Tweet
This thread describes a project that made a guitar amp designed to have a set of characteristics including some that are not so common with guitar amps. In addition the design implementation is constrained so that it can be built by modifying a readily available kit. For the kit I have chosen the tubedepot.com 18W clone (an EL84 amp). This kit requires that you drill a G10 card and install turrets, and so you have a very high degree of flexibility in what you actually build. This modification replaces/adds resistors and capacitors, and even replaces a preamp tube with a pin compatible one, but no significant chassis work was performed and of course the transformers were not changed. (I also needed to relabel the front panel, but that was not so hard with a label maker from an office supply store.) The same mods should be possible in another kit that uses EL34s and gets about 50 watts.
Summary of required characteristics:
1. Low noise front end without loss of stability
2. Able to drive a medium impedance (25K) volume pedal located iust before the last gain control (master volume) with no compromise in performance.
3. Single channel, high gain but flexible enough to adjust to a perfectly clean signal (total of three gain controls)
4. Blocking distortion free (no cut off of preamp stages with any gain setting)
5. Fender/Marshal type of bass-mid-treble tone stack.
Why these characteristics?
1. Low noise: The reason is primarily to accommodate non standard guitar outputs. First is the guitar using medium impedance pickups, where the voltage might be three times lower than, say, a typical single coil pickup (impedance 10 times or more lower). This is a passive guitar with 25K or 10K pots, often with a switched capacitor/resistor network for varying the resonant frequency that is set by pickup inductance/cable capacitance in a standard guitar. The tone is independent of the volume setting; because of the lower impedance; it is possible to use the volume control in ways you cannot in a standard guitar. Second is the wide dynamic range active guitar. This guitar buffers the pickups with an amplifier that can be as simple as a FET source follower. You can use very high output pickups and still get resonances at or above 5KHz because there is no cable capacitance. Of course you provide the capacitance you want across the pickup by switching or other method. The 25K pots have a very wide useful dynamic range, allowing one pickup to be set near or at clean and the other at very high gain (if you have separate volume controls), but you need...
2. The volume pedal. This enables the post-preamp distortion adjustment required to keep the volume within the right range when you change volume levels (perhaps by changing the pickup) on the guitar. Cable length can be conveniently long because the impedance is fairly low, and interference is not a problem because the level is high. (This is not a standard effects loop: the level is too high.)
3. High gain. Well your amp really ought to be able to give you feedback to the strings, even at reasonable volume levels and with no external pedals. That takes gain. And also you need a little extra gain for the medium impedance pickups.
4. Blocking distortion. Can be a problem with very high gain. Let’s kill this by design.
5. Tone stack. This is what I like, but the flexibility to use other tone stacks is there.
The schematic is shown here: (
).
Edit: pdf64 points out that there is a missing blocking cap on this schematic before the pre amp out. I will fix that as soon as I can.
Edit: Error fixed
1. The input stage: Eliminating the grid stop resistor can be achieved by using one or the other of two available two tube circuits. What is normally called a cascode (common cathode into common grid) has the highest possible gain and output impedance. That combination can be used for very high output impedance and gain, not what we wat in an amp with many stages. Let’s keep the gain well distributed, and so the other configuration is better: a cathode coupled pair, or elementary differential amplifier. The symmetry is broken here by ac grounding the left hand plate and grounding the right hand grid. Notice that every stage in the preamp uses the negative supply, but not all for the same reason. In this stage it allows the very high value of cathode resistor, assuring nearly perfect coupling between the two tubes, and it allows the grids to operate at true ground potential, simplifying the circuit. In either of the two configurations, you have the noise of two tubes, not what you want, but still a lot better than the grid stops. In principle a lower noise triode could be used if necessary, but that would require more current from the supply, and this seems good enough.
2. Second stage: Since this stage can be driven hard by the input stage through the first gain control, it uses long tail biasing to the negative supply (as do the two following stages). The dc gain, grid to plate, is less than unity, and so the operating point cannot shift much with a large dc input rectified by the cathode-grid “diode” when there is a large ac signal present. The downside of this is that an expensive high quality bypass capacitor is required. But a guitar amp can afford to cut some bass, and so a huge value is not required.
3. Third Stage: This stage receives its input from the second gain control and drives the tone stack. One section of a 12AT7 is used in common cathode mode (normal stage of gain). The output impedance is low enough to do a reasonable job of keeping the tone response accurate without the use of a cathode follower, and the approximately 3.5 ma plate current gives a stronger drive than the usual 12AX7 cathode follower. The cathode follower is not used because we do not have enough tube sections, and also, at least in my opinion, its internal feedback gives a sharper distortion in the “touch region” just on the non linear side of clean. (Of course it is nowhere as bad as an emitter follower).
