Advertisement

Blog

Build a high-voltage linear regulator

Editor’s note: Our guest bloggers are:

Donald Schelle is an analog field applications engineer for Texas Instruments and has more than a decade of engineering experience. He received a bachelor’s degree in electrical engineering from Lakehead University, Thunder Bay, Ontario, Canada. Donald can be reached at ti_donschelle@list.ti.com.

Jared Becker is an applications engineer for Texas Instruments. He received his bachelor’s degree in electrical engineering from Arizona State University. Jared can be reached at ti_jaredbecker@list.ti.com.

Introduction

Regulating from a very high voltage (>200V) to something usable by modern electronics (like 3.3 V) is a difficult proposition. For applications that require more than 1-2 mA of current, designers eschew a simple resistor/Zener diode combination in favor of a more complex solution. Managing power dissipation while maintaining a low-cost circuit can be challenging as well. The circuit in Figure 1 is a simple low-cost regulator that achieves high step-down ratios at output currents of 5 mA while minimizing cost; less than $0.50 in low volume.

Figure 1

Build a simple high-voltage linear regulator using a handful of low-cost components.

Build a simple high-voltage linear regulator using a handful of low-cost components.

In this article, we present detailed design calculations and tradeoffs needed to meet wide-input voltage-range requirements. We chose the following design parameters for our example regulator:

  • VIN(MAX) = 250V
  • VOUT = 3.3V
  • IOUT = 5 mA
  • TA(MAX) = 60o C (maximum ambient temperature)

Managing power dissipation

Managing design thermals is central to a functional first-pass prototype. Maximizing the input-voltage range requires careful component selection.

The pass transistor’s power dissipation will limit the output current. Equation 1 calculates the maximum current as a function of pass transistor power dissipation:

When the transistor is connected to a piece of 1-ounce copper (which acts as a heat sink) with dimensions of 15 x 15 mm, the chosen transistor (FCX458) is specified with ΘJA = 125o C/W. Given these parameters, the transistor will reach the maximum junction temperature of 150o C when the output current is 2.92 mA. Increasing the copper attached to the transistor to 50 x 50 mm reduces ΘJA to 60o C/W and yields a maximum output current of 6.08 mA. This is well within the design specification of 5 mA.

Power dissipation in the shunt resistor, RSH , also will be at a maximum when the input voltage is at its peak. Equation 2 calculates a minimum RSH resistance value based on the maximum resistor power dissipation:

Limiting power dissipated in the resistor to 350 mW yields a shunt resistance of 174 k Ω , which is a standard Energy Information Administration (EIA) 1 percent resistor value. The majority of power dissipated in the circuit occurs in either the shunt resistor, RSH , or the pass transistor, QP . Table 1 summarizes the power dissipation at the minimum and maximum design operating points.

Table 1

Power dissipation shifts between the shunt resistor and pass transistor depending on the operating point of the circuit.

Power dissipation shifts between the shunt resistor and pass transistor depending on the operating point of the circuit.

In general, power losses are lowest when the input voltage is at the minimum. Power losses in the resistor are highest when the input voltage is at the maximum. Power losses in the transistor are highest when the output current and input voltage are at their maximums.

Optimizing for minimum input voltage

You can optimize the design for the widest input supply range by choosing a transistor with a high beta, β , and a shunt regulator with a low minimum cathode current, IK(MIN) . A few popular choices for shunt regulators include the TL431 and TLV431. These devices feature IK(MIN) values of 1 mA and 80 µA, respectively. An improved ATL431 reduces the minimum cathode current to 40 µA. Equation 3 calculates the minimum input voltage as a function of the minimum cathode current:

Decreasing IK(MIN) by choosing an appropriate TL431 device and increasing the β of the transistor play pivotal roles in maximizing the input voltage range. Figure 2 plots the minimum input voltage as a function of β . Decreasing the minimum regulation current of the TLV431 device by 50 percent relaxes pass transistor β requirements equally so. Alternatively, you can maintain beta and gain additional input-voltage range. If suitable components are not available to meet the target input-voltage range, increasing the power dissipated in the shunt resistor will reduce the minimum input voltage at the expense of more heat dissipated in the design.

Figure 2

Minimum input voltage varies dramatically depending on the minimum shunt regulator operating current. The improved ATL431 reduces transistor beta requirements by almost 50 percent.

Minimum input voltage varies dramatically depending on the minimum shunt regulator operating current. The improved ATL431 reduces transistor beta requirements by almost 50 percent.

Testing the circuit

When testing circuits that implement high voltages, it’s important to pay careful attention to appropriate safety precautions. High voltages can be dangerous and you need to treat them with care, respect and intelligence. If you are unfamiliar with appropriate safety precautions, seek informed help before turning on the power.

