DIY Programmable Power-Supply for Vacuum Tubes – Part 1

Preface: up until now, all posts I’ve shared were completed in a single post. This was due to the fact I’ve waited until I was done with it and only then posted. This allowed me to assemble/verify (when needed), and was much more comprehensive for readers. However, lately I’m finding it more difficult to find the time to cross items off my “diy to-do” list. Quite a few items get stuck for long periods of time in the design stage, due to lack of time to move it forward and complete the board layout/assembly/testing. Therefore, I’ve decided to gradually post a few of these on the blog as parts of a project. This post will be the first of a few such projects that will be split into several parts. Hopefully, even sharing partial information such as schematics will prove useful to some readers. </end preface>

One of the items that was on my “wish-list” for quite some time is a programmable power-supply (PS) that will be fit for work with vacuum tubes. The main reason I need it is because I’m missing a high-voltage PS that can reach as high as 400V or over. Therefore, this was the main objective of the design I will present in this post. However, seeing as most transformers that are intended for these uses include a low voltage secondary winding for the heaters, it makes sense to have another channel that can supply the heater rail too.

The first step was to set the requirements out of the design. Since this is going to be used for development and not to power high wattage tube amplifiers at full load, I don’t see a need for excessive current capability on the high voltage rail. The rail voltage should be adjustable down to 50V, and up to 400V or slightly higher (450V at most). I want it to have an adjustable current limit just to be on the safe side, it can always be set to a sufficiently high value to be limited by other parameters. Finally, on the heater supply common values are 6.3V/12.6V, so these are the two modes which will be supported. I did want to use DC heater supply to reduce hum, and to be able to control it easily.

One of the main decisions when designing a high voltage PS has to do with the topology and type of high-voltage (HV) devices to use. Some people use vacuum tubes, others (like me), prefer to stick to MOSFET’s/BJT’s. However,when it comes to HV devices, there’s more to choosing the right transistor than simply looking at the maximum ratings for power and voltage. Most transistors are meant to be used as switching devices, and can therefore either pass significant current, or withstand high voltage. Once you start using these devices as something other than a switch, other limitations must be taken into account. Therefore, you must be sure to use transistors with forward-bias-safe-operating-area (FBSOA) ratings. It might be worth making sure these secondary-breakdown limitations are taken into account in the datasheet, as they will limit performance before power dissipation will for these applications. As you’d expect, devices with appropriate rating are typically more expensive than general-purpose devices. This is true for most devices in such a PS, including capacitors, trimmers, and any other parts where an appropriately rated components must be used for safety and reliability.

While there are other designs around the web, and even commercial products that could be purchased, I’ve decided to go for the DIY route and design one for my own use. The schematic I came up with can be seen below (click for high res):

As always when it comes to high voltages, and I can’t stress this enough, the voltages present in this schematic are high. Working with such high voltages can result in injury, and potentially even death. Again, these are potentially LETHAL voltages, if you aren’t qualified and comfortable to deal with these voltages, do not attempt to build this circuit. If you choose to do so, you are doing this at your own risk and responsibility.

Fig. 1. Power Supply Schematic for Both Rails – high voltages present, use at your own risk

The power transformer should have an appropriate secondary voltage, nominally ~325V as drawn. This will suffice for a 400VDC output voltage. If you’d like to push it up to 450V output, a higher voltage secondary should be used such as ~350V.
Voltage rating of parts are the minimum recommended values, and can obviously (preferably) be increased for larger margins. For instance, 630V (or higher) capacitors can be used for all 600V/500V capacitors drawn in the high voltage portion.

Operation Principle of High Voltage  Rail:

Lets start by going over the high voltage B+ rail. This is different than most designs I’ve seen, but its actually very simple to understand. U2 is a LT3081 voltage regulator, which uses a current reference instead of a voltage reference as most IC regulator. Therefore its output voltage  can be set using a single resistor, and the LDO then buffers this reference voltage to the output. In this case, R6 sets the minimum value of this reference voltage to 50V (50uA reference current is used inside the LT3081) when both trimmers are at their minimum position. This is the heart of the regulator, with all of the other parts providing over voltage protection, and adjustment of voltage/current.

Lets start with Q1/Q2, which are MOSFET’s with appropriate rating and power dissipation capability for this application.  These are used to limit the voltage and power over the regulator IC. Q2 is biased using D4 (12V zener) to leave sufficient voltage for the regulator to work properly, but keep it well within its comfort zone. Q1 is then biased using a resistive divider of R1/R2 at halfway between Q2 and the input voltage, to evenly spread the voltage and power between them. D2 (12V zener) is used as a protection device to limit the VGS value of Q1. Q2 doesn’t need it as it already has such a path via U2 and D4.

