The EL95 is a low-output power pentode, with a maximum anode dissipation of only 6W. In a single-ended application, which typically only run around 50% efficiency, a single EL95 would output around 3W, which is perfect for a bedroom or studio amp that you want to push into overdrive without making your ears bleed.

Experience tells us that power tubes perform best when they are operated close to the knee of the maximum dissipation curve. Looking at the characteristics curves for the EL95, we see that the knee is in the area of 200V. So we need to select a power transformer that will provide us a voltage in that area. Limiting ourselves to Hammond 269 series transformers, we have two options with 6.3 volt filament supply - the Hammond 269BX (300VCT @ 86mA) and the Hammond 269EX (380VCT @ 71mA, $45 @ Doberman). The equation to determine the rectified voltage from a transformer is:

Factoring a small forward diode voltage drop, the 269BX would rectify to 210V, while the 269EX rectifies to 266V. Having a higher supply voltage helps create more clean headroom in the preamp, so let's use the 269EX. We will use a "pi" filter after the rectifier to guarantee a ripple-free supply voltage, which will also provide a voltage drop. We'll visit the design of this filter later, but we can design it to attain whatever operating point we desire. Let's shoot for a Vb+ of 240V, which is one of the operating points specified in the EL95 data sheet. This is a very workable Vb+, right in the center between 0V and 550V, the maximum peak anode voltage. This should provide for fairly even voltage swing between these two extremes, making this output stage close to center-biased.

Anode Circuit

In a single-ended power amplifier, the anode circuit is simply the output transformer, connected to the anode at one end and the anode supply line at the other. All we need to do is determine the transformer's primary impedance, which we do by applying Watt's and Ohm's Law:

P = VI, V = IR
I = V / R
P = V^2 / R
R = V^2 / P
Zout = Va^2 / Pa(max)
Zout = 240^2 / 6.0
Zout = 9600 ohms

Fortunately, this is close to 10K, an option offered by the affordable and easily available Hammond 125SE series transformers. To choose which one, we need to know its quiescent DC bias current and power dissipation. This is the same as the anode current, which we can calculate using Ohm's Law:

V = I * R
Va = Ia * Zout
Ia = 240V / 10K
Ia = 24mA

The 125ASE is rated at 3W, 25mA DC bias. That's cutting it pretty close, both current and power-wise, which may cause distortion and "sag" that some people find desirable, but could perhaps cause earlier failure. The 125BSE is rated at 5W, 45mA DC bias. This would probably be a better choice. I will probably build with the 125CSE (8W, 60mA DC bias), which I can get for cheaper ($45 @ Doberman, $50 @ Mouser) than the BSE ($59 @ Mouser). Never hurts to have a little more transformer than you need.

Now let's plot the load line for this stage. This line, drawn on the tube's characteristics curves, represents the operating range of the tube, from minimum anode voltage (0V)/max anode current (24mA) to max anode voltage (240V)/minimum anode current (0mA). We plot these two points, and draw a straight line between them (blue line). You can see that this is well below the maximum dissipation curve:

We want to run the stage closer to the maximum dissipation curve, which will give us more output and drive. We can adjust this by biasing the stage appropriately. Because the transformer is a reactive load and has a very low DC resistance (Zout is only presented to AC signals), there is negligible DC voltage drop across the transformer, and the quiescent anode voltage will always be equal to Vb+. We can therefore draw a vertical line at Va = Vb+, and know that the bias point will fall somewhere on this line. We will select an operating point that is slightly below the maximum dissipation curve, say 24mA (purple dot). To get the load line for this new operating point, we simply slide the load line up the graph to intersect with this new operating point, maintaining the load line's gradient (purple line). As we do this, we can see that the anode voltage will not swing above its maximum peak rating (550V), which is important because it means that the anode will not be destroyed at the high end of its swing.

