AN-02. Circuit topology for tube power supply.

By Jac van de Walle


Perhaps the rectifier circuit is the most misunderstood part of a tube amplifier. I think that is because exceeding the data sheet limits doesn seem to do any harm, and also the tube doesn't get really hot from this. For most designers that seems good enough. However, rectifier tubes can suffer badly, without getting too hot. When peak current is too high, the tube will loose emission, and once that happens, it will develop higher forward voltage. That again develops more heat, and a self destroy process begins after a period of failure free operation. How long will it take before that happens? That is not a good question, but even so it may take 1000 hours. The right question is, what must we do, to prevent this, and get a multiple life time.

Why this is done wrong often

The problem with this is, a measurement of the plate current of the rectifier is needed, or at least use the diagrams from the tube data sheet. Using an oscillosope here, is not easy to do, even if you have one. On the other hand, just knock something together seems to work most of the time. So this is what most users do. There is a funny saying, why men like wood chopping so much. That is because you can see the result of your work immediately.

So the KEY thing is the peak current. If not using an oscilloscope, at least read keep a safe distance from maximum values, and use the deratign curves.

Even JJ, who publishes data sheets with as little as data as possible, publishes derating curves for the 5U4GB, using a large, and a small capacitor. That should be a clear sign, this is important.


Above is a cut out from JJ data sheet 5U4GB

A better data sheet for this is the RCA 5U4G. (But keep apart 5U4G and 5U4GB is specified different). The RCA data sheet is detailed, and yet easy to understand. You can see there, if you want maximum voltage, you can not take maximum current, and vice versa. Do you want the maximum capacitor, that is possible, but not at maximum voltage. You will find the trade offs in there.

Most unfortunately, such considerations are considered not necessary by a many. The internet is full of schematics, and an that is preferred above studying data sheets. A big risk is, these schematics are made often by people who just knocked it together, and they consider it a good design.

Also copying schematics from companies can be a mistake, because this category designers, has a strong tendency to replace large coils by oversized capacitors. They save some money on the hardware, and let the user replace the rectifier tubes more often.

Some of the most common errors:

  1. The DC resistance of the transformer Raa winding is an ESSENTIAL part of the circuit design. Any mistake with that, is slowly killing every good rectifier. People are often totally surprized this requirement exists at all, and transformer manufacturers never heard about it anyway. So on the market are very many unqualified transformers and amplifiers. Though this is clearly written in every tube data sheet ever since 100 years. Also a book about rectifier circuit design, if it doesn't mention this factor, stop reading it.
  2. Adding 2...5x more than the maximum allowed capacitance as a first capacitor for the rectifier. Just because 'it works', doesn't mean this is not a tube killer. (Also because of the next point)
  3. Draw excessive surge current from a rectifier tube during cold start. Caused by a low cost choke, which saturates at start up, too big capacitors, or other causes in the amplifier itself, at start up, such as DC coupled tube stages, adjustable bias, and other risky things.
  4. Combine the above by not measuring peak current, and start up current. Just copy schematics, makes you copy the mistakes too. Just to give the classical warning once more, when you see more than 33uF on a 5U4G, or more than 4uF on a 274B tube, such a schematic is made by an ignorant designer. So this can already by seen from picture files. Whereas transformer Raa Resistance can be measured simple with a low cost multi meter. Yet while peak current is hard to measure, at least you should make a simulation, for instance with 'PSU Designer', which is freeware for DIY, and the best program of it's kind. PSU Designer will work quickly and easy, and simulate peak current with Oscilloscope-like results, at any point of the circuit you select. For the hard core users there is LTSPICE freeware from Analog Devices. Though LTSPICE will take a very long time to understand when you are new to it. You can not see this when using PSU designer, but it is written with LTSPICE.

Have a look at the above overview. It's just numbers, but you can extract from this why it is the way it is.
  1. Let's begin with RGN2504 - RGN4004 compare. Now both tubes are sold for crazy prices on Ebay, specially the RGN4004 if NOS goes above 2000 Euro. Why is that? Well people see in the data sheet it can supply 300mA. So many of the larger amplifiers can be build with the tube. Whereas the "weaker" RGN2505 can supply only 180mA. It is goes somewhere between 1000 and 2000 Euro. What is behind these specifications? When looking at the output voltage, we see RGN2504 is the winner. It is just RGN4004 is a low voltage, high current tube, and RGN2505 is the opposite.
  2. Transformer DC resistance. This is something, you can surprise your transformer seller with. This is such a CRITICAL number! On the one hand it has to be high enough, to prevent tube sparking. On the other hand it has to be low, to prevent power loss. Also winding with such a very thin wore, is not always what they like. But such considerations may not bypass limits, the tube has. And no, there are not tubes which does no have such limits. If that limit it not specified, it means rather the tube vendor does not know himself. But it does not mean, with such tubes you can use them the same way as a silicon diode (which is the next item, below here). Look for this at AZ50. Which is by itself not better than RGN4004, it has only a better qualified set of data. Here, Philips brings the transformer resistance (Rt) in relation to the trade off between maximum current and maximum voltage. You can see AZ50 can almost to the 300mA/350V of the RGN4004. But it even exceeds the 500V/180mA of the RGN2504. How is that magic done? Can only tube really combine this? Yes it can when you pay attention to the transformer Rt. So at low Rt, 100Ohms you can go to 300mA. But for 500V (to prevent sparking) you need 200 Ohms Rt, which is less of a problem at 500V, because current is lower here anyway, and Rt loss is lower too.

Do not treat tube diodes like solid state diodes.

