silly me had thought "It's just one FET bolted to a heatsink. I don't need no stinkin' bypass cap!"

Yes, a separate power supply for 24V is ideal (and I do have one), however, I would like to design this to perform well at 24V.

[magic]

One Bob Cordell built quite a few FET power amps and wrote a bunch about it over the years on DIYA and in his articles and books. He recommended tight layout, base stoppers and local Miller compensation of the outputs IIRC.

Also on short circuits and overload protection, there is a big MOSFET vs BJT thread on DIYA where people tested some jellybean power FETs for pulsed SOA and found them to withstand quite a lot.

The question of SOA of switching power MOSFETs recurs on this forum every now and then but the discussion usually revolves around quoting app notes because it seems few people here are crazy enough to actually have tried to push the limits of jellybean parts