One of the things I noticed when searching the web for info about digitally controlling tube amps, is that there's a few people who post claiming to have done something, but nobody gets into the details (with the exception of the people who've read the patents on the tri-axis and explain how that works). So I'd like to keep that information flowing.
Anyhow, there are flaws with LDR's, which is why I suppose the major manufacturers don't use them - but this is a pet project, so I'm free to experiment and find out for myself how bad it really is.
So, to explain further my "findings" (which are probably obvious to anybody once they've actually begun working with circuits like this):
a. My table of voltages above is possibly misleading. The important part is the current over the 22k resistor I show in the hand-drawn schematic in the first post. The voltage drop across the resistor is from the value at the DAC pin (which is what is listed in the table) to about 0.65 (the voltage drop across the transistor). The transistor should pull down about 100 times that current through the LED.
b. The CdS cell I'm using is a GL5537. The data sheet I've got indicates there are two variants, a GL5537-1 and GL5537-2, but the supplier I purchased them from just labeled them GL5537. So I don't know which variant they are. I've got some GL5549 on order, I'm hoping they'll give me finer control in the 100k-500k range.
c. The LED's are super-bright green LED's. The 20 mA one is mouser part #
604-WP5603ZGDL/SD/G and the 30 mA one is mouser part #
941-C503BGANCB0F0791
d. The MAX529 impedance problems I had was because I wasn't using buffered mode. Now that I'm in buffered mode, the voltage dividers I use for REFH and REFL pins are working. The max output voltage is around 2.75v in buffered mode (I managed to get 3v unbuffered), so I'll have to lower the 22k resistor to up the current correspondingly.
e. That schematic in the first post should've had +5 connected to the diode on the REFL pin through a 10k resistor, to provide .65 volts. Doesn't really matter, since .65 volts is lower than the DAC likes to put out. But I hate not mentioning that error... also the CS pin on the DAC needs to be connected to the microcontroller, but that should be obvious
f. I'm using an ATmega1284p microcontroller. The worst part of setting that up was the external crystal oscillator - it took me a long time before I found a formula for determining the load capacitors to put on the crystal. The data sheet says to use 12-22pf for each capacitor, which means you have to be picky in which crystal you choose. Each capacitor needs to equal 2*(crystal load capacitance - stray capacitance), because they end up being in series across the crystal. I took a guess at only 2pf stray capacitance (very optimistic of me), used a 20Mhz 10pf crystal, which means each capacitor on the crystal should be (2*(10-2)) = 16pf, which is in the 12-22pf range. Not a standard size, but a pair of 15pf crystals worked, so there it is. Crappy internet advice just says "use a 22pf cap", but it really depends on the crystal's load capacitance. My first crystal had a 20pf load capacitance, which meant it really wanted about 36pf caps, which is too much capacitance for the chip per the data sheet. Anyhow enough complaining about that...
g. I'm still having problems getting the ATmega1284p ADC reading voltages properly in my test rig, so I've been using a multimeter to read the resistances of the CdS cells. Not sure why it reads voltages so low, maybe another impedance issue... a "real" external ADC might work better, but I've got no room left on the protoboards for more chips. Not a vital issue, but I do need to figure out how to reliably read voltages if I'm going to have circuitry to auto-calibrate the LDR's.
weirdo tubes amp
Moderators: pompeiisneaks, Colossal
Re: weirdo tubes amp
I have hardly the technical knowledge to follow what you are doing here but very very interested in your experience. Something along the lines of your project has long been my dream so I wish you great success!!!
My only and very humble suggestion... Why not use an off the shelf controller like an Arduino or PI board? This way at least you can skip the troubles with that part of the project and also utilize common interface elements along with programming assistance etc. Just a thought.
All the best.
My only and very humble suggestion... Why not use an off the shelf controller like an Arduino or PI board? This way at least you can skip the troubles with that part of the project and also utilize common interface elements along with programming assistance etc. Just a thought.
All the best.