4. Fourth Stage: The fourth stage drives the (25K) volume pedal when used; otherwise it drives a 27K resistor so that the gain remains nearly constant when a pedal is connected in full on position. The other section of the 12AT7 is used for this stage with 3.5 ma of plate current allowing a substantial signal to drive the pedal for good immunity to interference pickup and sufficient level so that the following phase splitter can overdrive the output tubes reasonably well. (Only the gain of the EL84s comes afterwards since the phase splitter has approximately unity gain.) This stage has more gain than necessary, and some attenuation has been inserted between the stack and its input. In fact, if this attenuator is omitted, the amp is not stable with all inputs, particularly not with a high Q single coil pickup with the treble up all the way. The attenuator makes the amp stable under all conditions I have tried (maybe not exhaustive). Since this attenuator comes right after the stack, you could use a different stack design with more inherent loss and perhaps other good properties as a result, and then adjust the 5. attenuator as required to maintain the right gain.
5. Phase splitter: A cathodyne is used; the negative supply increases the voltage output capability before overload, and allows the grid to bias the tube while at dc ground. The usual suggestions for avoiding nasty distortion have been followed, although the level of overdrive that can be achieved is not extremely high because of loss in the tone stack and the following attenuator.
Why a cathodyne splitter? The final gain control (“master” volume) should be after as much gain as possible, and the pedal is best located in series before it. If a splitter with gain is used, then the master might best be located after it (as suggested in a mod to the original kit). This is not so practical with an external pedal, however, and so the single tube splitter that has approximately unity gain has been used, and the gain from the “free” tube section is used earlier (the second stage).
Noise measurement: The noise measurement used a 10K metal film resistor as a ”cal”, and measurements of noise alone (input shorted) and noise plus cal (10K resistor plugged in) were alternated, combined, and used to find the noise resistance. A sequence of measurements was used in order to cancel out gain drifts between the two modes of measurement. The measurements were made with an eight ohm resistor connected to one speaker jack, and a recording interface connected to the other. The recording interface fed spectral analysis software. A very clean region of the spectrum at that time (one of many possibilities, of course) was chosen (2.666 KHz). All three gain controls were set to maximum. The measured noise equivalent resistance was 3442 ohms; this is slightly larger than the ideal (as it should be, but it is satisfactorily close) of 3125 ohms for two tubes with 1600 micromhos transconductance each.
Effect of long tail biasing on overload characteristics: The input stage has a clean output level of about 130 volts p2p. It was set to this level with a 1KHz sine wave, and the first gain control was adjusted to explore the effect of a large signal on the dc current through the second stage. This was done by monitoring the cathode voltage on the second stage: it was about +1.65 volts with no input, stays close to constant as the input is increased with the first gain control, and decreases to about -18 volts with the full 130 volts p2p input. This is a change in current of about .06 ma (60 microamps), or about 5.5%. Thus it seems that the long tail biasing is holding the operating point close to nominal as it is intended to.
When the same type of measurement is performed for the third stage, the first 12AT7 section, using the second gain control, the current changes by a similar percentage, but the change is in the positive, rather than negative direction. Anyone know why?
Power supply: Filament voltage for the preamp tubes is rectified and well filtered, using both available 6.3 volt windings in series to get a 12.6 V supply, reducing the relative loss from diode drops. The diode bridge is over kill, but I had it. A series resistor is used before the filter caps in order to get close to the correct voltage and reduce the peak diode switching transients somewhat.
The high voltage filter caps are larger than necessary, but at this point I have little motivation for figuring out which ones could be reduced by how much.
I have not connected the heater supply to dc ground, but just to ac ground with a capacitor. During normal operation it stays near ground because that is where the cathodes are. During turn on the preamp tube cathodes are at the negative supply until the heaters warm up. This exceeds the heater-cathode voltage spec for a short time, but no problems happen, and I have had this feature in another amp for years. The problem could be avoided with a double throw standby switch to switch both the positive and negative supplies on after the cathodes start releasing electrons.
Layout, wiring, etc.: One problem with this design is that the preamp tube sections cannot all run in ascending order of level because the phase splitter is a 12AX7 section. The solution used here is to put the 12AT7 second and put a shielding wall between it and the input stage tube. I have used pick guard plastic covered with conducting tape.
The dominant source of potential oscillations is feedback to the grid of the input stage. To minimize this, two conductor shielded cable is used between the jack and the tube. Both high and low side of the input signal run directly to the appropriate grids with the 1 M resistor connected directly between them. The shield is connected to the chassis near the tube. The jack is wrapped in insulating tape, and then conducting tape, which is connected to the shield.
The big G10 card has been used for the filter caps and less critical components. The filter caps for a particular stage are located near that stage in most cases, keeping the power supply connections to the stages short.
