Yes. You can't feed the screen regulators from the plates. The plate voltage can drop below the voltage needed to keep the gas regulator lit, so it will extinguish creating a step drop in the screen voltage. This will create a weird distortion that may not be musically useful. Connect the TBD resistors to a B+ supply.
It might be cliche, but now I'm picturing Back to the Future. This may be one of my favorite amp stories of all time. The plexi safety screen cracked me up.
Once I really get some experience with this stuff, one idea that really intrigues this dilettante is building a guitar amp around the GU-81M. I don't really want to mess around with a 2kV power supply, though - and it's not like I could do anything with 700W output- so I need to research how it behaves when you starve the plate down to, say, 500V. Orrrrrr, maybe ... it looks like the screen can dissipate 120W, so maybe use that for the output?
Gotcha, makes sense. I guess I was figuring that the voltage wouldn't swing below 150V, but that just goes to show how much I've got left to learn.
In a shunt configuration, should I worry about feeding them voltage close to their operating point (assuming it's enough to light them)? Or can I tap off something higher?
Updated. Bumped up the screens to 258V so I can get two different colors!
Maybe three, if I bias the 815 hot enough ... 😎 ��
Oh, and I decided to make a real VR tube building block for DIYLC. Surprised there's not one out there already. Or maybe not surprised that I wasn't smart enough to find one.
Maybe three, if I bias the 815 hot enough ... 😎 ��
Oh, and I decided to make a real VR tube building block for DIYLC. Surprised there's not one out there already. Or maybe not surprised that I wasn't smart enough to find one.
Okay, coming back here after working on solving the 3B28 plate delay challenge on the other thread. For the record, at this time I'm planning on using an ICM102 to delay energizing the main PT primary, while the 3B28s' filament trafo is connected directly to the switch. Here's the current state of the power supply:
Two things to watch out for:
You cannot put a capacitor in series with the regulator tubes in most cases. You get a relaxation oscillator instead a regulated voltage.
You may get a bit of noise with the 3b28 with a capacitor input supply. The 3b28 has to reach 50 anode volts before it starts to conduct and will snap down to the 14 volt drop each cycle causing a current spike. Choke input supplies avoid this. Possibly a little bit of series resistance between cathode and the first cap will help.
You cannot put a capacitor in series with the regulator tubes in most cases. You get a relaxation oscillator instead a regulated voltage.
You may get a bit of noise with the 3b28 with a capacitor input supply. The 3b28 has to reach 50 anode volts before it starts to conduct and will snap down to the 14 volt drop each cycle causing a current spike. Choke input supplies avoid this. Possibly a little bit of series resistance between cathode and the first cap will help.
My next challenge: Bias. My original plan was to use a single fixed bias supply, but then I realized that I don't particularly care about squeezing every last watt out of this, so I started thinking about cathode bias.
Then I saw this note in the 815 datasheet:
I'm planning on having the screens at around 200V. With the shared cathode, that means I'm back to fixed bias, now with individual adjustments per side. No bias tap on the PT. I see three ways to get the supply:
I'm leaning toward back-biasing, using Aiken's "improved zener" circuit. Low component count, regulated bias to go with the regulated screens, and it just seems clean to me.
Any advice?
Then I saw this note in the 815 datasheet:
I'm planning on having the screens at around 200V. With the shared cathode, that means I'm back to fixed bias, now with individual adjustments per side. No bias tap on the PT. I see three ways to get the supply:
- Voltage multiplier on the unused 5V winding on the PT
- Tap the HV, rectify, and divide down to what I need
- Back-bias, which is what the original circuit used
I'm leaning toward back-biasing, using Aiken's "improved zener" circuit. Low component count, regulated bias to go with the regulated screens, and it just seems clean to me.
Any advice?
Good to know! The only cap near that is the filter cap for the 300V node, and that's on the other side of a series resistor that I added after I posted the last schematic. Is the filter cap something that would cause an issue?
Interesting, hadn't considered that. I'm not sure if PSUD2 is modeling for that, but here's a closeup of the voltage at the first cap (the sawtooth), and the voltage at the transformer. I definitely see those big spikes at the trafo, but it doesn't seem to be reflected in the waveform at the cap. And I don't know if that reflects what you're talking about, because I don't see a matching spike at the tube.
Here's an extreme closeup of the modeled current at the trafo:
Adding a series resistor smooths out those spikes (and moves them down from the peak of the waveform), at the cost of dropping the voltage, by almost 50V with 100R. But it doesn't seem to change the shape of the waveform at the cap.
I've got the tubes and sockets, so I can just buy the caps and wire it up with a dummy load, hook it up to the scope, and see. And if the caps don't do the trick, I can redesign as necessary.
Thanks for the insight!
Back-biasing
Using Aiken's description of back-biasing and messing around with learning LTSpice, I've come up with a circuit that I think will work. My requirements are adjustability around a nominal -15V, and individual balance per tube.