Screen-Grid Circuit

Current flow from the anode is controlled by the current supplied to the screen-grid. Therefore, to obtain our desired quiescent operating point (240V, 24mA), we need to determine this current (Is). If we look at the data sheet, we see that the screen current (Is2) is 4.5mA when the anode current (Ia2) is 24mA. Hey, that's exactly our target quiescent bias point! For reference, that is a 5.33:1 ratio between plate and screen currents, and if we wanted to find the screen current at another quiescent anode current, we would calculate it using this ratio:

Ia / ls = Ia2 / Is2
Is = 24mA * (4.5mA / 24mA)
Is = 4.5mA

There are two resistors that control this current. The first is a choke or dropper resistor (Rd) in the power supply that forms, with a filter capacitor, an RC smoothing filter between the anode circuit and the screen. It provides isolation and a small voltage drop. This value is usually 470R or 1K, to keep the voltage drop small. In the power supply of most guitar amplifiers, the preamp supply lines are downstream from this resistor, meaning that the preamp will also draw current across this resistor, increasing the voltage drop. If we estimate the preamp current draw at 5mA, that's a total of 9.5mA of current draw across this resistor. Applying Ohm's law, we calculate the quiescent voltage drop (Vdq) across this resistor:

V = IR
Vdq = 9.5mA * 1K
Vdq = 9.5V

The second resistor that controls screen current is the screen-grid stopper (Rg2). This resistor is here to prevent screen current from going through the roof and destroying the grid when the anode voltage goes low. Following Merlin's method (which I admit I don't completely follow), we look at the load line and see that the -8V curve crosses the load line at 30mA. Because the function of the screen-grid stopper compresses the curves to prevent rapid current increase, we want the -4V curve to compress down to where the -8V curve is. So, on the mutual characteristics graph, we plot -4V, 30mA (green dot):

We can see that this point is below the Ig2 = 200V curve, we will call it 175V. This means that we need to drop Vb+ - Vg2 = 240 - 175 = 65V across the two resistors under peak conditions. According to the load line, the peak anode current is 48mA, so using Ohm's law:

V = IR
65V = 48mA * Rg2
Rg2 = 65V / 48mA
Rg2 = 1354R

We will add a little buffer and go with 2K, which will add a little extra compression. As this resistor's function is to protect the tube from overload, it should have a high power rating: 3W or better. Using Ohm's law and the previously calculated quiescent screen current (Isq = 4.5mA), we can determine the voltage drop across this resistor:

Now that we know the current demand of the screen circuit, we can design the pi filter in the power supply. The purpose of the pi filter is to remove ripple from the rectified voltage, "smoothing" it into a stable DC voltage. The first capacitor after the rectifier is the most important, and is called the "reservoir" capacitor. It is typically in the range of 33uF - 220uF, and should be rated to tolerate the voltage it expects to see. In this case, we'll use a 47uF 350V cap. A resistor is connected between the rectifier output and the Vb+1 supply line, which connects to the non-anode side of the output transformer. Another filter cap is connected between Vb+1 and ground. The resistor determines the voltage drop between the rectifier output and Vb+1. At DC, there is no voltage drop across the transformer secondary, so this resistor forms a voltage divider with the impedance of the anode, Za (10K). However, we also need to factor in the fact that the pentode's screen and the preamp also draw current across this resistor, increasing the voltage drop. So, estimating:

The closest standard value is 750R, which is plenty close.

Biasing

We plot the quiescent anode voltage (Ia = 24mA) and screen voltage (Vsq = 221.5V) on the mutual characteristics curves (blue dot), which shows us our desired grid-to-cathode bias voltage Vgk = -7.5V. The cathode current is equal to the anode current plus the screen current:

Ik = Ia + Isq
Ik = 24mA + 4.5mA
Ik = 28.5mA

And applying Ohm's law we get the cathode resistor value (Rk):

V = IR
Vgk = Ik * Rk
Rk = Vgk / Ik
Rk = 7.5V / 28.5mA
Rk = 263R

The nearest standard value is 270R, which is close enough. Power dissipation (Pk) across this resistor is:

P = VI
Pk = Vk * Ik
Pk = 7.5V * 28.5mA
Pk = 0.21W

So we should use a 1/2W resistor or better.