Re: weirdo tubes amp
I don't like the Arduino for a few reasons:
a. USB firmware upload uses a UART. I want my UARTs for MIDI, one RX for midi in, one TX for midi out to custom footswitch, one TX for midi thru. I can share them w/ the USB interface but that seems like it's going to be error prone
b. I want to use JTAG debugging
c. I need a lot of pins to drive the UTFT display
d. The Arduino libraries aren't a great fit for what I'm doing, so if I'm writing my own I2C and SPI drivers anyhow, there's not a lot of use left for Arduino
That's why I'm going straight to an ATmega1284p. The only issue with getting that set up was the crystal & its capacitors, and the $100 I had to spend on a JTAGICE3 programmer to get my JTAG debugging.
I don't know the Raspberry PI platform, so I can't say much about it.
a. USB firmware upload uses a UART. I want my UARTs for MIDI, one RX for midi in, one TX for midi out to custom footswitch, one TX for midi thru. I can share them w/ the USB interface but that seems like it's going to be error prone
b. I want to use JTAG debugging
c. I need a lot of pins to drive the UTFT display
d. The Arduino libraries aren't a great fit for what I'm doing, so if I'm writing my own I2C and SPI drivers anyhow, there's not a lot of use left for Arduino
That's why I'm going straight to an ATmega1284p. The only issue with getting that set up was the crystal & its capacitors, and the $100 I had to spend on a JTAGICE3 programmer to get my JTAG debugging.
I don't know the Raspberry PI platform, so I can't say much about it.
Re: weirdo tubes amp
I put an op amp source follower between the ATMega1284P's ADC pin and the LDR measurement, so high resistances don't monkey with the ADC. And then fixed a bug in the measurement code that was probably 99% of the problem. So the prototype of the control circuits for the LDR's is done and functional.
The GL5549 LDR's showed up, so I tested those with the 30 mA LED's. They are better than the GL5537's, and I'm able to have finer grained control over the higher resistance values. Things are fairly linear from 6.1k to 76k, but then the last 20% of the DAC's output range is spent on a fairly exponential curve from 100k to 1M.
I think I can simulate a logarithmic pot fairly well, although overall impedance will change if I want a large impedance (> 100k), and can simulate linear pots up to 100k. The max logarithmic impedance I can see working would be around 200k to 250k.
I then through together a second one, and it had half the resistance of the first. So I'm going to have to hand-test each of these and bin them... bleah.
Does this match with anybody else's experience, if anyone here's monkeyed around with using LDR's as potentiometer replacements?
My next step is prototyping the digital tuner. I want a built-in tuner on my amp, just because. So I've got a daughter board with a dsPIC33E, which the ATmega1284p will communicate with through I2C, and the dsPIC33E is responsible for audio processing. It also allows me to add some digital effects if I add some external RAM (delay, flange) if I feel like it, although that wouldn't be very tube purist. So maybe not. More interestingly, I might be able to sample the signal at various points and have a built-in audio frequency oscilloscope, since I've got that full-color LCD display. Ambitious, but I think I'll put the hardware in and see if I ever get to programming it. The only bit I'll be seriously prototyping beforehand is the tuner though.
Once that prototype is done (waiting on a 10uf ceramic cap from Mouser before it's finished), I can design the PCB's for the digital stuff and the ridiculous power supply.
The GL5549 LDR's showed up, so I tested those with the 30 mA LED's. They are better than the GL5537's, and I'm able to have finer grained control over the higher resistance values. Things are fairly linear from 6.1k to 76k, but then the last 20% of the DAC's output range is spent on a fairly exponential curve from 100k to 1M.
I think I can simulate a logarithmic pot fairly well, although overall impedance will change if I want a large impedance (> 100k), and can simulate linear pots up to 100k. The max logarithmic impedance I can see working would be around 200k to 250k.
I then through together a second one, and it had half the resistance of the first. So I'm going to have to hand-test each of these and bin them... bleah.
Does this match with anybody else's experience, if anyone here's monkeyed around with using LDR's as potentiometer replacements?
