My HV tube project (inspired by the GTFO & Voltage Multiplier threads)

[ubersam]

I'd like to request corrections or suggestions from those more knowledgeable on this subject than I, before I commit the schematic to a layout.

After reading the aforementioned threads, I was inspired to start my own HV smps powered tube preamp project. A couple of years later, I finally managed to draw something up. It started out simple, a copy of the Fender Bassman 5F6A preamp section without the dual channel input. My intention is to use it with a dedicated poweramp, or in the effects return of an amp. But as with many things that I start simple, it evolved as I looked at other preamp schematics. Particularly the Fender 5E6A and the Marshall 1959. It looked to me like the differences (in the preamp section) were primarily different Cathode bypass capacitor values and/or Cathode resistor values, also, different treble bleed (bypass?) caps on the volume control. I thought it would be simple enough to make them switchable, so I drew it up. Then came playing with the Duncan tonestack calculator and LTspice simulations of the tonestack, which led me to make the treble caps switchable too. I initially drew the switching as a rotary switch to switch between different values: https://www.dropbox.com/s/r5o6npdztf8p78q/SMPS%20TUBE%20PRE%20-%20PREAMP%20V1.pdf?dl=0

I've done similar things to pedals but then I thought, low voltage vs. high voltage. The same scheme might not work as well in a high voltage circuit. So, I looked at more schematics for other examples. I found some that had a large value resistor in series with the cap. So I copied that: https://www.dropbox.com/s/fes0qg7b4iivqby/SMPS%20TUBE%20PRE%20-%20PREAMP%20V2.pdf?dl=0

The first schematic has about 19 resistors less but I'm not sure how well it will perform, i.e. will there be switching noises (clicks, pops, etc.)? The second has additional resistors which I think will help minimize switching noises. But will the difference be significant enough to warrant the additional parts?

In simulation, a 100mV/1KHz input will generate a 17V/-16V (33V_PP) output, and a 1V 1KHz input will generate a 115V/-82V (197V_PP) output. Should I consider padding the output? Something like a 470K||470p before/after C39? Or perhaps a local feedback on the output buffer?

I appreciate your input/s.

[GibsonGM]

Hey Sam,

I love these projects   Neat ideas.    I can see what you're driving at....the ultimate switchable everything, ha ha!  Very ambitious and a lot of work.  A little unconventional, but that may be what you want.     Generally, one would experiment with bypass and coupling caps on the board, and end up with permanent values before finishing off the project, but if you really want to make them switchable, that's up to you!   I think you'll find that you set the switches to 1 or 2 values and then leave them there a lot...     

It's a lot of work, but if you understand that the payoff might not be as magical as you could think, it's your baby....your thinking about the differences (cathode bypass, coupling and bleed caps..) is right-on - knowing WHY they are those values - what the designer intended - is the priceless thing! 

I think you'll have switch pops where there's DC no matter what, altho the resistors should make them less prominent. Not a big deal unless you plan to do a lot of switching on the fly??   Have you simmed the resistors in LT Spice to see what affect they could have?

As for the output, if you're running this into an amp, I'd definitely limit the output to whatever that amp can tolerate!   Tubes sound so good driving an amp, it's easy (for me anyway) to keep upping the output, which theoretically could damage some input circuitry. 

You have your work cut out for you, Sam...if you haven't built something easier, like a duo-triode booster on HT or something, maybe you'd want to hold off and do so?  But I won't try to keep a man from his dreams!   This may be exactly what you want to do      Merlinb ("Valve Wizard") lays out a pretty good argument for why they don't make preamps switchable this way in his book, and I agree - once set, they're pretty much in the sweet spot already and tweaking won't help.  Final choice is yours, tho.

[Transmogrifox]

I'm really interested in this too.  I'm in the process of prototyping my own HV tube box.

This was inspired by my brother buying a new guitar, and he doesn't have any more than a PA for an amp and he said his friend's tube box sounded really good direct into his PA -- so I knew what to make him for Christmas.  Now it's February and it's still an unfinished project -- got my 300V SMPS working, then I loved my simple self-oscillating boost converter so much I made a boost to 25V to series 4x6.3V heaters -- and that is working.

