weirdo tubes amp

[shoggoth]

So in this thread (http://ampgarage.com/forum/viewtopic.ph ... highlight=), boots explained he had a bunch of leftover tubes from some radios, and had a bunch more loctal radio tubes sitting around. So I took them all, and now I'm going to build an amp from them (well once my Zenith conversion is done - one of the pentodes gave up the ghost, so I'm waiting for a replacement before finalizing some of the voicing).

So here's what boots sent:

1R5 - heptode, directly heated cathode
1U4 - pentode, directly heated cathode
1U5 - diode & remote cutoff pentode, directly heated cathode
35W4 - half-wave rectifier
3V4 - 1/4w pentode, directly heated cathode
2x 50C5 - 7w pentode, 50v heater
2x 7B7 - remote cutoff pentode
7F7 - dual triode
7F8 - dual triode
7G7 - pentode
7H7 - remote cutoff pentode
7V7 - 4w pentode
7W7 - 4w pentode

So that's a lot of goofy tubes. I bought a bunch of loctal sockets off eBay cheap (because they were missing brackets - no problem there, the necessary little flanges can be cut into the chassis using CNC), and a bunch more sockets from ApexJr (7 pins are $0.35 each!).

The power transformer cost $3.95 from ApexJr (3" X 2 5/8" w/ mounting ears, 115Vac input with voltage taps of 48V, 40V, 24V and 8V, 1 amp estimate). Multiple Greinacher voltage doublers can get me to the voltages I need, and the 1a current works for everything except the 6.3v heaters. I'll have to get a separate PT for that. For the output transformer, I'm using a Classictone Princeton Reverb style 15W transformer. That cost me $31.

Remote cutoff pentodes and heptodes are a challenge to use. With a pair of RC pentodes, though, I can theoretically put together a nice push/pull compressor, and a third can be the cathodyne phase inverter - the small signal voltages from the guitar means it won't have that horrible remote-cutoff distortion.

The heptode is just a straight-up heptode, no remote cutoff features, so I can use it as a modulator. I'll use one of the pentodes in a lamp-stabilized Wein bridge, because more vacuum is always better, and have a tremolo that way. Not sure if a lamp Wein bridge will get to the low frequencies I want for a tremolo, it'll be an experiment, or if I can even get a Wein bridge oscillator working with a single pentode. I'll find out though.

So, the topology:

Cathodyne phase inverter / first stage - 7H7 remote cutoff pentode
Push/pull compressor - pair of 7B7 remote cutoff pentodes
Gain - 1U4 pentode. Preamp gain happens right after this stage
Side chain feeding control voltage back to compressor - 7W7 pentode, through a voltage doubler
For "high gain" channel only - two more gain stages using both halves of 7F7 dual triode
Cathode follower to feed FX loop - 7G7 pentode
FX loop recovery - 3V4 pentode
Modulator - 1R5 heptode
Oscillator - 7V7 pentode
1/2 of 7F8 dual triode - gain stage feeding master volume
1/2 of 7F8 dual triode - cathodyne phase inverter
2x 50C5 - power amp

This leaves the 1U5 diode/RC pentode. There's only so many remote cutoff tubes you can stick in an amp, so this one's being left behind. The 35W4 half-wave rectifier is also being ignored - it'd just be an act of masochism to jam this into the amp.

There's a couple of big experiments here - I want to couple the push/pull compressor directly to the following gain stage instead of using a transformer, so I'll have to figure that out. I might have misjudged the relative gain between stages and be lacking somewhere. The input stages might be very noisy. And the oscillator to feed the heptode modulator could be a pain to design.

But even with those challenges, that's no reason not to make things much, much more complicated. I also want to try driving things with microprocessors - having presets with the effects turned on/off is fairly easy, but I think it would be cool to store the knob settings as well.

