When testing an amplifier, the first thing I generally do is to check voltages. This is what I advise others to do as well. However, the real trick is in determining if the measured voltages are acceptable or not. And herein lies the difficulty.

This has turned into a long and rambling post. This is because even though the word “voltage” may seem simple, the complexities of understanding what is being measured and to what those measurements are being compared are many and varied. And there are a lot of issues to be addressed before a real understanding may be reached. It’s also very important to understand your design. Without a good understanding, the measurements won’t have any real meaning.

Before going into what is actually measured, I’ll talk a little about the different categories of “voltages” and what each mean. This discussion really applies to many of the circuit parameters we measure, but the case of voltages is representative of all of them.

Types of Voltages

When you place the leads of a meter on circuit elements the voltage you read is, for want of a better word, truth. This is what the voltage actually is. The question to be asked is then how does “truth” compare to the other voltages we use. And by other voltages, I mean our assumptions about voltages at each stage of the design and build process. Because these are not what many seem to think they are.

The first category I like to discuss is design voltages. When we look at a schematic or perform a load line analysis, we make assumptions about what the voltage levels should be. This applies to B+ voltages, filament voltages, bias voltages, electrode voltages and everything else we put into a design. These voltages can be both assumptions, like the B+ voltage and filament voltages, and educated guesses, like bias voltages, based on our knowledge (and assumptions) of the circuit elements and how they operate. But design voltages are just what the name implies, they are design points. They are not any kind of objective truth.

The second category I like to discuss is characteristic voltages. These are the voltages we obtain from data sheets and component markings. These voltages are statistical in nature. The value marked on a component is a characteristic, not a measurement. As such it is based on a statistical set of measurements extrapolated to all similar components. When data sheets contain numbers for µ or gm, when they publish characteristic curves, these numbers are statistical approximations based on a sample of production. And since data sheets rarely, if ever, contain variance numbers for these things, they are actually of limited value. They guide the design, but are not hard and fast rules.

Next comes the most contentious category, published voltages. These are the voltages actually printed on schematics and contained in design documentation. For these it is very important to use lots of judgement and common sense. There are many reasons for this.

The first is due to errors. Not everything gets written down properly. Especially for noncommercial and DIY designs. Mistakes are made. Things get written down incorrectly. If the analysis says the number is wrong, then it probably is.

The next reason is what I call the “Leo Rule” after Leo Fender. If you look at a schematic for a commercial product that contains one hundred numbers (voltages, currents, component values, etc.), then at least five of them will be incorrect. I don’t know why this rule holds but it seems to rather well. There are rumors that, at least of the part of Fender, these errors were deliberate to keep others from copying their amps. But that seems unlikely. More probable is that in most commercial products, the documentation gets developed concurrently with the product. And, more often than not, changes made in one branch don’t always get reflected in the other. A lesson I learned well over a 30+ year Engineer career.

Another reason is lack of knowledge. This is a significant problem in the DIY world. Many DIY projects found on the internet, instead of being the result of careful design and prototyping, are the result of hacked together designs, copied topologies, and wild assumptions. Many of these designs show numbers which at first glance most should know are incorrect. I contend that this is one reason for the widespread use of the (improperly named) SRPP signal stage. It is not because this topology makes a superior audio stage (it doesn’t) but because it is easy to get one that functions without truly knowing why.

The final category to be discussed is measured voltages. Regardless of what was designed, intended, or assumed measured voltages are a reflection of what is actually happening in the circuit. This can be a difficult fact to accept. The most often heard phrase when measuring circuit voltages is “that’s not right”. Of course the measured voltage is the voltage. It may be unexpected, it may be a deviation from the design point, it may not match published values, but it is, in all probability, correct for the circuit you’re testing. Whether it’s correct for the circuit you think you’re testing is another question altogether.

Voltage Measurement

With the understanding of what is meant by “voltage” the next topic is how do we go about determining if the voltages measured are acceptable or indicative of some error or problem. I’ll discuss my typical approach to initial voltage testing to highlight my methodology.