I'm not going to make him one with a pile of switch options, but I would be interested in knowing what values you stick with.   Besides, I might make one for myself with a pile of switches.  At the very least I would probably use 2-pin jumpers to make experimenting easy.

Have you done any audio simulations?   That can help you make some decisions before you commit to a layout.

I have simulated mine in LTSpice using a .wav file input, and it generates a .wav output.  I must say the Duncan Amps Triode models are pretty darn good.  I have had success with this simulation method also with a FET distortion box.  The sound wasn't exactly the same, but it was close enough that there weren't any major changes from the original design.

The basic effect of frequency emphasis/deemphasis will show up in the simulation .wav output, while the touch dynamics and all that stuff -- well you have to feel it in your fingers anyway.  A simulation can't replicate that

[ubersam]

 

...I can see what you're driving at....the ultimate switchable everything, ha ha!  Very ambitious and a lot of work...Generally, one would experiment with bypass and coupling caps on the board, and end up with permanent values before finishing off the project...

Yeah, it does look like a lot of work. But if I break it down into modules, it shouldn't be so bad, i think. Having the ultimate switchable everything wasn't really the goal. That is more for convenience. I could build using some default values and swap them out as I go. But knowing me, I'd forget how it sounded/felt by the time I unsolder the old part and swap a new value in. Having them on a switch makes it easier to go back and forth with the values. That is what I do when breadboarding low voltage projects. I just hope the click or pops wont blow my speakers. Eventually though, I would like to settle on some specific values, maybe one or two things on a switch.

...I'm not going to make him one with a pile of switch options, but I would be interested in knowing what values you stick with...

I'll post an update when I get that far. At this time I still need to create a component pattern for the rotary switches and the noval socket.

Have you done any audio simulations?   That can help you make some decisions before you commit to a layout...

That is interesting. I did not know something like that could be done. I'll have to look into that. I still have my old Firebox and a Shure E-something mic (flat, square mic).

[GibsonGM]

LT Spice is a free program (Transmog mentions it above), you can run all this in simulation and actually put a wav thru, and get an approximation of what it will sound like!  Keep it short, tho, takes some time to process

I look forward to hearing what you come up with, Sam!  I'm the sort that reads up on what 'works'...what combinations of cathode bypass, coupling caps etc, work for what 'styles', and WHY....I tack them up, play with them, and unsolder/solder in stuff as I go.  This is (like I said) very ambitious, and I bet you'll learn a ton by doing it!   Don't forget to mess around with load lines and do some freq. calculations, if you haven't already!

**EDIT**

PS - about the switch pops...don't know if they exist, but make-before-break rotary switches would prevent that...if you can't find/get them, how about a kill switch on the output before switching?  Just to mute the output (could ground the output...).

Or, if you know what voltage level you want at the output into your amp...say 7V P-P max for discussion purposes...maybe you could put a pair of clippers in there set up for 10V?  Thus sending anything over that to ground rather than into your amp.  That was a common way ham radio guys kept spikes out of their headphones...so a transient wouldn't hurt your amp.   They can be made 'adjustable' by using a reverse bias voltage and a pot....just some thoughts.

[Transmogrifox]

into that. I still have my old Firebox and a Shure E-something mic (flat, square mic).

You can also do a quick web search for clean electric guitar samples if you're under-ambitious about setting up your recording stuff.  Lots of this stuff exists around the web because many companies post samples of their effects, usually one "clean" and some audio files of "wet" so you could rip off the clean recordings and put them through LTSpice.

If you're using a simulation program other than LTSpice then you will have to check the documentation to find out if it provides the ability to input a wav file and output a wav file.

As GibsonGM says, long audio samples take a long time to simulate.  The samples I use for simulation are kept to 10 seconds or shorter.  It seems 5 seconds is the minimum for really being able to form an impression of the sound.

[Transmogrifox]

By the way, here is the self-oscillating SMPS I am using:


The prototype I actually built uses 10k for R1 and 100 uH for L1.  The 10k gives more gain for more certain startup and the lower inductance brings switching frequency down to about 145 kHz, which makes this more efficient.