There are couple of ways this could be done:

a. Motorized pots. I can't find a source for these. If I could, I'd go for dual gang pots, so I could read the resistance of the pot to know its location, and then turn the motor until I found the resistance I was looking for. Doubt I'll go this route though, since it doesn't involve blinky lights, and I love blinky lights

b. Dual-gang pots, use the resistance of the pot to know its location, and store that. When a preset is loaded, turn on an indicator light to show the user when they've turned to the right spot. Could circle the knob with lots of little lights, and have amber/green led's showing the general location, and then go green when you're spot on. The H&K coreblade does something like this. It seems really cumbersome to me, but should be very accurate.

c. Digital potentiometers. These are typically low-voltage parts, and low-resolution as well (usually 8 bit). Signal voltages tend to get high w/ 12ax7s - I don't have those though. My signal voltages are smaller, I could probably get away with the AD5290 (100k, 30v) or AD5282 (200k, 15v). They're surface mount parts though and I hate that, and they couple the tube amp ground to the digital ground, which might stink. They also make zipper sounds if you don't change their values at the zero-crossing of the signal, but hey I've got a microprocessor - I can just wait to send the change until I see a zero-crossing of the signal. That's no big deal.

d. LDR potentiometers. A pair of LDR's can be hooked up as a potentiometer. They can be calibrated exactly in a table stored in the microprocessor code, so that I know which control voltage results in which resistance, which makes their non-linearity no problem at all. The issues left are thermal variance, variance due to aging of the CdS cell, and variance due to aging of the LED. If thermal variance is proportional in both cells comprising the two halves of the pot, then no issue there - the voltage divider remains the same. The LED and CdS cell aging may have issues though - one side is likely to be lit more than the other in general, and they may change unequally on both halves of the pot. There's also distortion at high signal voltages, but this amp won't have super high signal voltages, so I'm not too worried about that.

Mesa's Triaxis amp uses what is effectively a digital potentiometer and four CdS cells activated by two LED's, with one digital potentiometer providing feedback to the other, so that the value of the digital potentiometer is matched. This has the benefit of eliminating nonlinearity (which I don't care about since I can have a table mapping control voltages to resistances), and of eliminating aging and thermal issues (which I do care about). It's also patented, and relies on matching the CdS cells, which seems like a hassle. I think I'll skip that work.

I can think of two approaches to the LDR pots:

a. Put them in the circuit where they are needed. The control voltage shouldn't have much noise on it, so just run the control lines straight to them
b. Put them together in an enclosed location, have shielded wires or whatnot run from the tube amp over to the LDR pots. Put a Peltier cooler in with the pots, and have it keep the temperature in the LDR pot chamber (LDR chamber pot?) fixed at 25c or something. Should reduce thermal variance, but you've now got possible condensation to deal with and increased signal wire runs in the amp.

Not sure what I should do in this respect, suggestions are welcome.

The knobs on the panel will be rotary encoders. Since they rotate freely forever, there's no indicator on the knob itself - indication can be done in two ways I can think of

1. A circle of LED's around the knob, lighting up to show the "rotation"
2. A 3- or 4-digit numeric display telling you exactly what value you've chosen

I'm going with #2 - better to be exact, I think. Plus, with CNC you can make the front panel look really good. Instead of switches to change between options, I'll be able to use rotary encoders as well - turn to change between channels. I can use the LED displays to print which channel you're on. Unfortunately I couldn't find a source for 1/4" 14- or 16-segment LED displays (or good driver chips for them either), so I'll just use 3-digit 7-segment displays and try to render letters onto those as best I can.

I've also got tiny LCD displays I picked up cheap to show the list of presets stored, and probably display other info if I get ambitious - maybe some ADC's could show voltages throughout the amp, and maybe a graphic equalizer display while you're playing. Who knows. I'll worry about those details later.

So, totally over-complicated and pointless, but it's a hobby, not a job, and I've got no deadlines. A lot of the parts have been ordered (& arrived) for the digital stuff, and I'm doing PCB layout now. It'll take a while though, so updates will come very slowly.