I always start with a check out of the mains wiring and power transformer with no tubes installed. I measure the mains voltage at the transformer input, and the open circuit voltages on all the transformer secondaries. This information provides me with immediate information about transformer wiring and operation. It is important to include the measurement of the mains voltage at this point. Mains voltage can vary by ±5% or more depending on location and grid conditions. Knowing the input conditions sheds light on voltage measurements to be taken later on.

These “open circuit” transformer voltages will all be higher than normal. Most transformers are rated for voltage at load. As an example, the 660v secondary of the Edcor XPWR178 measures almost 700v with no load (≈6% high). Also note that variations in mains voltages from the rated condition will be reflected on the output voltages as well. It is important to make sure that the circuit can handle this higher than normal voltage until currents start flowing in all stages. The filament voltages will also be higher than normal, sometimes by a significant margin. But this is to be expected. Most receiving tubes can handle a fairly wide variation in operating filament voltage.

The next check is with the tubes installed, the output transformers loaded with dummy loads, that the inputs resistively terminated. Just some shorting plugs works well for this or, if the volume control is on the inputs, simply turning the volume to zero will do. This effectively grounds the grids of the first signal stage. All voltages are measured referenced to the signal ground. I also never directly measure grid voltages. The loading of most meters will cause these voltage to read in error. It is better to calculate grid voltage from bias voltages of other tube electrodes. Remember that vacuum tube voltages are referenced to the cathode terminal, not to signal ground. So if measured dc voltages at the plate and cathode referenced to signal ground are 110v and 3.5v respectively, then the plate voltage (i.e. the voltage referenced on the data sheet plate characteristics and the load line analysis) is really 106.5v and not 110v. And, if the grid is properly grounded with an appropriate grid resistor, the grid voltage, and hence the bias voltage, is -3.5v (i.e. the voltage relative to the cathode) and not zero volts.

At startup I also measure the B+ as the tubes all come to temperature. Normally the B+ will rise above the steady state level to some extent. How much depends on the warmup characteristics of the rectifier tube and the power tubes. You should check to make sure that this transient B+ voltage does not exceed the voltage ratings of either the filter capacitors or the coupling capacitors.

Once everything has come to temperature, at least 60 seconds and maybe up to a couple of minutes, the bulk of voltage measurements may then be taken. These include plate, screen, and cathode voltages, and voltages within the power distribution filters. Grid voltages should be inferred from the topology and the other electrode voltages.

Voltage Assessment and Interpretation

The overarching rule in assessing voltages is that they must be congruous with each other and what we know about the circuit. For example, it you have a two stage B+ filter with series resistive elements, then the current through both series elements should be the same. As such, if they are of the same value, then the voltages measure across them should be equal. Otherwise current is flowing somewhere it shouldn’t be. However, it is important to remember that large electrolytic capacitors can have leakage currents in the single digit milliamp range. These currents are normal. Further, the current that you calculate flowing through these resistors (ΔV/R) should be close of what is calculated for the entire circuit. I’ll discuss what “close” actually means in a minute.

The best way to begin assessing voltages is to look at individual tube operating points. For cathode bias stages, the total current through the tube flows through the cathode resistor. Since the voltage across the bias resistor is easily measured, the current can be calculated by dividing the cathode voltage by the resistor value. With the plate voltage, cathode voltage, and cathode current in hand (or the plate current if you are not using a triode) an operating point can be plotted on the plate characteristics. This operating point should be “close” to the design value.

So what does “close” actually mean? First it should be within about 10% to 15% of the calculated values based on the actual B+ voltage, the actual plate load resistance, and the calculated bias voltage. It doesn’t have to exactly match because it’s a measurement off of a single tube. There exists a lot of variance in most published characteristic curves. More important is that all the measurements are self consistent. For example, in a common cathode triode stage, the calculated plate current should match the calculated cathode current. The total of voltage drops across the plate load resistor, the plate to cathode path of the tube, and the cathode bias resistor should add up to the B+ voltage. If they don’t, then something is wrong.

Summary

The overarching lesson here is that before you start to measure voltages, you should have a good idea about what the measured values should be. This idea should be based on the design and analysis of the circuit. By carefully measuring voltages and checking them against what was calculated, any issues with the operation of the circuit should stand out.

As always, questions and comments are welcome.