Three small component cards were used. One is for the tone stack components and the following attenuator. It is mounted near the treble potentiometer. Thus it is easy to change the tone stack with little impact on other wiring. A card is also used for the cathodyne components. It is located between the cathodyne stage and the output tubes. A third card is used for components associated with the fourth stage. It is mounted near the tube on the vertical side of the chassis.
Summary of required characteristics:
1. Low noise front end without loss of stability
2. Able to drive a medium impedance (25K) volume pedal located iust before the last gain control (master volume) with no compromise in performance.
3. Single channel, high gain but flexible enough to adjust to a perfectly clean signal (total of three gain controls)
4. Blocking distortion free (no cut off of preamp stages with any gain setting)
5. Fender/Marshal type of bass-mid-treble tone stack.
Why these characteristics?
1. Low noise: The reason is primarily to accommodate non standard guitar outputs. First is the guitar using medium impedance pickups, where the voltage might be three times lower than, say, a typical single coil pickup (impedance 10 times or more lower). This is a passive guitar with 25K or 10K pots, often with a switched capacitor/resistor network for varying the resonant frequency that is set by pickup inductance/cable capacitance in a standard guitar. The tone is independent of the volume setting; because of the lower impedance; it is possible to use the volume control in ways you cannot in a standard guitar. Second is the wide dynamic range active guitar. This guitar buffers the pickups with an amplifier that can be as simple as a FET source follower. You can use very high output pickups and still get resonances at or above 5KHz because there is no cable capacitance. Of course you provide the capacitance you want across the pickup by switching or other method. The 25K pots have a very wide useful dynamic range, allowing one pickup to be set near or at clean and the other at very high gain (if you have separate volume controls), but you need...
2. The volume pedal. This enables the post-preamp distortion adjustment required to keep the volume within the right range when you change volume levels (perhaps by changing the pickup) on the guitar. Cable length can be conveniently long because the impedance is fairly low, and interference is not a problem because the level is high. (This is not a standard effects loop: the level is too high.)
3. High gain. Well your amp really ought to be able to give you feedback to the strings, even at reasonable volume levels and with no external pedals. That takes gain. And also you need a little extra gain for the medium impedance pickups.
4. Blocking distortion. Can be a problem with very high gain. Let’s kill this by design.
5. Tone stack. This is what I like, but the flexibility to use other tone stacks is there.
The schematic is shown here: (
Edit: pdf64 points out that there is a missing blocking cap on this schematic before the pre amp out. I will fix that as soon as I can.
Edit: Error fixed
1. The input stage: Eliminating the grid stop resistor can be achieved by using one or the other of two available two tube circuits. What is normally called a cascode (common cathode into common grid) has the highest possible gain and output impedance. That combination can be used for very high output impedance and gain, not what we wat in an amp with many stages. Let’s keep the gain well distributed, and so the other configuration is better: a cathode coupled pair, or elementary differential amplifier. The symmetry is broken here by ac grounding the left hand plate and grounding the right hand grid. Notice that every stage in the preamp uses the negative supply, but not all for the same reason. In this stage it allows the very high value of cathode resistor, assuring nearly perfect coupling between the two tubes, and it allows the grids to operate at true ground potential, simplifying the circuit. In either of the two configurations, you have the noise of two tubes, not what you want, but still a lot better than the grid stops. In principle a lower noise triode could be used if necessary, but that would require more current from the supply, and this seems good enough.
2. Second stage: Since this stage can be driven hard by the input stage through the first gain control, it uses long tail biasing to the negative supply (as do the two following stages). The dc gain, grid to plate, is less than unity, and so the operating point cannot shift much with a large dc input rectified by the cathode-grid “diode” when there is a large ac signal present. The downside of this is that an expensive high quality bypass capacitor is required. But a guitar amp can afford to cut some bass, and so a huge value is not required.
3. Third Stage: This stage receives its input from the second gain control and drives the tone stack. One section of a 12AT7 is used in common cathode mode (normal stage of gain). The output impedance is low enough to do a reasonable job of keeping the tone response accurate without the use of a cathode follower, and the approximately 3.5 ma plate current gives a stronger drive than the usual 12AX7 cathode follower. The cathode follower is not used because we do not have enough tube sections, and also, at least in my opinion, its internal feedback gives a sharper distortion in the “touch region” just on the non linear side of clean. (Of course it is nowhere as bad as an emitter follower).