My next step is prototyping the digital tuner. I want a built-in tuner on my amp, just because. So I've got a daughter board with a dsPIC33E, which the ATmega1284p will communicate with through I2C, and the dsPIC33E is responsible for audio processing. It also allows me to add some digital effects if I add some external RAM (delay, flange) if I feel like it, although that wouldn't be very tube purist. So maybe not. More interestingly, I might be able to sample the signal at various points and have a built-in audio frequency oscilloscope, since I've got that full-color LCD display. Ambitious, but I think I'll put the hardware in and see if I ever get to programming it. The only bit I'll be seriously prototyping beforehand is the tuner though.
Once that prototype is done (waiting on a 10uf ceramic cap from Mouser before it's finished), I can design the PCB's for the digital stuff and the ridiculous power supply.
Re: weirdo tubes amp
Well, since I'm blazing new ground with all the digital nonsense, I think I'll change the low voltage portion of the power supply to an SMPS. My main concern is heat dissipation, linear regulators are very inefficient so that's a lot of heat build up inside the chassis. Going SMPS, I won't have to actively cool the power supply, heatsinks should be enough.
Since I'm designing the SMPS myself, I'll put in robust LC pi filters to get rid of the 140khz hash. Should be noise free, the hifi guys sometimes use SMPS for their heater supplies, and they're pretty finicky.
For the high voltage, I don't see how SMPS would help much at all. It's an unregulated power supply, so it's pretty efficient already. So that stays old-school.
Since I'm designing the SMPS myself, I'll put in robust LC pi filters to get rid of the 140khz hash. Should be noise free, the hifi guys sometimes use SMPS for their heater supplies, and they're pretty finicky.
For the high voltage, I don't see how SMPS would help much at all. It's an unregulated power supply, so it's pretty efficient already. So that stays old-school.
Re: weirdo tubes amp
This thread:
https://tubeamparchive.com/viewtopic.ph ... t=photopot
Post #9 or so links to a PhotoPot circuit I came up with some time ago. It relies on an actual resistor (2 of them) to scale the pot value, a digital pot to move the wiper and produce the desired taper, 2 LDRs (this was the part that got difficult - matching the things) and a pair of op-amps to servo the mess.
You could - or if there was enough of a market for them, I could - manufacture dual-element matched LDRs using an LED, two CdS cells and something to make sure both cells are equally illuminated. Still need a jig to match/sort cell pairs, though.
Hope this helps!
https://tubeamparchive.com/viewtopic.ph ... t=photopot
Post #9 or so links to a PhotoPot circuit I came up with some time ago. It relies on an actual resistor (2 of them) to scale the pot value, a digital pot to move the wiper and produce the desired taper, 2 LDRs (this was the part that got difficult - matching the things) and a pair of op-amps to servo the mess.
You could - or if there was enough of a market for them, I could - manufacture dual-element matched LDRs using an LED, two CdS cells and something to make sure both cells are equally illuminated. Still need a jig to match/sort cell pairs, though.
Hope this helps!
Re: weirdo tubes amp
That would be this post I presume:
http://music-electronics-forum.com/t1015/#post8215
That's the Triaxis method with a digital pot instead, I think ?
I didn't want to do the opamp feedback thing, it seems messy. I plan to do some temperature tests to see if I can maintain a consistent voltage divider in different conditions, if I can't then I suppose feedback becomes a necessity.
Since I can digitally calibrate, I don't need to worry about linearity, I can just pick out different led currents to make the curve I want.
http://music-electronics-forum.com/t1015/#post8215
That's the Triaxis method with a digital pot instead, I think ?
I didn't want to do the opamp feedback thing, it seems messy. I plan to do some temperature tests to see if I can maintain a consistent voltage divider in different conditions, if I can't then I suppose feedback becomes a necessity.
Since I can digitally calibrate, I don't need to worry about linearity, I can just pick out different led currents to make the curve I want.
Re: weirdo tubes amp
Randall Smith might claim it infringes his patent, but not really the TriAxis method. He relied on hand-selected value pairs for each pot. This one uses identical circuits, scaled by two resistors (you need a pair of those setups to make a whole pot).