Then I used an identical version to make a regulated 25V supply for 4 series heaters by changing the stackup of zeners to ~25V and the inductor to 47 uH.  Both work like a charm and they're super simple (I think).

[ubersam]

I have been using LTspice to sim the freq. response and voltages. Although, I'm not familiar with it enough to know where to plug the .wav in. Heck, I didn't even know that was possible.

I thought the same about the make-before-break rotaries. I remember seeing those at mouser a few years back when I bought a handful of the break-before-make kind for another project. IIRC, they have the same footprint as the ones I have. So if need be, it shouldn't be too hard to take the b-b-m out and pop an m-b-b in.

As for the output levels, I've looked at the schematic of the Triaxis to see how they did the output. That preamp has a very hot output. I remember having to set the output pot no higher than 8:00/8:30. I've also looked at schematics of a couple of power amps to see their inputs and maybe determine how much they can handle. I need to look a little more because it hasn't quite sunk in yet. At this time, I'm leaning more towards a voltage divider kind of padding.

By loadline calculations, do you mean something like this: http://www.ampbooks.com/home/amplifier-calculators/12AX7/, I've also downloaded a blank 12AX7 curve/chart thing to try out. I've done the freq. calcs. too using AMZ's calculator. I've also set up an excel spreadsheet with the formulas plugged in, that's more so that I can play with it at work and still look like I'm doing work.

I've read stuff on MerlinB/Valve Wizard's site, tons of info there. That's were I got the idea to use the last triode as an AC coupled cathode follower as opposed to another DC coupled cf.

[ubersam]

Here's the voltage multiplier that I will be using: https://www.dropbox.com/s/9j3wsnb9ulu2rsv/555-smps.pdf?dl=0 - I was seeing a huge current spike during simulation which seemed to coincide with the speed at which the inductor was charging. So I simulated slowing down the rate at which the voltage rose from 0-12v and ended up adding Q4, a couple of resistors and a large-ish cap. Since I was already at it, I added R.G.'s reverse polarity protection, and a bi-color/direction LED to indicate polarity of 12v supply. I'd probably build this first, separate from the preamp. Still debating that though.

[merlinb]

The first schematic has about 19 resistors less but I'm not sure how well it will perform, i.e. will there be switching noises (clicks, pops, etc.)?

Switching cathode-bypass caps will pop unless they have discharge resistors. The tone caps are much smaller, so the pop will be smaller, and maybe not noticeable. Having said that, I think you have way too many switchable caps, and you will hardly ever use them. Simple 2-way switches would provide more than enough options, and you can also get waaay more versatility from a shift control, i.e. replace R26 in the tone stack with 33k and add a 100k pot (variable resistor) in series with it.

An arc-protection diode for the cathode follower wouldn't go amiss either.

In simulation, a 100mV/1KHz input will generate a 17V/-16V (33V_PP) output, and a 1V 1KHz input will generate a 115V/-82V (197V_PP) output. Should I consider padding the output? Something like a 470K||470p before/after C39? Or perhaps a local feedback on the output buffer?

Whether you pad or not is up to you, but you'll definitely want to add a pair of 10V back-to-back zeners from the output to ground, to clamp high voltages. You don't want to be sending hundred-volt transients down the cable to 9V pedals!

[GibsonGM]

^^ This ^^   

And this might be helpful:  http://electrostud.wikia.com/wiki/Using_WAVE_files_as_input/output_in_LTSpice

[ubersam]

...Having said that, I think you have way too many switchable caps, and you will hardly ever use them. Simple 2-way switches would provide more than enough options, and you can also get waaay more versatility from a shift control, i.e. replace R26 in the tone stack with 33k and add a 100k pot (variable resistor) in series with it.

It does have many switchable caps now but that is only to make auditioning the different values easier for me. So the switches will see plenty of use in that regard but I do agree that once I settle on the one or two values that I like, the switches will not have as much use. At that time, I will just replace the switches/caps with the cap value that I like. I do like the shift control idea. Much simpler than the switchable treble caps. I ran a sim last night and it looks like it moves the LP cutoff

An arc-protection diode for the cathode follower wouldn't go amiss either.