I built a super simplified model in LTSpice that represents the amp as a single 3.7k resistor to represent 123mA current @ 460V, and then tweaked the values until I got to a range of -12V to -18.5V at the wiper of the balance pot. As a sanity check, I also made sure that B+ dropped by the amount of the zener.
The simulation shows about 118mA flowing through the zener. I don't know enough about zeners to know which specification I need to make sure can handle that current, but I'll figure that out.
I know it's unconventional, but it's the simplest bias circuit available to me, with the added benefit of regulation. So ... what have I messed up here, either theoretically or in the implementation? Aiken says to connect only the main filter cap, but would it make sense to connect all of them?
One last thing - I know that the 50R resistors in the simulation don't do anytyhing, I just put them there for the visual.
Using Aiken's description of back-biasing and messing around with learning LTSpice, I've come up with a circuit that I think will work. My requirements are adjustability around a nominal -15V, and individual balance per tube.
I built a super simplified model in LTSpice that represents the amp as a single 3.7k resistor to represent 123mA current @ 460V, and then tweaked the values until I got to a range of -12V to -18.5V at the wiper of the balance pot. As a sanity check, I also made sure that B+ dropped by the amount of the zener.
The simulation shows about 118mA flowing through the zener. I don't know enough about zeners to know which specification I need to make sure can handle that current, but I'll figure that out.
I know it's unconventional, but it's the simplest bias circuit available to me, with the added benefit of regulation. So ... what have I messed up here, either theoretically or in the implementation? Aiken says to connect only the main filter cap, but would it make sense to connect all of them?
One last thing - I know that the 50R resistors in the simulation don't do anytyhing, I just put them there for the visual.
Last edited:
I have done this myself and it works. There are some points of caution. When a Zener blows, it almost always shorts out. This will remove your bias and make your output tube very angry.....red faced kind of angry. So it's best to use something way oversized in the Zener department.
The Zener diode carries all the current drawn by the entire amplifier. Simulations and data sheets might show a current of 123 mA. These simulation do not include Mr. Metalhead "Setting the Controls for the Heart of the Sun" on his pedal board and thrashing away with the master volume on 11. There is no amplifier class for this.....C- maybe, but expect peak currents at least twice what you expect, as you left the simulation world of class AB when the amp crossed into clipping.
Putting a fat cap across the Zener will blunt the effects of typical guitar amp mayhem, but it will also delay the bias voltage rise, which depends on tube current to develop anyway. I would try something in the 100 uF range.
I am including the spec for the 5 watt Zeners that I use in my tube amps. The right most column (Izm) is the maximum current spec. Vz-nom is the nominal voltage for a given diode. The 1N5357B is a 20 volt diode with a 237 mA max current spec. This assumes some heat sinking through the leads usually into a PC board. It is best to run the Zener at half this current or less.
Zener diodes do NOT like to be operated in parallel since they never match, and one will eat most of the current. Wiring them in series is OK, and a trick that I often use. I have made series Zener strings with taps for several different voltages each buffered with a mosfet for screen and other voltage regulators in an amp.
In this case I would use two 1N5347A 10 volt Zeners in series giving you a maximum current spec of 475 mA.
The Zener diode carries all the current drawn by the entire amplifier. Simulations and data sheets might show a current of 123 mA. These simulation do not include Mr. Metalhead "Setting the Controls for the Heart of the Sun" on his pedal board and thrashing away with the master volume on 11. There is no amplifier class for this.....C- maybe, but expect peak currents at least twice what you expect, as you left the simulation world of class AB when the amp crossed into clipping.
Putting a fat cap across the Zener will blunt the effects of typical guitar amp mayhem, but it will also delay the bias voltage rise, which depends on tube current to develop anyway. I would try something in the 100 uF range.
I am including the spec for the 5 watt Zeners that I use in my tube amps. The right most column (Izm) is the maximum current spec. Vz-nom is the nominal voltage for a given diode. The 1N5357B is a 20 volt diode with a 237 mA max current spec. This assumes some heat sinking through the leads usually into a PC board. It is best to run the Zener at half this current or less.
Zener diodes do NOT like to be operated in parallel since they never match, and one will eat most of the current. Wiring them in series is OK, and a trick that I often use. I have made series Zener strings with taps for several different voltages each buffered with a mosfet for screen and other voltage regulators in an amp.
In this case I would use two 1N5347A 10 volt Zeners in series giving you a maximum current spec of 475 mA.
How much power are you expecting from this? Also what sort of output transformers do you have?
I see the tube can do 40-50 watts push pull but that is class AB2.
I see the tube can do 40-50 watts push pull but that is class AB2.
Yup, definitely understand that.
I don't see anybody but me ever playing this guy, and I'm more of a roots-rock and jam band guy, but we're talking about pennies in parts, so definitely worth it. Who knows, maybe I'll become a metalhead when I hit 50 in a couple years?
The cap would only need to be specified for the Zener drop plus margin, right? So a 50V cap would work?