If you don't like op-amps, you'd need an ADC input per side and a tight loop to stay servoed. An op-amp is a lot simpler way to correct for thermal drift and wear.
It'd be worth following both threads, as discussion improved the sketch-based design quite a bit.
If you don't like op-amps, you'd need an ADC input per side and a tight loop to stay servoed. An op-amp is a lot simpler way to correct for thermal drift and wear.
It'd be worth following both threads, as discussion improved the sketch-based design quite a bit.
Re: weirdo tubes amp
So I did a highly controlled experiment, where I took two LDR's I had in my cold basement, set them to give a voltage divider with "wiper" around 27%, and then ran a heat gun on them briefly.
I had been hoping that they would change by relatively the same amount, so the voltage divider would be preserved. That was not the case, the "wiper" rose to 35% of the voltage across the LDR's. So either a feedback mechanism is necessary, or temperatures must be controlled in the chassis.
The only thing I don't like about the feedback mechanism is the assumption that two CdS cells are going to behave identically.
Maybe digital potentiometers are the way to go. I didn't want to go near them because the +/-15v AD7376 pots are only available in surface mount components. They also only have 128 steps.
Pairs of 10-resistor arrays with reed switches to short them out could also work, for pure tube tone and lots of clickity-clacks whenever you turn a knob. That just seems ridiculous though.
Could use JFET's to short resistors, but they've got small resistances when "on" and may distort too much. Possibly not more than a digital potentiometer though.
I'm going to sleep on this and see what inspiration strikes.
I had been hoping that they would change by relatively the same amount, so the voltage divider would be preserved. That was not the case, the "wiper" rose to 35% of the voltage across the LDR's. So either a feedback mechanism is necessary, or temperatures must be controlled in the chassis.
The only thing I don't like about the feedback mechanism is the assumption that two CdS cells are going to behave identically.
Maybe digital potentiometers are the way to go. I didn't want to go near them because the +/-15v AD7376 pots are only available in surface mount components. They also only have 128 steps.
Pairs of 10-resistor arrays with reed switches to short them out could also work, for pure tube tone and lots of clickity-clacks whenever you turn a knob. That just seems ridiculous though.
Could use JFET's to short resistors, but they've got small resistances when "on" and may distort too much. Possibly not more than a digital potentiometer though.
I'm going to sleep on this and see what inspiration strikes.
Re: weirdo tubes amp
Remote-cutoff pentodes as attenuators? One each for Treble, Middle and Bass, followed by a mixer stage, then another RC5 for volume?
No idea what the control voltage looks like. I suppose LDR-based control of the control voltage would work, but now the distortions of the resistive elements are out of the signal path.
No idea what the control voltage looks like. I suppose LDR-based control of the control voltage would work, but now the distortions of the resistive elements are out of the signal path.
Re: weirdo tubes amp
I was thinking along very similar lines this morning - tubes are voltage controlled resistors.
Remote cutoff tubes won't work, the signals have to be very small or else you get distortion, as the tube can't distinguish between a DC control voltage and a large signal swing.
But, a heptode would work very nicely - put a negative control voltage from a DAC onto the oscillator grid to attenuate the signal. A dual control pentode might work as well, by putting a control voltage onto the suppressor grid. [EDIT - I meant dual control, suppressor grids don't do much on regular pentodes or so I've read]
The gain and volume controls are trivial to implement in this manner. Other controls (treble/bass for the Baxandall tone stack, frequency controls for the modulator, compressor controls) may be harder to implement. I'll have to think some more about those.
Remote cutoff tubes won't work, the signals have to be very small or else you get distortion, as the tube can't distinguish between a DC control voltage and a large signal swing.
But, a heptode would work very nicely - put a negative control voltage from a DAC onto the oscillator grid to attenuate the signal. A dual control pentode might work as well, by putting a control voltage onto the suppressor grid. [EDIT - I meant dual control, suppressor grids don't do much on regular pentodes or so I've read]
The gain and volume controls are trivial to implement in this manner. Other controls (treble/bass for the Baxandall tone stack, frequency controls for the modulator, compressor controls) may be harder to implement. I'll have to think some more about those.