Anyhow, here's a schematic of the LDR pot that I'm thinking of. It's not complete, I didn't connect all the control lines to the MCU and didn't add the power filter caps, but it should give you an idea. This is using an 8 bit DAC to control a current sink, I might need to switch to higher precision. I'll breadboard a test circuit and see what precision I can get when the CdS cells show up in a few weeks. As it stands, one of the MAX529 chips and two ULN2075B's should be able to control four LDR pairs.

[xtian]

Holy cheese toast, man! This is a dozen thread's worth of madness in one! Subscribed!

[tubeswell]

Schematic please.

[shoggoth]

It's not built yet - no schematic yet. I have messy block diagrams but they don't say more than the topology I posted.

The digital stuff will probably be built first (because until I design the display circuit boards I can't design the chassis), but I'm not putting together that schematic until I breadboard the CdS cells, and they haven't arrived yet.

I've got some days off next week, I'll see about getting preliminary schematics done on the compressor & modulator stages.

Anyhow my explanation of the Triaxis quad-LDR method is a bit incoherent, it was late at night when I posted. Just check the patent for how Mesa did it - https://www.google.com/patents/US520854 ... CDYQ6AEwAA

[M Fowler]

I have a bunch of odd tubes let me know if you run out and need a tube.

[shoggoth]

I have a bunch of odd tubes let me know if you run out and need a tube.

Thanks for the offer - hopefully this batch is sufficient!

I think for the next amp I build I'll choose tubes based on their design characteristics instead of weird experiments to see if any random assortment of tubes can be made into a sweet amp.

[shoggoth]

The Mouser order showed up with all the parts I need to build a test rig for the microcontroller, displays, & digitally controlled LDR's - got it about halfway assembled, I'll probably wrap it up by Tuesday. Then, on to programming!

Now, for schematics - start with the basics. I've attached the schematic for the power section. It's done weird, because these are the transformers I've got on hand. The 52v transformer was only $4 from Apex Jr, and the Antek toroidal 150v transformer was one I ordered by mistake.

The 52v transformer will be responsible for the 50v heaters on the 50C5 power tubes. I'm also putting a voltage quadrupler circuit on it, that will deliver around 140v according to my math. It could end up higher or lower in reality, I haven't had a lot of luck getting math and actual voltages to line up, but it's probably in the ball park. That 140v is about perfect for the 50C5 tubes, and then I'll use resistors to drop the voltage for the various 90v tubes.

The Antek toroid I'm going to wire up backwards - 115v into the 150v taps. This gives me around 240v DC, which is what I need for the rest of the tubes. It also gives me 10v DC for the low voltage supplies.

There's a bunch of directly heated 1.4v tubes mixed in with 6.3v tubes, so I don't want AC heater wires lying next to the 1.4v DC wires - that will be way too noisy. So the 6.3v supply will be regulated DC. And there'll also be 5v and 3.3v DC for the various digital components.

My only question right now with the power supply is fusing. With the fuses in the schematic as shown, they will pop instantly due to inrush current filling the capacitors. I don't know if slo-blow fuses are a good idea, or if they'd survive the inrush current - so maybe I should move the fuses after all the big caps and before the dropping resistors ?

[shoggoth]

Whoops, just noticed in that schematic I didn't change from ideal capacitances to affordable and available capacitances - those 4700 uf's in the high voltage circuits are going to end up being a 470 uf in the 140v section and a pair of 330 uf's in the 245v section.

[boots]

Wow Shoggoth, that makes my head spin! Good luck. I hope you'll post some sound clips when it's up & running. I'm wondering what a pair of 50C5's will sound like.

[shoggoth]

Huh, I'm out-clevering myself. I did my arithmetic wrong on the quadrupler, that's 280v coming out. Twice what I want. I should just use a voltage doubler instead. That's what I get for designing circuits at 11 pm.

Maybe I can tap the quadrupler in the middle somewhere to get the 140v, keep the 280v for the higher voltage circuits, and replace the toroid with a smaller 9v transformer.