Cathode Bypass Capacitor

The cathode bypass cap boosts the gain of the stage by removing negative feedback, and has the added bonus of introducing a low-frequency roll-off. To determine the roll-off point, we use the following:

f = 1 / (2 * pi * R * C)

If we want this stage to roll off signals below audible frequencies, which rob the amplifier of power, we choose a roll-off point below 20Hz. 10Hz gives us a little buffer.

10Hz = 1 / (2 * pi * 270R * Ck)
Ck = 1 / (2 * pi * 270R * 10Hz)
Ck = 58.9uF

The closest standard values are 47uF and 68uF, we will use 68uF, which will make the roll-off point 8.66Hz. The cap should be rated for at least 3x the expected cathode voltage (7.5), so we will use a 25V capacitor.

Grid-Reference Resistor

The grid-reference resistor sits between the grid and ground, biasing the grid at ground potential, allowing comparison to the slightly more positive cathode voltage. This value also constitutes the load seen by the previous preamp stage. Consulting the data sheet, we see that the maximum value for the grid-reference resistor (Rg1) is 2Meg. We'll use 1Meg, because the role of this resistor will probably be performed by a 1Meg-A master volume pot.

Grid-Stopper Resistor

This guy prevents high-frequency oscillation, possible values might be on the tube's data sheet, but aren't in the case of the EL95. We'll take a wild guess of 2K for design purposes, and tweak the value during implementation if oscillations occur.

Tube Amps## Table of Contents

This design will follow Merlin Blencowe's single-ended output stage design procedure. Many thanks for his great site and the assistance he lends at the AX84.com forum.

## EL95 Specs

(from NJ7P.org and Frank's Electron Tube Data Sheets)Pa(max) = 6W

Va(max) = 300W

Va(peak) = 550V

Rg1(max) = 2Meg

Za = 10K

## Operating Voltage (Vb+)

Experience tells us that power tubes perform best when they are operated close to the knee of the maximum dissipation curve. Looking at the characteristics curves for the EL95, we see that the knee is in the area of 200V. So we need to select a power transformer that will provide us a voltage in that area. Limiting ourselves to Hammond 269 series transformers, we have two options with 6.3 volt filament supply - the Hammond 269BX (300VCT @ 86mA) and the Hammond 269EX (380VCT @ 71mA, $45 @ Doberman). The equation to determine the rectified voltage from a transformer is:

Vrect = (Vsec / 2) * SQRT(2)

Vrect = (Vsec / 2) * 1.414

Factoring a small forward diode voltage drop, the 269BX would rectify to 210V, while the 269EX rectifies to 266V. Having a higher supply voltage helps create more clean headroom in the preamp, so let's use the 269EX. We will use a "pi" filter after the rectifier to guarantee a ripple-free supply voltage, which will also provide a voltage drop. We'll visit the design of this filter later, but we can design it to attain whatever operating point we desire. Let's shoot for a Vb+ of 240V, which is one of the operating points specified in the EL95 data sheet. This is a very workable Vb+, right in the center between 0V and 550V, the maximum

peakanode voltage. This should provide for fairly even voltage swing between these two extremes, making this output stage close to center-biased.## Anode Circuit

In a single-ended power amplifier, the anode circuit is simply the output transformer, connected to the anode at one end and the anode supply line at the other. All we need to do is determine the transformer's primary impedance, which we do by applying Watt's and Ohm's Law:

P = VI, V = IR

I = V / R

P = V^2 / R

R = V^2 / P

Zout = Va^2 / Pa(max)

Zout = 240^2 / 6.0

Zout = 9600 ohms

Fortunately, this is close to 10K, an option offered by the affordable and easily available Hammond 125SE series transformers. To choose which one, we need to know its quiescent DC bias current and power dissipation. This is the same as the anode current, which we can calculate using Ohm's Law:

V = I * R

Va = Ia * Zout

Ia = 240V / 10K

Ia = 24mA

The 125ASE is rated at 3W, 25mA DC bias. That's cutting it pretty close, both current and power-wise, which may cause distortion and "sag" that some people find desirable, but could perhaps cause earlier failure. The 125BSE is rated at 5W, 45mA DC bias. This would probably be a better choice. I will probably build with the 125CSE (8W, 60mA DC bias), which I can get for cheaper ($45 @ Doberman, $50 @ Mouser) than the BSE ($59 @ Mouser). Never hurts to have a little more transformer than you need.