Re: weirdo tubes amp
I gave it some more thought.
The biggest problem I can see is with potentiometers/rheostats where one of the sides isn't at ground potential. There's plenty of solutions that can deal with one end grounded, but control voltages for everything except LDR's have to be referenced to one end of the variable resistor. Putting together floating voltages and optoisolators to allow the CPU to talk to whatever's generating the control voltage is just ridiculous.
So, that leaves LDR's and digital potentiometers (and equivalent constructions from individual resistors and switches) for "floating" potentiometers.
When one end is grounded, there are more options.
Gain, volume: I'm liking heptodes here - a tube solution for a tube amp, plus it keeps tubes in the signal path
EQ: Don's idea of using what amounts to a graphic EQ with a separate control tube per band is probably the way to go. Fortunately I've got a giant pile of heptodes, just have to see if I've got enough amperes in the high voltage section to power them all
Presence / depth / resonance: It's late and I don't want to think about these right now. Maybe just use relays to select from a few reasonable values.
Compressor threshold & intensity: could probably use a JFET here as a voltage controlled resistor (see http://graffiti.virgin.net/ljmayes.mal/comp/vcr.htm). It's in the side chain, so no worries about tone effects. Could also use a digital potentiometer, but a JFET keeps digital control wires out of the tube section of the amp - it just has to be fed a constant control voltage. Might be able to adapt the circuit to use triodes at low voltages though, which would be MORE TUBEY, and thus better.
Modulator (i.e. heptode-mixed tremolo): I wanted to build an Wien bridge pure sine wave LFO, because so much vacuum! Lots of little light bulbs to build up thermal mass to get stable low frequencies. The problem is there's a pair of rheostats in that circuit, that need to provide identical values, and one of them is floating. I'm not giving up on this yet though - I expect I need to rethink this circuit in terms of tubes and their varying resistances.
The biggest problem I can see is with potentiometers/rheostats where one of the sides isn't at ground potential. There's plenty of solutions that can deal with one end grounded, but control voltages for everything except LDR's have to be referenced to one end of the variable resistor. Putting together floating voltages and optoisolators to allow the CPU to talk to whatever's generating the control voltage is just ridiculous.
So, that leaves LDR's and digital potentiometers (and equivalent constructions from individual resistors and switches) for "floating" potentiometers.
When one end is grounded, there are more options.
Gain, volume: I'm liking heptodes here - a tube solution for a tube amp, plus it keeps tubes in the signal path
EQ: Don's idea of using what amounts to a graphic EQ with a separate control tube per band is probably the way to go. Fortunately I've got a giant pile of heptodes, just have to see if I've got enough amperes in the high voltage section to power them all
Presence / depth / resonance: It's late and I don't want to think about these right now. Maybe just use relays to select from a few reasonable values.
Compressor threshold & intensity: could probably use a JFET here as a voltage controlled resistor (see http://graffiti.virgin.net/ljmayes.mal/comp/vcr.htm). It's in the side chain, so no worries about tone effects. Could also use a digital potentiometer, but a JFET keeps digital control wires out of the tube section of the amp - it just has to be fed a constant control voltage. Might be able to adapt the circuit to use triodes at low voltages though, which would be MORE TUBEY, and thus better.
Modulator (i.e. heptode-mixed tremolo): I wanted to build an Wien bridge pure sine wave LFO, because so much vacuum! Lots of little light bulbs to build up thermal mass to get stable low frequencies. The problem is there's a pair of rheostats in that circuit, that need to provide identical values, and one of them is floating. I'm not giving up on this yet though - I expect I need to rethink this circuit in terms of tubes and their varying resistances.
Re: weirdo tubes amp
So here's how I'm thinking of building my voltage-controlled sine wave oscillator. Negative control voltages will all be generated using a microcontroller, DAC's and op amp inverting voltage buffers.