Testing the circuit yielded good results. An output capacitor is required in order to keep the regulated voltage from oscillating. The circuit in Figure 1 shows 10 µF, but we were able to achieve stability with a capacitance as low as 4.7 µF. No-load to full-load transient responses were also within tolerance. The scope capture in Figure 3 (trace No. 2) shows approximately 200 mV of ripple during the transient event. Also key is that the circuit is designed to be functional across semiconductor process limits for the ATL431. Since the typical IK(MIN) is much lower than the maximum, the measured minimum input voltage was much lower than calculated.

Figure 3

With a 200V input (yellow trace), the output voltage (blue trace) maintains regulation during a full-scale 0-5 mA rising load transient commencing after the trigger output (purple trace).

With a 200V input (yellow trace), the output voltage (blue trace) maintains regulation during a full-scale 0-5 mA rising load transient commencing after the trigger output (purple trace).

Automation

We used Excel to automate the design calculations presented here. You can download the free calculator tool here. Note that the tool requires additional add-ins to select standard resistor values. [1]

Reference

1 Donald Schelle. Calculate Standard Resistor Values in Excel, EDN, January 2013.

13 comments on “Build a high-voltage linear regulator

  1. D Feucht
    January 12, 2016

    I really like minimalist circuits like this. And you did not omit thermal analysis! Good work.

    I did not fail to catch your comment in passing about needing an output capacitor to stabilize the feedback loop. Might you consider writing a sequel that does a feedback analysis of the loop stability?This would move the analysis to a deeper level.

    I am taunting you a little with this question because with a TL431 in the loop, this is a seemingly simple circuit! I have a 6-part article in the queue for Planet Analog that goes deeply into the TL431, for later this year. However, do consider writing such an article. This one is a good start! I would be glad to discuss more of this with you in background mode. Get ahold of me through http://www.innovatia.com if interested.

  2. Hooey0
    January 13, 2016

    I agree that a minimalist approach is generally a desirable attribute for a design.

    However, I do have a comment on the circuit as shown.

    The ATL431 has a set of ranges of output capacitance versus current necessary to ensure stability. These ranges are clearly shown on the datasheet. Since there isn't a capacitor across its output, I doubt that this design would satisfy all of the stability requirements.

    A number of years ago, I was presented with a circuit that used a similar shunt regulator, the LM431, also without a capacitor across the output. I cold-called the late and great Bob Pease when he was at National Semiconductor and discussed the circuit with him. In his own inimitable way, he said that the LM431 was a “piece of crap”, and very prone to instability. He recommended the LM4041, which had no output capacitor requirements for stability. We subsequently redesigned the circuit and used the LM4041 with no problems.

    The bottom line is read the datasheet and make certain that you adhere to all of the requirements of the device. If you don't, you're looking for trouble. 

     

  3. CC VanDorne
    January 13, 2016

    I also appreciate the minimalistic approach.  In fact, that's pretty much what I thought EE's were trained to do: meet the spec while keeping parts count down and thus BOM cost low.

    But my requirements are different.  What changes in this design if I want a high voltage output, and with more current?  In other words regulating say 500Vdc down to 450V, or 250Vdc? And how would one make that variable on the fly?  I found some designs on line but they are not so minimalistic.  In this case it looks like putting a pot where there are now two feedback resistors would suffice, no?  Please advise.

    Oh, and if you are asking yourself what fool would want a 250V output!?  Only those crazies who still mess with tubes would want that!  Yep, guilty.

  4. Scott Elder
    January 13, 2016

    The circuit might be running afoul of the pulse power limits.  I didn't see mention of the input rise time, but if it is on the order of ~300us @60C Ta, this is pushing the limits.

    Other than that, it looks great.

  5. AlecS
    January 19, 2016

    I looked up the FCX458 datasheet and it's a SOT89 package.  The typical recommended pad layout would leave a little less than 10 mils of clearance between the collector and emitter.  IPC2221A recommends at least 24 mils of clearance even at voltages just above 30V.  While there is apt to be great deal of difference between IPC221A and the actual breakdown voltage, I am surprised you were able to test this at a 200V input without any arcing (and presumably downstream destruction) occurring.  Is the FCX458 available in a different package that affords more creepage/clearance?

     

    I'd be paranoid about designing something like this in to a PCB unless I had well known environmental conditions and could manage to use something like Paschen's law to convince myself I was several multiples away from the physics allowable minimum.  Aren't there safety issues here without having some sort of galvanic isolation?  Shouldn't safety trump the simplicity/cost factor?

  6. Don Schelle
    January 25, 2016

    Hi AlecS,

    Thank you for the well thought out reply. I found yours especially concerning, considering that safety is something we take very seriously.

    While we tested our circuit above 200V (I believe our power supply stopped at 220V), we did not see any arcing. Our test board was also haphazardly constructed, which initially added further mystery around this subject.