Now let's plot the load line for this stage. This line, drawn on the tube's characteristics curves, represents the operating range of the tube, from minimum anode voltage (0V)/max anode current (24mA) to max anode voltage (240V)/minimum anode current (0mA). We plot these two points, and draw a straight line between them (blue line). You can see that this is well below the maximum dissipation curve:

We want to run the stage closer to the maximum dissipation curve, which will give us more output and drive. We can adjust this by biasing the stage appropriately. Because the transformer is a reactive load and has a very low DC resistance (Zout is only presented to AC signals), there is negligible DC voltage drop across the transformer, and the quiescent anode voltage will always be equal to Vb+. We can therefore draw a vertical line at Va = Vb+, and know that the bias point will fall somewhere on this line. We will select an operating point that is slightly below the maximum dissipation curve, say 24mA (purple dot). To get the load line for this new operating point, we simply slide the load line up the graph to intersect with this new operating point, maintaining the load line's gradient (purple line). As we do this, we can see that the anode voltage will not swing above its maximum

peakrating (550V), which is important because it means that the anode will not be destroyed at the high end of its swing.## Screen-Grid Circuit

Current flow from the anode is controlled by the current supplied to the screen-grid. Therefore, to obtain our desired quiescent operating point (240V, 24mA), we need to determine this current (Is). If we look at the data sheet, we see that the screen current (Is2) is 4.5mA when the anode current (Ia2) is 24mA. Hey, that's exactly our target quiescent bias point! For reference, that is a 5.33:1 ratio between plate and screen currents, and if we wanted to find the screen current at another quiescent anode current, we would calculate it using this ratio:

Ia / ls = Ia2 / Is2

Is = 24mA * (4.5mA / 24mA)

Is = 4.5mA

There are two resistors that control this current. The first is a choke or dropper resistor (Rd) in the power supply that forms, with a filter capacitor, an RC smoothing filter between the anode circuit and the screen. It provides isolation and a small voltage drop. This value is usually 470R or 1K, to keep the voltage drop small. In the power supply of most guitar amplifiers, the preamp supply lines are downstream from this resistor, meaning that the preamp will also draw current across this resistor, increasing the voltage drop. If we estimate the preamp current draw at 5mA, that's a total of 9.5mA of current draw across this resistor. Applying Ohm's law, we calculate the quiescent voltage drop (Vdq) across this resistor:

V = IR

Vdq = 9.5mA * 1K

Vdq = 9.5V

The second resistor that controls screen current is the screen-grid stopper (Rg2). This resistor is here to prevent screen current from going through the roof and destroying the grid when the anode voltage goes low. Following Merlin's method (which I admit I don't completely follow), we look at the load line and see that the -8V curve crosses the load line at 30mA. Because the function of the screen-grid stopper compresses the curves to prevent rapid current increase, we want the -4V curve to compress down to where the -8V curve is. So, on the mutual characteristics graph, we plot -4V, 30mA (green dot):

We can see that this point is below the Ig2 = 200V curve, we will call it 175V. This means that we need to drop Vb+ - Vg2 = 240 - 175 = 65V across the two resistors under peak conditions. According to the load line, the peak anode current is 48mA, so using Ohm's law:

V = IR

65V = 48mA * Rg2

Rg2 = 65V / 48mA

Rg2 = 1354R

We will add a little buffer and go with 2K, which will add a little extra compression. As this resistor's function is to protect the tube from overload, it should have a high power rating: 3W or better. Using Ohm's law and the previously calculated quiescent screen current (Isq = 4.5mA), we can determine the voltage drop across this resistor:

V = IR

Vsq = 4.5mA * 2K

Vsq = 9V

So the quiescent screen voltage will be:

Vsq = Vb+ - Vdq - Vsq

Vsq = 240V - 9.5V - 9V

Vsq = 221.5V

## Pi Filter

Now that we know the current demand of the screen circuit, we can design the pi filter in the power supply. The purpose of the pi filter is to remove ripple from the rectified voltage, "smoothing" it into a stable DC voltage. The first capacitor after the rectifier is the most important, and is called the "reservoir" capacitor. It is typically in the range of 33uF - 220uF, and should be rated to tolerate the voltage it expects to see. In this case, we'll use a 47uF 350V cap. A resistor is connected between the rectifier output and the Vb+1 supply line, which connects to the non-anode side of the output transformer. Another filter cap is connected between Vb+1 and ground. The resistor determines the voltage drop between the rectifier output and Vb+1. At DC, there is no voltage drop across the transformer secondary, so this resistor forms a voltage divider with the impedance of the anode, Za (10K). However, we also need to factor in the fact that the pentode's screen and the preamp also draw current across this resistor, increasing the voltage drop. So, estimating:

Vrect = Va + Vpi

Vpi = 266V - 240V

Vpi = 26V

V = IR

Rpi = Vpi / (Ipre + Iscreen + Ia)

Rpi = 26V / (5mA + 4.5mA + 24mA)

Rpi = 26V / 33.5mA

Rpi = 776R

The closest standard value is 750R, which is plenty close.

## Biasing

We plot the quiescent anode voltage (Ia = 24mA) and screen voltage (Vsq = 221.5V) on the mutual characteristics curves (blue dot), which shows us our desired grid-to-cathode bias voltage Vgk = -7.5V. The cathode current is equal to the anode current plus the screen current:

Ik = Ia + Isq

Ik = 24mA + 4.5mA

Ik = 28.5mA

And applying Ohm's law we get the cathode resistor value (Rk):

V = IR

Vgk = Ik * Rk

Rk = Vgk / Ik

Rk = 7.5V / 28.5mA

Rk = 263R

The nearest standard value is 270R, which is close enough. Power dissipation (Pk) across this resistor is:

P = VI

Pk = Vk * Ik

Pk = 7.5V * 28.5mA

Pk = 0.21W

So we should use a 1/2W resistor or better.

## Cathode Bypass Capacitor

The cathode bypass cap boosts the gain of the stage by removing negative feedback, and has the added bonus of introducing a low-frequency roll-off. To determine the roll-off point, we use the following:

f = 1 / (2 * pi * R * C)

If we want this stage to roll off signals below audible frequencies, which rob the amplifier of power, we choose a roll-off point below 20Hz. 10Hz gives us a little buffer.

10Hz = 1 / (2 * pi * 270R * Ck)

Ck = 1 / (2 * pi * 270R * 10Hz)

Ck = 58.9uF

The closest standard values are 47uF and 68uF, we will use 68uF, which will make the roll-off point 8.66Hz. The cap should be rated for at least 3x the expected cathode voltage (7.5), so we will use a 25V capacitor.

## Grid-Reference Resistor

The grid-reference resistor sits between the grid and ground, biasing the grid at ground potential, allowing comparison to the slightly more positive cathode voltage. This value also constitutes the load seen by the previous preamp stage. Consulting the data sheet, we see that the maximum value for the grid-reference resistor (Rg1) is 2Meg. We'll use 1Meg, because the role of this resistor will probably be performed by a 1Meg-A master volume pot.

## Grid-Stopper Resistor

This guy prevents high-frequency oscillation, possible values might be on the tube's data sheet, but aren't in the case of the EL95. We'll take a wild guess of 2K for design purposes, and tweak the value during implementation if oscillations occur.

## Design Schematic

## Build Photos