The input filter only uses capacitors (C1/C2), there will be non-negligible ripple over it. Therefore, C3 is used to low-pass-filter (LPF) the voltage at the gate of Q1, which then serves as a low impedance buffer for this filtered voltage and reduces ripple significantly.  The gate of Q2 doesn’t have a similar capacitor to prevent a voltage stress over U2 in case of a sudden change of the output voltage. However, the corner frequency of this single LPF if low enough (<1Hz) to filter input ripple effectively.

The next part is U2 and it surrounding components. U1 is a constant current source with the resistors around it used to set its current (~750uA in this circuit). Since U2 generated a proportional-to-absolute-temperature (PTAT) current, a complementary-to-absolute-temperature (CTAT) source is needed. This is achieved by Q3 which is a small-signal BJT in a diode connection. As temperature rises its forward voltage drop falls, and the current R5 will drop. Overall, this circuit generated a mostly temperature independent, or zero-TAT (ZTAT) current. The reason we need this ~750uA current is to amplify the effectiveness of RV1/RV2 voltage setting resistors. If this current wasn’t present, to achieve reference voltage of 400V for the LT3081 would need a 8Mohm potentiometer (50uA reference current of the IC). The closest standard value is 10Mohm, and the selection of these potentiometers is very limited. By adding the 750uA current to the 50uA of the LT3081 we can use a X16 lower value of potentiometer, which results in 500Kohm which is significantly more common. If you can source an appropriately spec’d 10Mohm potentiometer, you can always use it and omit U1 and its supporting components. R6 is placed before the connection point of the additional current which allows a high value (1Mohm) resistor to be used which makes the LPF of C5/C6 more effective at reducing reference noise from U1. C5/C6 serve the additional function of adding a “soft-start” to the regulator by heavily filtering the reference voltage. D3 is there to offer reverse polarity protection and offer a discharge path for the reference node in case of sudden change of surrounding voltages.

The LT3081 in the package used for this design offers a few additional pins with extra functionality. These include output current monitoring, and temperature monitoring. For this design there is no need for them and they are therefore not used. However, one feature of the T3081 which is employed in this design is the integrated current limiting which is set by a single resistor. To allow the output current to be adjusted all the way down to 0A across all manufacturing spread of the LT3081, a smaller than nominal 330R is used as the minimum value, with an additional 1K potentiometer. The need to support the spread of LT3081 means that part of the range of RV3 will be cropped for the nominal case. This also means that for the nominal case, maximum value of the current limit will actually be >300mA. This doesn’t mean that the PS is designed to support currents of over 150mA as noted in the schematic sheet. However, given sufficient cooling, it could indeed do this, especially if the output voltage is high enough so that the power dissipated across Q1/Q2 is kept low. Please note that RV3 (just like RV2) is operated at a significant potential above ground, and must therefore be an appropriately rated part for safe usage. Notes on the schematic include an appropriate part for RV2, and a similar part (of lower resistance) can be used for RV3.

C7/C8 are used for output filtering. Excessive capacitance isn’t needed here as the output is regulated. However, you can use higher capacitance value if you choose to. I would recommend against using much lower capacitance, or else the stability of the regulator could be compromised. Finally is the circuit around Q4. Other than including the LED for visual notification of the PS  being active, it serves as the minimum load path for the regulator. To keep the LT3081 happy, a minimum load of 2mA nominal (5mA max) is needed. This is achieved by Q4 and its supporting components. R8/R9 pass a small current (~2mA) that is used to bias D5 (zener diode) and supply the base current of Q4. The voltage of D5 (-VBE of Q4) over R10 generates a ~9mA current that is needed for the minimum load of the regulator (and a headroom for D3 operation). Q4 could have obviously been a MOSFET of the same type of Q1/Q2, and is only a different component to reduce cost due to the reduced current needed for this device.

Operation Principle of Low Voltage  Rail:

Now lets move to the low voltage heater supply. I wanted to create a modular design that could be scaled as need for up to 8A@6.3V and 4A@12.6V (and in theory even higher by a simple parts replacement) in case multiple heaters will be run from this rail. However, I didn’t much want to design a discrete voltage LDO so that stability consideration won’t need to come into play when designing the feedback loop and compensation network that LDO’s typically include. Luckily the MIC29xxx family from Microchip offer very low dropout voltage at various current capabilities. The strongest device in the family offers a 7.5A current handling capacity, but unfortunately its fairly expensive. Therefore I’ve decided that a modular approach with a few lesser capable parts would be well suited for this application. In this case I’ve opted for 4x3A LDO’s which are much cheaper, and even when you consider the cost of additional parts, sums up to roughly the same cost as a single 7.5A LDO. The modular approach has its benefits though, such as easier thermal design and the ability to save cost and build a lower current version if you choose to.