To give you some sensitivity for this item, take very good note of the following. When comparing tubes with solid state devices, the elementary difference is: With solid state devices it is more difficult to build such for high voltage, as it is for high current. With tubes, it is reversed. It is difficult to build such for high current, and high voltage was never a problem. We have to realize, when stressing exactly this natural limit, the result will cause problems. So solid state diodes have less of a problem with repeating peak current for 25 years of use, but they definitely will have a reliability problem, with repeating the peak voltage for a life time of 25 years.

With tubes, this above situation is reversed. Tubes have little problems when the peak voltage, but they will develop problems quickly when the peak current is reached continuously. So design considerations are totally another, and it would be completely wrong to apply the design considerations for solid state diodes to vacuum diodes as well, saying 'a diode is a diode'.

With solid state devices we all know, you can not exceed the peak voltage. If you do this only once, they can be dead..

Even greater care should be taken with tubes, since these are devices with limited lifetime, and they are expensive. If you abuse them with peak current at start up, or exceed it during normal operation, lifetime will be shortened severely. We repeat the same as before: As this is impossible to calculate with precision, good designers will measure it with a good instrument, or alternatively over dimension the design ratings. This is in order to stay safely away from the damaging limits.

Many good books have been written about tube power supplies, and of course we can not compress this in a few pages here. Keep three elementary things in mind, to get good tube lifetime: 1) Respect the maximum first capacitor value in the datasheet, and be aware when you use a maximum factor on the one end, you get a compromise on the other. 2) Understand and apply the derating curves, meaning at maximum current, you can not use the maximum voltage, and vice versa. 3) Use the MINIMUM copper resistance needed for a TUBE power supply transformer. Transformer building companies can not know what you need, so without specifications from your side, you get just 'something'. Which is only fine when you need 'anything'. This is so for solid state, but this is not so for tubes. You need to specify the copper resistance to them, and if they can't do the value the tube requires, you need to add the difference by using external resistors.


These rules above, need to be respected to get good tube life, and failure free operation. For maximum lifetime of any object, and sure for electron tubes, keep some distance from maximum limits. That is: maximum capacitor value, maximum voltage, maximum current, and minimum copper resistance.

Low residual hum

With a solid state power supply, the last final bit of hum can be eliminated by using larger capacitors. This is no problem for the diodes, as they do not suffer from transient current. With a tube power supply, the last final bit of hum, is not present because capacitors are too small, but because the choke is too small. Moreover, if circuits are not ideal you would be tempted to tweak components values, but better would be to use an improved circuit. The most commonly used circuit is Circuit #1, as below here, but this is the most unintelligent. In very many cases, simply use circuit #2 already brings an improvement. We call this 'star grounding'. Tough Circuit#2 is far from ideal, and Circuit #4 does the same even better and with less wiring. Circuit #3 shows one of those typical little things, you can observe when somebody does not treat a tube like a solid state diode. This makes sure the left side of the tube heater wire has the same current (and the same temperature by that) as the right side of the heater wire. Whereas Circuit #4 is such a logical and easy thing to do, but people do not understand why, and take Circuit #1, this is really painful. It is allowed to use Circuit #4, also if you do not understand why it is better, as it is free of patent rights. The words are: No ground loop risk, and electrical stray field. Circuit #5 is the ultimate circuit. Only possible with Lundahl chokes. This is the most beneficial circuit ever, for lowest hum. The choke is not only used here as an inductor, but also creates inductive decoupling of the complete rectifier parts from the amplifier circuit. There is no direct connection any more! All hum signals, no matter how they try to flow, must pass a choke now. Please also read the short explanations on the right, below here. After that I hope you understand why we say it is so silly to find Circuit #1 in most 'professional' designs.

Circuit 1: This is the most commonly used. What is not so nice about this circuit, will get clearer when you understand the benefits of the other circuits here.
Design Quality of Circuit1: One Star *

Circuit 2: Added here is a bleeder resistor, a ground connection point, and if needed an external windings resistor. The ground connection point prevents any faulty ground path. This is called 'star' grounding.
Design Quality of Circuit 2: Two Stars **

Circuit 3: This avoids the rectified DC current flows through the transformer heater winding. This will reduce Transformer hum. However circuit #4 is more simplified and will do the same.
Design Quality of Circuit 3: Three Stars ***

Circuit 4: This circuit refers the Choke AC and DC electrical field to ground, where they can cause less problems. Also the ground path for the first capacitor is now forced in a correct way, same as in circuit 3.
Design Quality of Circuit 4: Four Stars ****

Circuit 5: This is how to connect a double coil Choke, such as the Lundahl LL1673 or similar products. Beware the polarity of the connections. (the dots in the circuit diagram here). The importance of this circuit is very high. We have complete inductive separation of the transformer from the amplifier. Definitely, the transformer capacitance from primary to secondary can not inject an AC hum current into the amplifier any more.
Design Quality of Circuit 5: Five Stars *****

Bleeder Resistor:

This is to empty all capacitors after power off. Using no bleeder resistor is safety issue, as voltage may stay on the capacitors for a long time. Also use of no bleeder is a reason for a sparking rectifier, as a repeated switch on, with partially charged capacitor may trigger a spark. Make sure the resistor discharges all capacitors to less than 40V in one minute. Do not mount the bleeder directly on a capacitor, because this heats the capacitor, and this is not good for reliability.


Peak current is the killing factors for a tube rectifier. First, there is the cold start up peak current. That can be 0.5 to 2 seconds, in which the amplifier and empty capacitors draw a multiple of normal current, doing so when the rectifier is barely working, and the wear out this moment is equivalent or larger than the whole rest of the session. The second factor is continuous peak current, so when the amplifier has warmed up. The way to measure this is by using a current probe, or by inserting a small resistor in the tube circuit, and make differential voltage measurement across it, using two probe tips.