So that's an interesting point. I'm going to reuse the original turret boards for the amp circuit, but I'm planning on building the power supply on something like veroboard, both for stability and so it's easier to use radial caps. I can lay that out so that the Zeners keep their full leads.
Thanks so much for the recommendation! I realized that my target operating point is actually at -18V, so I need to make some adjustments, but that datasheet will be invaluable.
I'll admit that I'm still learning on this stuff, but based on the operating point, it should be around 45W. The OT is from the donor Hammond AO-14, which uses two 6V6s. So it's the right impedance (8K P-P), but that's definitely pushing it hard.
On the other hand, I'll be using it at home and in the studio, without an attenuator, so I don't expect I'll ever be asking it for more than 20W once it's off the bench. Regardless, I'll definitely keep an eye on the OT temp.
The 815 spec sheet I have shows a push pull amplifier with 500 volts on the plate with 125 volts on the screen. Load is 8K plate to plate resulting in 54 watts out.
This is class AB2 though.
Grid is biased at -15 volts.Peak grid to grid volts is 60 meaning the grid has to be driven +15 volts peak.
You may get reasonable class AB1 power with 225 volts on the screen but not 54 watts.
This is class AB2 though.
Grid is biased at -15 volts.Peak grid to grid volts is 60 meaning the grid has to be driven +15 volts peak.
You may get reasonable class AB1 power with 225 volts on the screen but not 54 watts.
Last edited:
I failed to grasp that the output transformer was meant for a pair of 6v6s and you were playing with the 20 watt range.
I would probably go for about 200 volts on the screens and 450 on the plate, No need for AB2. I doubt you would have bias problems so you can have shared cathode bias about -25 volts.
May be best to trade the 3b28s for a 5u4 rectifier. Plate voltage drops about right and you avoid some noise problems.
I would probably go for about 200 volts on the screens and 450 on the plate, No need for AB2. I doubt you would have bias problems so you can have shared cathode bias about -25 volts.
May be best to trade the 3b28s for a 5u4 rectifier. Plate voltage drops about right and you avoid some noise problems.
That's pretty much where I'm at ... 200V on the screen, and around 450 on the plates, a touch higher or lower depending on how I handle biasing. At -25V bias, though, that puts dissipation at 16.6W, which seems a little cold (66% max). All of this is still in the realm of theory, though, since I haven't even started to build.
I recognize that the comment about individually biasing each side when using over 125V on the screens is directed at use cases that are a little more demanding than a guitar amp. So maybe it's overkill to implement a fixed bias supply. Haven't made a decision on that yet.
The donor amp used a 5U4. But as silly as it sounds, this build is very much about the looks, and the 3B28 is the right look. If it causes noise issues, I'll work through that.
A 50 ohm 10 watt resistor between the 3b38s and the first capacitor may enough to kill your noise problem with the 3b28.
You may need a 10K stopper for the 815 grids. Also try to get keep the plate wiring as far away from the grids and driving stage as you can. Also you probably want a bypass cap on the screen going back the the first cap negative and not through any other ground wiring.
Should work fine if you do that.
You may need a 10K stopper for the 815 grids. Also try to get keep the plate wiring as far away from the grids and driving stage as you can. Also you probably want a bypass cap on the screen going back the the first cap negative and not through any other ground wiring.
Should work fine if you do that.
Started work on the preamp topology. No values yet, just figuring out what components need to go where. Notes:
Would appreciate any and all feedback on the draft. I'm going to start slowly filling in component values this week.
- Topology is: EF37 input stage -> Gain control -> 1/2 6C8G cold clipper -> attenuator -> HY615-based cascode -> MOSFET source follower -> tone stack -> 1/2 6C8G recovery -> Pre-PI MV -> FX loop -> PI -> Post-PI MV -> cathode-biased 815 in push-pull
- Yup, sticking with cathode bias. So much easier than any other biasing method, and I'm not sure it's going to matter - the amp should be nice and firm already, what with a regulated screen supply, big ol' reservoir caps, and xenon rectos.
- I stuck a Fender BF tone stack in for now. Might keep it, might not. Between the source follower feeding it and a healthy recovery stage, I can do pretty much whatever I want.
- Yes, both Pre- and Post-PI MV. Might want different flavors, and I don't know how that 6J6 is going to sound when overdriven.
- Multi-star grounding scheme. Just a spitball for now. Trying to keep everything that's fed from one power node grounded together, except that the power amp uses two different voltages and I grounded it together.
- Had the wrong lower VR tube - the 0B3 is an octal, and I'm completely out of places to stick a big socket without cutting any new holes. So screen voltage is going to drop to 180 from the planned 200, because I want two different colors of glow tube.
- Don't know what I'm going to do for the loop. Maybe I'll buy a kit? I find the loop one of the least interesting parts, but I do really want to have one.
Would appreciate any and all feedback on the draft. I'm going to start slowly filling in component values this week.