Looking at the schematic, in-front of the bridge rectifiers are 2 switches (actually two poles of the same 4 pole switch) which are used to select the connection of the secondary windings from the power transformer for 6.3V/12.6V voltage. Following rectification this is filtered on the bulk capacitors, a capacitor per regulator (4 in this design). the 22,000uF per capacitor is probably more than you’ll even need, and can safely be reduced to 15,000uF or perhaps even lower. Next is the heart of the regulator built around the MIC29303 LDO’s. The LDO’s are all set (independently) to ~2.75V output voltage. To generated the 6.3V output voltage, their ground node is not connected to ground, but instead to a programmable voltage set by U4 as we’ll explain in a moment.
The first question that comes to mind is why would you want to do this? There are a few reasons for this actually. Most important is the fact that multiple LDO’s are used, with their output currents summed together at the load. Due to mismatch between LDO’s their outputs can’t be shorted together directly, and require ballasting resistors (R17/R20/R23/R26). The higher the mismatch, the larger the ballasting resistors must be, and higher the output impedance of the regulator board. By using a small multiplication factor of the LDO’s internal reference (lower output voltage), the difference in output voltages can minimized, and the ballasting resistors can be made smaller. The resistors in the feedback network are fairly low in value, and are used as the minimum load path for the regulators to maintain regulation.
If you prefer to have a lower current capability at reduced cost, you can always remove 1/2/3 LDO’s and their supporting circuity, according to the max load current. Since the LDO’s have some mismatch, I would recommend to use 2.5A per LDO at most so that a 2 LDO’s build would support up to 5A at most. This will keep each LDO at ~3A limit even in the presence of mismatch/variations.

Now lets move to U4 and its surrounding circuitry. TL431 is a shunt regulator with a 2.5V reference voltage. The resistive divider to the left of it is used to set the desired voltage. RV4 is a potentiometer used to allow a +/-0.5V tuning of the output voltage around the nominal 6.3V/12.6V values. R11 is a switchable resistor (another pole of the single 4PDT switch) which can be shorted for 6.3V, or placed as part of the feedback loop for 12.6V. C18/C25 act as a bypass capacitor to reduce noise and allow a return path for the feedback network of the LDO’s at higher frequencies for stability. C26 acts as a bypass for SW3 in case it is placed far from the board and has significant inductance in the wires running to it. Q5 is activated as the current through U4 rises over ~4mA by R14, and conducts the bulk of the current running down this path. The need for Q5 arises from the fact that the LDO’s used can have currents of over 30mA per device flowing out of their ground pin at high output load currents. Therefore U4 is unable to dissipate this heat, and Q5 is needed.

Finally, at the right most edge of the circuit are the output filtering capacitors, and LED for visual notification. When the PS is active, either D9 or D10 will light up according the the set voltage (final pole o f the 4PDT switch). D8 is a fault notification LED. It can be activated by either on of the LDO’s in case of a fault. The MIC29xx3 FLG pin support multiple errors notification, including short circuit protection/thermal shut-down protection/voltage dropout/etc. You can choose to omit it if you don’t want to use this functionality.

(Optional) Reducing Voltage Stress Over RV2:

One of the things that were somewhat annoying to me in the topology above is the working voltage requirement of RV2 which should be able to handle ~400V. While appropriate parts could be bought, they tend to be fairly expensive. Therefore, I wanted to add an option to reduce the working voltage of this component while keeping a similar circuit topology. One possible solution is shown below in Fig. 2 (click for high res).

Fig. 2. Modified Schematic with Reduced Voltage Over RV2 – high voltages present, use at your own risk

The circuit above adds a few extra components (R30/R31/R32/R33/Q6/D11), but could actually reduce the total BOM cost. For instance, RV2 could now be a lower rated part like the 93R1A-R22-A23L. The modified circuit now places RV1/RV2 not only in the programming (reference) current path, but also as the control voltage for the gate of Q6. This device buffers the voltage from its gate (with an offset of a few volts due to its threshold voltage) to the point below R31, in turn connected to R30 and then the output voltage. Therefore,as RV1/RV2 are adjusted to higher resistance value, more voltage will be dropped across R30/R31, and a higher current will flow through them (<1mA max). This current flows through Q6 and into R33, which then causes the reference voltage of U2 to rise. As you must have already figured out, this is a positive feedback loop. Yes, positive feedback, the kind you don’t want to have in stable systems. However, since R33 is lower in resistance than R30/R31 in series, the loop gain is <0dB, and the voltage will converge to a defined value. D11 is there to limit voltage stress over the gate of Q6. The interesting thing about this circuit is that the VDS value of Q6 is defined by the voltage over R32 (and an additional VGS of Q6 which is only a few volts). This means that in the circuit above Q6 will only see ~25V over it, and can be a small and cheap transistor. This modified version puts a voltage of ~270V over RV2, and therefore allows cheaper parts to be used. Please note that the resistors around U1 were also adjusted to support the modified topology.