[shoggoth]

More 11:00 pm circuit work. Here's a corrected power supply schematic. Still using the voltage quadrupler, but tapping into it at the halfway point to get the 140v I was looking for.

I've tossed the idea of using the 150v Antek toroid, I'll just get a 5A 9v transformer for the low voltage supply, that'll save a couple of pounds.

[shoggoth]

A few more revisions. I've added some indicator LED's to the filter caps to show that they're charged and dangerous (an idea I found on Merlin's website - note that the 6.3v supply is from his book as well), and replaced the cap discharge 100k resistors with voltage dividers that I can feed to an ADC, so I can monitor voltages from the microcontroller.

I also dropped in a thermal sensor IC (the TC74A7), I'll have those sprinkled throughout the amp so I know how hot things are getting inside the chassis. This one will be hanging around all the regulator heat sinks - maybe I should attach it to the one of the bridge rectifier / voltage regulator / power transistor heat sinks directly? Ambient temp may be good enough.

I'll power a pair of quiet 12v fans to push air over the tubes & transformers. There's a lot of tubes, and the transformers will be running close to their limits, and those temperature-sensitive LDR's are going to be in the chassis, so I want to keep air circulating.

All the low voltage filter caps are sized correctly now, I think.

QUESTION - are these the spots people normally fuse the B+ power supplies (and low voltage supply as well) ? I worry about inrush current popping the fuses as those filter caps charge up. Note the transformers are near their limit, so maybe that will restrict current? Or maybe they'll overdeliver for the second(s) it takes to charge the caps.

[tictac]

You can use a CL-150 inrush current limiter on both B+ legs if you're worried about inrush current... There are some HiFi guys doing this for classis Harmon Kardon and Marantz tube amps...

TT

[shoggoth]

Thanks, I'll check out the data sheets on that CL-150.

I had some math errors (typical) in the power supply, so I've got to redo parts of that, but what's there is the general idea. When I get around to it, I'll update that part of the schematic.

I just completed putting together the test rig for the digitally-controlled LDR pairs. It tests:

a. LED display
b. Rotary encoders
c. Debounce chip
d. I2C i/o chip
e. DAC driving current sink driving LEDs driving LDRs

Those are the primary components of the digital portion of the amp. I'll be tossing in some temperature sensors and voltage monitors too, but I think I'll just breadboard those when I test their circuits. The LCD display I already tested w/ a breadboard, very straightforward to interface with.

You can see a crappy phone photo of the test rig below. One protoboard wasn't enough for all the chips, so I put the CPU and power regulator on one board, and did the rest of the test rig on a daughter board. The gray box is the AVR programmer.

Now I just need to get this bad boy programmed. That will probably take a while, all I've done with microcontrollers is some minor Arduino stuff.

[shoggoth]

The test jig works well enough, although now that I'm trying to adjust references voltages into the DAC it's being finicky (I'm probably not accounting for some impedance in the chip, I'll have to go through the data sheet at some point and figure it out).

But it's good enough to measure some of my hand-made LDR's.

Voltage / 30mA LED Resistance / 20mA LED Resistance
2.99v / 970R / 1.5K
2.84v / 4.6k / 9.7k
2.72v / 7.7k / 19.k
2.6v / 12.3k / 36k
2.48v / 23.5k / 84k
2.38v / 50k / 202k
2.24v / 118k / 500k
2.18v / 350k / 1.45M

The resistance gets unstable above that 200k, but I didn't have enough coupling capacitors so I don't know how stable the DAC was. That's a bit disconcerting, I don't want to be introducing noise.

So I'm liking the 30mA LED with the CdS cells I've got. I'll need to figure out my issues with shrinking the DAC voltage range to 2.1v - 3.0v to get the most usable values out of the 8 bit precision it gives me.

I think I'm going to have to add relays to each LDR to pull it out of the circuit so I can calibrate. I don't know that I have to do it on startup of the amp, maybe a manual calibration performed periodically will be enough, with value stored off in some EEPROM.