    I did lookup Table 6-1 (Electrical Conductor Spacing) in IPC-2221A and it does indeed say in column B2 that for elevations less than 10000 feet (we tested this board in Dallas, which is well below that altitude), you do need to maintain at least 0.6mm spacing when operating below 150V and 1.25mm when below 250V. Note that column B2 applied to uncoated bare conductors.

    Reading further I believe that the spec which should be applied here is B4/A6, which according to IPC-2221A is “most commonly used in commercial, nonharsh environment applications in order to obtain the benefit of high conductor density protected with permanent polymer coating (also solder resist), or where the accessibility to components for rework and repair is not required.”

    Column B4, section 6.3.4 of IPC-2221A states that (at any elevation) it is intended “When the final assembled board will not be conformally coated, a permanent polymer coating over the conductors on the bare board will allow for conductor spacing less than that of the uncoated boards defined by category B2 and B3”. The minimum spacing in this case is 0.4mm at voltages less than 250V.

    Column A6, section 6.3.6 of IPC-2221A states that (for elevations less than 10,007 feet), “External component leads and terminations, that are not conformal coated, require electrical clearances states in this category. The minimum spacing in this case is 0.8mm at voltages less than 250V.

    Looking at the SOD-89 footprint for the FCX458 (URL excluded due to forum rules), I see that base, emitter, and collector of the device are all on separate pins on the same physical face, thus it is conceivable that at some point during the operation of the circuit that 250V will be across two of these terminals. The pin-spacing on these terminals is 1.5mm on centre. The width of each pin is 0.54mm (max) or 0.62mm (max) depending on which pin you look at. Let's assume that worst case all pins are 0.62mm (max) to simulate the worst case clearance and bake in a little conservative judgement. 1.5mm – 0.62mm = 0.88mm. This clearance is well within the common combination B4/A6 spec as laid out in IPC-2221A.

    For the more cautious readers, I'd refer to column A7, section 6.3.7 of IPC-2221A which states (at any elevation) that “External Component Lead/Terminations” have a “Conformal Coating”. In this case the minimum clearance requirement drops to 0.4mm. You could argue that conformally coating a board adds cost, though you could also argue that many industrial designs already conformally coat their boards and thus the cost is already baked in. The point here is that conformally coating a board does add a significant amount of margin when looking at clearances in high-voltage applications.

    In either of the cases above the circuit should function as intended and meet safety specs as outlined in IPC-2221A. I'd like to thank you for writing in. I wasn't aware of these safety specs, and you have taught both myself and the readers about an additional design facet when working with high voltages.

    Don

    P.S. To the other forum visitors, I do intend to reply to all posts; however, my day job and family life is a little hectic right now and I have limited time to post. I'll get there. Please be patient.

  7. Don Schelle
    January 25, 2016

    Hi Scott,

     

    Thank you for writing in.    Always good to hear from our readers.

     

    I'm a little confused on what you mean by pulse power limits.    Are you referring to the load current transient rise time?    If so, would that mean a 0-5mA load transient that occurs in 300us?

     

    While we applied a load transient to the output, our test equipment was rather crude and we didn't capture the load current rise-time.    We also didn't put the circuit in an oven to achieve 60oC.

     

    If my assumptions above are correct, would you please explain how you came to the conclusion that a 0-5mA transient with a rise time of 300us @ 60oC would be “pushing the limits”?   We'd be very interested in any shared learnings that you may have.

     

    Don

  8. Don Schelle
    January 25, 2016

    Hi CC VanDorne,

     

    Thank you for writing in.   Admittedly, learning about tube electronics has been one of those 'bucket list' items that I've never gotten around to yet.   One day….    🙂

     

    Using a potentiometer in the feedback network is an acceptable solution to 'tune' the output voltage to a desired level.   If you want to get fancy, you could use a digital potentiometer to achieve the same function; though finding a digital potentiometer with an input voltage range of 250V might be like searching for a unicorn.   In lower voltage designs, I've successfully implemented this approach.

     

    Another way to program the output voltage via digital means would be to use a third resistor in the feedback network.    The third resistor tied to a DAC allows more precise control and the output voltage function is linear and predictable.    The third resistor will limit current flowing into the DAC and it 'should' work at those high voltages.   I've written an article on this subject called “An Easy Way to Roll Your Own Programmable Power Supply”.   I'd include a url here, but forum rules prevent it.    Google will have to be your friend.   🙂    Also, if you're going to design something like this, my usual caveat applies.    There is no substitute for a prototype.   

     

    Thanks again for writing in.   I never thought this little article would generate the kind of foot-traffic that it has.   Until next time.

     

    Don

     

     

     

  9. Don Schelle
    January 25, 2016

    Hi D Feucht,

     

    Thank you for writing in.   Loop stability is certainly a concern here, and you can spend hours optimizing the circuit to achieve ideal transient response while mainting some semblance of stability.  