(Optional) Reducing Voltage Drop over Rectifier:

For the heater supply, especially when set for 6.3V output voltage, there is very little headroom left for regulation. This both necessitates the use of a very low dropout voltage IC, such as the one used in this design, and optimization of the rest of the power path. 6.3VAC, when rectified will generate ~8.9VDC (assuming it isn’t loaded much, and ideal diodes are used). However, to maintain reasonable regulation, you must not only keep the voltage over the IC higher than the dropout voltage (typically quite significantly higher for good regulation, but thankfully the heater isn’t too sensitive and we can flex on this demand), but consider additional effects. For instance, we would like to be able to set the voltage to 6.8VDC in this design, just to have some flexibility. There is also the voltage drop of the resistance of the inductor’s secondary windings, and the forward voltage drop of the rectifier, and the droop over the bulk capacitors as they drain between halves of the power line cycle. There are a couple of other parameters that I’ve neglected here too. So overall, we have very limited headroom.
Since this is a 2×6.3VAC transformer, there are two ways of wiring it for 6.3V operation. This is with either 2 windings in parallel and a bridge rectifier, or as a center-tapped (CT) transformer with a single diode on each of the windings. For the first option, we have the benefit of halving the winding resistance, while for the second there is the benefit of having one less diode drop. The answer as to which is better really depends on the resistance of the winding and the type of diodes being used. I’ve drawn the first option in the schematic, but you could always choose differently if you think this isn’t the best combination for the parts you have available.

For the schematic above, assuming~8.9VDC is the maximum value we can achieve from a 6.3VAC signal, and that we would like to support a 6.8V output voltage setting (at max load this will drop to 6.6V by the ballasting resistors of course), we have ~2.1V left. Lets leave ~500mV for the regulators as they will heat-up at these conditions and therefore dropout will increase to >400mV. We are left with ~1.6V. For the 4×22,000uF capacitors we have in the schematic, and a 50Hz line frequency, the droop over the capactitors can be as high as 1V, although in practice we can expect something closer to ~0.8V.  This leaves us with ~0.8V at most for the rectifier, and this is assuming there are transformer isn’t so heavily loaded that its output voltage drops lower than expected. Therefore, the only way to meet these specifications is to either use a slightly higher secondary voltage (perhaps 7-8VAC), or use a transformer with sufficiently higher VA rating. One more modification we could implement to help meet these requirement would be to replace the rectifier diodes with low voltage drop type, such as Schottky diodes. This is in-fact what I intend to do, by replacing the rectifier drawn in the schematics above with SDT40A100CT diodes. This, along with a transformer with a reasonable VA overhead (~50%) should be able to deliver on the full 6.8V@8A output setting.


As stated in the preface of this post, this is only the first part of (hopefully, if I find the time) a series of posts on the topic. The schematic presented here was drawn by me, but I’m yet to build it and verify the operation. This will probably take some time before I carry out the board layout, order the parts, and start assembling and testing it. In the meantime I hope it will be of value/interest to others. Feel free to contact me with any questions/feedback you might have regarding the schematic via the contact page.

The complete schematic is attached here again in PDF format:
The modified version with reduced voltage stress over RV2 is attached below:

2 thoughts on “DIY Programmable Power-Supply for Vacuum Tubes – Part 1”

  1. Greetings :)!

    I came across your website thanks to Google. Thank you for such a wealth of useful information! I’ve been building solid state headphones amps for several years and would now like to try a few vacuum tube designs. I’ve completed a simple OTL and will start to experiment with some more complex designs. For this purpose, I’d first like to construct an adjustable power supply, and this looks like an excellent candidate.

    Did you ever build this circuit?

    Thanks a ton!

    1. Hi Richard,
      Thanks for reaching out. Unfortunately I never found a sufficient need/time to do this. Whatever little time I found to dedicate to this hobby I’ve spent on other projects as posted on the blog.
      Hope I will be able to revisit this in the future, but I can’t say when (if) this will be.

Leave a Reply

Your email address will not be published. Required fields are marked *