I do like the Wien bridge. An explanation - the R1/C2 and R2/C1 combinations make a low pass and high pass filter (or is that R1/C1 and R2/C2? doesn't really matter for the purposes of this discussion since they're all equal). The values of R1 and R2 should be equal, and C1 and C2 should be equal. This means that the frequency determined by 2*pi*R1*C1 is the least attenuated, and thus the frequency that the circuit will oscillate at. The circuit has a weakness - amplitude control. Too high and boom, square waves, too low and it peters out. It has to be just right. Normally this is done by using lamps as a poor man's thermistor, but they don't work so well at low frequencies as their temperature will start to vary. See the top half of the first hand-drawn crappy schematic below.
The strength of the Wien bridge is that it has great frequency control - if you can get good matches between R1 and R2, and C1 and C2, then that's your frequency. So if I can drive voltage-controlled resistance accurately, I can pick and choose frequencies and get perfect sine waves. With the guitar tuner software I plan to write, I can then have the modulators track to the guitar input frequency, allowing for crazy ring modulation effects that sound musical on all notes - and probably more tricks.
How I plan to do this is with tubes, now that LDR's have lost their shine for me. The low-pass filter will be done by varying the plate resistance of the heptode in conjunction with capacitor C1, while the high-pass filter will be done by varying the plate resistance of a triode in conjunction with capacitor C2.
The two resistances that need to be kept in sync are:
Heptode: (R1 * (heptode plate resistance + R2)) / (R1 + R2 + heptode plate resistance)
Triode: (R3 * triode plate resistance) / (R3 + triode plate resistance)
So, by calibrating the plate resistances at different control voltages and storing them in the microcontroller, I should be able to accurately set the resistances. If the heptode and triode resistance ranges are way out of whack, I can choose C1 and C2 as different values, to make it easier to get the low-pass and high-pass filters to have the same 3db cutoff frequency.
For amplitude control, I'll use a remote cutoff pentode - that's what they do, after all. I copied the amplitude control voltage part of the circuit from Fred Nachbaur's Dogzilla amp (see http://www.dogstar.dantimax.dk/tubestuf/dzindex.htm). This should work regardless of frequency, as opposed to the light bulbs - although light bulbs would've been cooler. I drive the RC pentode and the amplifier control voltage circuit with a cathode follower to provide current and keep from loading down the high-pass filter and ruining the frequency control.
What's not pictured in the schematics - I'll need some relays & voltage dividers to allow me to use the microcontroller & an ADC to measure plate voltages at different control voltage, so I can figure out control voltage causes what resistance. I'll also have relays to swap out different values for C1 and C2, allowing different decades of frequencies to be generated.
I also drew up how to do things with a triode instead of a heptode, theoretically at least - use the control voltage to bias the triode as shown in the diagram below.
Finally, I think it's possible to do this all very simply with a phase shift oscillator, although frequency control won't be as good and there will be distortion - but on the other hand if I just wanted to drive a tremolo instead of having a wider array of modulation effects at my disposal, this would probably work fine - see the bottom half of the schematic diagram below.
What I haven't done yet is size out all the resistors & capacitors - too much math too late at night. The risk is that the size of capacitors may be too big for the current the tubes can supply, so I'll have to pay attention to that.
Choosing tubes is relatively straightforward - more gain means more variation in plate voltage, which means more variation in plate resistance. Probably a 12dw7 so I can get a high gain triode for the high pass filter, and a high current triode for the cathode follower. Haven't looked too deeply into heptodes yet, but many are really high gain, so I'm hoping I can find a suitable one.
I do like the Wien bridge. An explanation - the R1/C2 and R2/C1 combinations make a low pass and high pass filter (or is that R1/C1 and R2/C2? doesn't really matter for the purposes of this discussion since they're all equal). The values of R1 and R2 should be equal, and C1 and C2 should be equal. This means that the frequency determined by 2*pi*R1*C1 is the least attenuated, and thus the frequency that the circuit will oscillate at. The circuit has a weakness - amplitude control. Too high and boom, square waves, too low and it peters out. It has to be just right. Normally this is done by using lamps as a poor man's thermistor, but they don't work so well at low frequencies as their temperature will start to vary. See the top half of the first hand-drawn crappy schematic below.