     

    Understand that we write these articles on our own personal time, and while we enjoy doing it, there is a large amount of time that we can't dedicate to other things.    It's hard to say “no” to a little boy asking you to go outside and play soccer for a few minutes.   🙂

     

    When we built and tested the circuit, we initially included no output capacitance.   Unsurpisingly the output voltage oscillated.    Rather than go back to the drawing board for endless hours of calculations, we slapped the 'easy button' and grabbed a 10uF electrolytic capacitor off the shelf.    Problem solved!    We then reduced the capacitance until the output started oscillating again.    Below 4.7uF we ran into problems, hence the comment at the end of the article.

     

    I'm looking forward to your 6-part article regarding the TL431.   It sounds exciting!    What I find most fascinating is that among what I call 'old-school' engineers, the TL431 is very common and well known.  But many fresh-out-of-school engineers have never heard of it.    Admittedly, I didn't learn about this part until 10 years into my career.   While I described one application here, the part is surprisingly versatile.

     

    Thanks again for writing in.  

     

    Don

     

     

  10. AlecS
    January 26, 2016

    Thank you very much.  That is a fantastic and very thorough response to my question.

     

    I'm not overly knowledgeable about IPC-2221A or high-voltage design in general.  I essentially had to deal with both exactly once in the past when working on a high-reliability industrial design and sort of just had to do the best I could for a first timer.  We tested a lot of bare, uncoated, milled breadboard mockups at near 300VAC as a proof of concept that we would have more than sufficient spacing once everything was laid out on an actual production board.

     

    As a note for others to avoid potentially violent arcing problems I had during initial testing, beware of solder flux contamination left behind when you are hand working/reworking high-voltage boards like this.  At first I was baffled that the seemingly more than adequate (according to IPC-2221A) spacing used was insufficient and left me with arcing issues.  After wasting the better part of a day pondering, someone with better eyes and more experience than me noticed the slight yellowish discoloration left behind by the solder flux (despite the blackened and charred board surface).  I'm glad he noticed and saved me from running off to mill another board with even more ridiculous trace/component spacing.  After applying a little board cleaner and scrubbing, it retested fine without issues.

  11. Don Schelle
    January 26, 2016

    Hey AlecS,

     

    Appreciate the response regarding solder flux residue as I share your frustrations.   Early in my career, I spent a number of years developing evaluation boards for a major silicon manufacturer.  We too would often mill prototype boards and hand assemble; sometimes (almost always) using copius amounts of solder flux.   

     

    When problems inevitably arise (at any working voltage) I quickly learned that my first answer would be a question.   “Did you thoroughly clean the board?”.   For new engineers, the answser was often 'no'.   Our observations of these failed boards always lead back to flux residue being conductive at some level.    Sometimes those levels would be minor leakage and it may go unnoticed, or not (think precision analog circuits).   Other times the residue would be more conductive and cause mysterious circuit operation.  Without data, our explanations for these problems were limited to theories; however, the most prevailing theory was that naturally occuring moisture in the air would be trapped in the flux residue as it dried.        

     

    After being burned (metaphorically) a number of times, I became very OCD about cleanliness.   My mother would be so proud.   After every rework and especially the initial prototype assembly, I would clean every PCB using the following method:

     

    1. Throw the PCB into a heated ultrasonic bath which contains a mixture of flux-off.   5-10 minutes usually did the trick.

    2. Use a toothbrush to clean every component.   When handing the toothbrush to the uninitiated I found myself using the phrase “go to town on the board with this brush”.  🙂

    3. Wash the board with cold water.

    4. Bake the board to ensure that no water or moisture remains.    Secretly, myself and others would grab the air-hose to blow excess water off the board.   There can be potential problems with using compressed air on electronics, but we won't go into that here.

     

    Once I adopted the above protocol, the mysterious/weirdo events with my boards dropped dramatically.   Empircally I would put the number at about 80% and the problems that remained, were explainable using scientific means.   Long story short, my circuits began to work in the real world like you would expect them to work on paper.   Happy days!

     

    Thanks again for writing in!

     

    Don

     

     

  12. Scott Elder
    January 30, 2016

    Hello Don,

    I was referring to when the input voltage rises to 250V in about ~300us.  This causes a current slew through the transistor hence transient power across the device.  The output is essentially a dead short at turn on.

    The datasheet shows transient pulse power limits down to about 100us.  It seems that the design is on the edge (if not over) of those limits.  Especially if one considers high beta, high ambient temperature.

    Scott

  13. EdwardThirlwall
    September 17, 2018

    I am always in awe as to how much knowledge you possess in this field. I guess you would need to have keen interest in this industry to be able to draw on so much knowledge about it. It would not be an easy task to take on if you simply do not have any interest in this topic to be able to move forward to learn a thing or two about it.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.