4. Fourth Stage: The fourth stage drives the (25K) volume pedal when used; otherwise it drives a 27K resistor so that the gain remains nearly constant when a pedal is connected in full on position. The other section of the 12AT7 is used for this stage with 3.5 ma of plate current allowing a substantial signal to drive the pedal for good immunity to interference pickup and sufficient level so that the following phase splitter can overdrive the output tubes reasonably well. (Only the gain of the EL84s comes afterwards since the phase splitter has approximately unity gain.) This stage has more gain than necessary, and some attenuation has been inserted between the stack and its input. In fact, if this attenuator is omitted, the amp is not stable with all inputs, particularly not with a high Q single coil pickup with the treble up all the way. The attenuator makes the amp stable under all conditions I have tried (maybe not exhaustive). Since this attenuator comes right after the stack, you could use a different stack design with more inherent loss and perhaps other good properties as a result, and then adjust the 5. attenuator as required to maintain the right gain.
5. Phase splitter: A cathodyne is used; the negative supply increases the voltage output capability before overload, and allows the grid to bias the tube while at dc ground. The usual suggestions for avoiding nasty distortion have been followed, although the level of overdrive that can be achieved is not extremely high because of loss in the tone stack and the following attenuator.
Why a cathodyne splitter? The final gain control (“master” volume) should be after as much gain as possible, and the pedal is best located in series before it. If a splitter with gain is used, then the master might best be located after it (as suggested in a mod to the original kit). This is not so practical with an external pedal, however, and so the single tube splitter that has approximately unity gain has been used, and the gain from the “free” tube section is used earlier (the second stage).
Noise measurement: The noise measurement used a 10K metal film resistor as a ”cal”, and measurements of noise alone (input shorted) and noise plus cal (10K resistor plugged in) were alternated, combined, and used to find the noise resistance. A sequence of measurements was used in order to cancel out gain drifts between the two modes of measurement. The measurements were made with an eight ohm resistor connected to one speaker jack, and a recording interface connected to the other. The recording interface fed spectral analysis software. A very clean region of the spectrum at that time (one of many possibilities, of course) was chosen (2.666 KHz). All three gain controls were set to maximum. The measured noise equivalent resistance was 3442 ohms; this is slightly larger than the ideal (as it should be, but it is satisfactorily close) of 3125 ohms for two tubes with 1600 micromhos transconductance each.
Effect of long tail biasing on overload characteristics: The input stage has a clean output level of about 130 volts p2p. It was set to this level with a 1KHz sine wave, and the first gain control was adjusted to explore the effect of a large signal on the dc current through the second stage. This was done by monitoring the cathode voltage on the second stage: it was about +1.65 volts with no input, stays close to constant as the input is increased with the first gain control, and decreases to about -18 volts with the full 130 volts p2p input. This is a change in current of about .06 ma (60 microamps), or about 5.5%. Thus it seems that the long tail biasing is holding the operating point close to nominal as it is intended to.
When the same type of measurement is performed for the third stage, the first 12AT7 section, using the second gain control, the current changes by a similar percentage, but the change is in the positive, rather than negative direction. Anyone know why?
Power supply: Filament voltage for the preamp tubes is rectified and well filtered, using both available 6.3 volt windings in series to get a 12.6 V supply, reducing the relative loss from diode drops. The diode bridge is over kill, but I had it. A series resistor is used before the filter caps in order to get close to the correct voltage and reduce the peak diode switching transients somewhat.
The high voltage filter caps are larger than necessary, but at this point I have little motivation for figuring out which ones could be reduced by how much.
I have not connected the heater supply to dc ground, but just to ac ground with a capacitor. During normal operation it stays near ground because that is where the cathodes are. During turn on the preamp tube cathodes are at the negative supply until the heaters warm up. This exceeds the heater-cathode voltage spec for a short time, but no problems happen, and I have had this feature in another amp for years. The problem could be avoided with a double throw standby switch to switch both the positive and negative supplies on after the cathodes start releasing electrons.
Layout, wiring, etc.: One problem with this design is that the preamp tube sections cannot all run in ascending order of level because the phase splitter is a 12AX7 section. The solution used here is to put the 12AT7 second and put a shielding wall between it and the input stage tube. I have used pick guard plastic covered with conducting tape.
The dominant source of potential oscillations is feedback to the grid of the input stage. To minimize this, two conductor shielded cable is used between the jack and the tube. Both high and low side of the input signal run directly to the appropriate grids with the 1 M resistor connected directly between them. The shield is connected to the chassis near the tube. The jack is wrapped in insulating tape, and then conducting tape, which is connected to the shield.
The big G10 card has been used for the filter caps and less critical components. The filter caps for a particular stage are located near that stage in most cases, keeping the power supply connections to the stages short.
Three small component cards were used. One is for the tone stack components and the following attenuator. It is mounted near the treble potentiometer. Thus it is easy to change the tone stack with little impact on other wiring. A card is also used for the cathodyne components. It is located between the cathodyne stage and the output tubes. A third card is used for components associated with the fourth stage. It is mounted near the tube on the vertical side of the chassis.
Attached Files
I think you would need to have an RC lpf in the grid to make it effective and economic. You could of course do the same thing with the original stage and still have lower noise.
Comment