The strength of the Wien bridge is that it has great frequency control - if you can get good matches between R1 and R2, and C1 and C2, then that's your frequency. So if I can drive voltage-controlled resistance accurately, I can pick and choose frequencies and get perfect sine waves. With the guitar tuner software I plan to write, I can then have the modulators track to the guitar input frequency, allowing for crazy ring modulation effects that sound musical on all notes - and probably more tricks.
How I plan to do this is with tubes, now that LDR's have lost their shine for me. The low-pass filter will be done by varying the plate resistance of the heptode in conjunction with capacitor C1, while the high-pass filter will be done by varying the plate resistance of a triode in conjunction with capacitor C2.
The two resistances that need to be kept in sync are:
Heptode: (R1 * (heptode plate resistance + R2)) / (R1 + R2 + heptode plate resistance)
Triode: (R3 * triode plate resistance) / (R3 + triode plate resistance)
So, by calibrating the plate resistances at different control voltages and storing them in the microcontroller, I should be able to accurately set the resistances. If the heptode and triode resistance ranges are way out of whack, I can choose C1 and C2 as different values, to make it easier to get the low-pass and high-pass filters to have the same 3db cutoff frequency.
For amplitude control, I'll use a remote cutoff pentode - that's what they do, after all. I copied the amplitude control voltage part of the circuit from Fred Nachbaur's Dogzilla amp (see http://www.dogstar.dantimax.dk/tubestuf/dzindex.htm). This should work regardless of frequency, as opposed to the light bulbs - although light bulbs would've been cooler. I drive the RC pentode and the amplifier control voltage circuit with a cathode follower to provide current and keep from loading down the high-pass filter and ruining the frequency control.
What's not pictured in the schematics - I'll need some relays & voltage dividers to allow me to use the microcontroller & an ADC to measure plate voltages at different control voltage, so I can figure out control voltage causes what resistance. I'll also have relays to swap out different values for C1 and C2, allowing different decades of frequencies to be generated.
I also drew up how to do things with a triode instead of a heptode, theoretically at least - use the control voltage to bias the triode as shown in the diagram below.
Finally, I think it's possible to do this all very simply with a phase shift oscillator, although frequency control won't be as good and there will be distortion - but on the other hand if I just wanted to drive a tremolo instead of having a wider array of modulation effects at my disposal, this would probably work fine - see the bottom half of the schematic diagram below.
What I haven't done yet is size out all the resistors & capacitors - too much math too late at night. The risk is that the size of capacitors may be too big for the current the tubes can supply, so I'll have to pay attention to that.
Choosing tubes is relatively straightforward - more gain means more variation in plate voltage, which means more variation in plate resistance. Probably a 12dw7 so I can get a high gain triode for the high pass filter, and a high current triode for the cathode follower. Haven't looked too deeply into heptodes yet, but many are really high gain, so I'm hoping I can find a suitable one.
You do not have the required permissions to view the files attached to this post.
Re: weirdo tubes amp
I didn't look at this for the longest time, because I came to the conclusion that my custom symbol library in KiCad was too big. Eventually I bit the bullet, and rewrite my libraries, and I'm back at it.
I'm switching the low voltage portion of the power supply to be a bunch of buck regulators. I won't know the amps for the heater or relay portions until I get a bunch more done, so no sense uploading that schematic yet.
I do have a rough draft of the compressor done. Need to do the math to figure out the cap & resistor values still, but the basic idea is there. The input signal is split by a cathodyne phase inverter, and fed into a push-pull amp with remote cutoff pentodes.
These drive an output transformer. I couldn't figure out a way to do it transformerless and still have common mode cancellation. I wanted high bandwidth, and Edcor's high bandwidth transformers are ultralinear, so hey, it's an ultralinear remote cutoff amp. I don't know what that will do (if anything) to the RC characteristics, I guess I'll find out. The output signal is taken from the output transformer.