Would this be the resistor and diode in series from grid to cathode on the DC coupled cathode follower?

Whether you pad or not is up to you, but you'll definitely want to add a pair of 10V back-to-back zeners from the output to ground, to clamp high voltages. You don't want to be sending hundred-volt transients down the cable to 9V pedals!

Like the clippers that Mike was suggesting? I have some 9V and some 24V (or 48V?) zeners. I don't plan to plug this into pedals but into the input of a dedicated power amplifier, or an amplifier's effects return. I have a little mixer that I used with the Triaxis to run parallel effects with a Lexicon unit. I might plug into/through that as well, or the Lexicon unit.

And this might be helpful:  http://electrostud.wikia.com/wiki/Using_WAVE_files_as_input/output_in_LTSpice

Excellent! I gonna have to try that this weekend.

[GibsonGM]

Clippers like I suggested - yes.  Zeners really are the best way, Merlin is right - regular diodes work but not the same way. The 9V is probably your best bet if that's what you have, will get you an output of about 18V P-P before they kick in and clamp the voltage if you set them up back-to-back like clipping diodes on the output - saving your power amp input       Find out what the ideal input level is for the power amp....the 9 or 10V area is a SAFE one for most equipment, but may not be what the poweramp 'wants'.   Too high an input can blow out your input circuitry, like the input coupling cap in the amp.

Your tube project COULD, in theory, put about 2/3 the supply voltage, if you don't attenuate!!   So you'll want a voltage divider at the output, PLUS the zeners.    I'd use an output cap, too (didn't see if you have one).   Set the divider to get you down into that (maybe) 15V RMS AC (as measured on your meter) before you plug into the amp and you're golden.   You can then put a pot behind THAT to control actual level. 


The arc protection is the R and diode, yup.  There could be some arc potential there; those are high voltages on the grid.   Try to get as much advice from Merlin as you can - he is VERY tube-oriented and I've learned a volume of info from him!   

[Transmogrifox]

Here's the voltage multiplier that I will be using: https://www.dropbox.com/s/9j3wsnb9ulu2rsv/555-smps.pdf?dl=0 - I was seeing a huge current spike during simulation which seemed to coincide with the speed at which the inductor was charging. So I simulated slowing down the rate at which the voltage rose from 0-12v and ended up adding Q4, a couple of resistors and a large-ish cap.

*EDIT:  Removed comments about inrush limiting (fast 12V power on FET).  Further thinking this is actually not a bad idea and accomplishes the same end as a more elaborate soft-start * /EDIT

Anyway, a few thoughts to consider (may be all are unfounded and the SMPS as it stands is fine)
A higher switching frequency will reduce peak switching currrents and keep intermodulated noise and duty cycle jitter further from the audio band.  

Something like a 150 kHz to 200 kHz switching frequency is perfectly reasonable.  

Another advantage of a frequency an order of magnitude higher than anything audible is the filtering requires much smaller capacitors.  The actual PCB real-estate for a 400V 1 uF cap is much less than  the typical 10k-47 uF ladders to several hundred uF caps used to filter 60 Hz.

However, to realize a design with small HV filter capacitors is how clean your 12V supply is, and how tight your regulation loop runs.  The 120 Hz or any other audible noise on 12V will be fed through the SMPS.  The higher the feedback loop gain, the better the rejection to 12V fluctuatons.  A big resistor divider on a BJT base is pretty low gain feedback, so your SMPS will be pretty much a slave to whatever 60/120 Hz noise is on the 12V input.

You may consider a simulation with a typical 12V supply that you intend to use, including the AC rectification and filtering to determine the amount of 120 Hz rejection exists in your design.

If 120 Hz couples into the first gain stage then it will be /hella/ noisey and you will be confused about where all the 120 Hz (or 100 Hz if in Europe) hum comes from.