The control voltage is generated in a separate sidechain. First it hits a heptode, that controls the threshold of the compressor - so it's just a tube simulating a potentiometer. The microcontroller (not shown) controls a DAC which feeds a voltage to an inverting op-amp, which then enters as THRESHOLD_CV into the schematic, driving grid 1 of the heptode. Grid 3 is the signal taken from one end of the cathodyne phase inverter.
The sidechain amplifier amplifies that even further, sends its signal through a voltage doubler, and drives a bunch of relay-driven resistors and caps to control the attack and release. Out of that mess comes a negative control voltage that is applied to the grids of the RC pentodes.
I'm switching the low voltage portion of the power supply to be a bunch of buck regulators. I won't know the amps for the heater or relay portions until I get a bunch more done, so no sense uploading that schematic yet.
I do have a rough draft of the compressor done. Need to do the math to figure out the cap & resistor values still, but the basic idea is there. The input signal is split by a cathodyne phase inverter, and fed into a push-pull amp with remote cutoff pentodes.
These drive an output transformer. I couldn't figure out a way to do it transformerless and still have common mode cancellation. I wanted high bandwidth, and Edcor's high bandwidth transformers are ultralinear, so hey, it's an ultralinear remote cutoff amp. I don't know what that will do (if anything) to the RC characteristics, I guess I'll find out. The output signal is taken from the output transformer.
The control voltage is generated in a separate sidechain. First it hits a heptode, that controls the threshold of the compressor - so it's just a tube simulating a potentiometer. The microcontroller (not shown) controls a DAC which feeds a voltage to an inverting op-amp, which then enters as THRESHOLD_CV into the schematic, driving grid 1 of the heptode. Grid 3 is the signal taken from one end of the cathodyne phase inverter.
The sidechain amplifier amplifies that even further, sends its signal through a voltage doubler, and drives a bunch of relay-driven resistors and caps to control the attack and release. Out of that mess comes a negative control voltage that is applied to the grids of the RC pentodes.
Last edited by shoggoth on Wed Apr 02, 2014 2:47 am, edited 1 time in total.
Re: weirdo tubes amp
So as I start to do the math on that schematic, I realized I forgot to bias the cathodyne input stage.
A little arithmetic later, and I found that I'd have the grid at +90v and cathode at +93v. Well that's bad news for the guitarist plugging into that.
So
a. I could add a coupling capacitor between guitar input & grid. But if the coupling capacitor shorts, the guitarist gets electrocuted. So that doesn't seem like a good idea, unless the guitarist really sucks.
b. I could power the plate at +125 and the cathode at -125. I'd need to develop a negative supply that could deliver a clean 10mA or so at -125v. Pain in the butt, I'm sick of adding exotic voltages.
c. I could pull the non-inverted voltage off the pentode's screen maybe? I don't get unity gain (esp from the plate), and I expect the voltage swings may be unbalanced, so I'll have to attenuate the signal to match and reach unity gain. I need that small input signal so I don't overpower the remote cutoff tubes, they will sound crappy if the input signal is too big.
Anyhow if you don't see the PDF attachment (it's on the previous post), it's because you're not logged in. They get hidden if you aren't logged in.
A little arithmetic later, and I found that I'd have the grid at +90v and cathode at +93v. Well that's bad news for the guitarist plugging into that.
So
a. I could add a coupling capacitor between guitar input & grid. But if the coupling capacitor shorts, the guitarist gets electrocuted. So that doesn't seem like a good idea, unless the guitarist really sucks.
b. I could power the plate at +125 and the cathode at -125. I'd need to develop a negative supply that could deliver a clean 10mA or so at -125v. Pain in the butt, I'm sick of adding exotic voltages.
c. I could pull the non-inverted voltage off the pentode's screen maybe? I don't get unity gain (esp from the plate), and I expect the voltage swings may be unbalanced, so I'll have to attenuate the signal to match and reach unity gain. I need that small input signal so I don't overpower the remote cutoff tubes, they will sound crappy if the input signal is too big.
Anyhow if you don't see the PDF attachment (it's on the previous post), it's because you're not logged in. They get hidden if you aren't logged in.