If you put several Zeners or TVS's stacking up to 350V in series with the BJT feedback resistor, then you'll get higher feedback loop gain since the output AC will essentially appear through a much smaller resistor value to the base.  This will largely regulate the powerline hum  out of the system and reduce the need for filter capacitors.

Obviously too much feedback gain will cause you to have to consider gain/phase loop stability stuff, but the cheap 'n' dirty is to just back off on the gain until it's stable (tighter regulation requires more rigid stability analysis).  The second cheap 'n' dirty trick is the dominant pole compensation.  If you don't care for tight regulation above  120 Hz, you can set a dominant pole in the feedback at say, 250 Hz and have a really sloppy reaction to transient, but generally washes the higher order effects to less than unity gain. 

Anything you lose in regulation precision you need to compensate with large HV caps and RC networks.  You still have to think about 120 Hz noise unless you are certain your 12V input is really well regulated and immune to power line frequency fluctuations.

How much is too much?  Suppose even though your output may be limited to 10V p-p, you generally use it like stompbox, so something like 2V p-p is a "hot" output.  You might want to aim for something like less than 1% AC (120 Hz) noise to be substantially inaudible.   This means AC noise at output must be 20 mV p-p.  Looks like you're shooting for something on the order of 350V DC on HV rails, so your output signal is 350 V p-p on the final stage.  It gets divided down from there, so the 1% rule still applies.   AC noise at the final amp stage should be 3.5V p-p to be tolerable.

This seems easy if it was just a cathode follower, but consider 2 gain stages ahead of it = 60*60 = gain of 3600!.  Now suppose the 120 Hz noise couples into the first stage at a gain of 1/4, all things considered, the max 120 Hz noise you can tolerate is 3.5 V/900 = ~ 4 mV p-p (ouch!).

If you have 1 V p-p on your SMPS output, you need 4 mV/3.5 = -58 dB rejection at 120 Hz.

This means you either need 3 RC filters with cut-off at 12 Hz, or one RC filter with cut-off 0.12 Hz to get this much rejection at 60 Hz...you get the idea that more gain means more aggressive filtering needed to cut the hum.

I think I have used too many words to say you should consider running simulations with a worst-case noisy 12V supply to make sure the 120 Hz hum won't become audible.  An SMPS doesn't guarantee against that unless your regulation is really tight.

[ubersam]

Update: Finally finished it:

[tubegeek]

Update: Finally finished it:

Sweet project, excellent build!

How does the "end user" (as opposed to the "applications engineer") like it? Any comments on the sound and the most useful options?

[ubersam]

Thanks! So far the impressions are good. I've never played through a real 5F6A but the cleans are bright, chime-y & snappy, definitely Fender sounding, IMO. I added a master volume so I can turn up the 'volume' to get overdrive without having to get really loud. It gets a nice classic rock crunch with the 'volume' around 1:00~2:00, think Rolling Stones Brown Sugar but a little brighter. 2:00~3:00 gets to AC/DC-ish sound, specially if I engage the 2nd triode stage's cathode bypass cap. One of the caps cuts off around 285Hz, like the Marshall 1959. It gives more drive but loses some bottom. Another cap cuts around 81Hz, which is useful to give more drive but keep some of the bottom.

With the 'volume' turned up high, and the 2nd triode stage's cathode bypass cap engaged, the bottom can get tubby. If rolling back the bass knob isn't enough, that's where the 1st triode stage bypass cap selection comes in handy. There are six choices, the first having no bypass cap then 5 different cap values. the cutoffs are 80, 60, 40, 30, and  1 (to mimic the 5F6A). So far I haven't decided on which cutoff is preferred in this situation. I like 80hz in most cases, an in some I like 40hz.

For the bright switch, I have multiple values on a rotary switch but I've been preferring the 220pF cap most of the time. But if things get too treble-y, I'd switch it off in conjunction with rolling back the treble.

For the mid-shift, I opted for having multiple treble caps on a switch instead of the pot in series with the 56K resistor. I haven't played much with this so for now I've mostly been using the Fender value, switching to the Marshall value to compare.

I only have hum-bucker loaded guitars so I haven't done a proper single coil test. I also want to test it with my basses.