SMPS for vacuum tube power amplifiers.(status: back at it)

[Circlotron]

Heres another look at one of the power pentodes.

6CA7 / EL34. Mmm...

[SK_Caterpilar_SK]

If input impedance is an issue, you could also use a low power pre-amplifier before the transformer, which would only need a small power transformer. Another possibility is to have a differential input stage for the valve amplif and couple the signal via Y-rated capacitors, which are safe to connect between the mains and exposed metalwork or a low voltage signal.

I knew you wouldn't like my suggestion, but that was the point. I personally like your project, but as mentioned above, many thermionic purists won't. The fact that it sounds the same and the oscilloscope output is identical doesn't matter. It's phycological. If I told you it's possible to use digital signal processing to make an amplifier with exactly the same characteristics as a one with glowing filaments, would you be interested? Probably not. Many people like old technology. This is no different to steam enthusiasts vs those working on cutting edge electric locomotives.

Added complexity. The preamplifier stage is  not necessary and adds probably a lot more distortion than desired.

I know it is possible to use digital processing to replicate the characteristics of a tube amplifier no doubt. But it aint the real thing, aint the visual appeal. I however did design a lot of chip/transistor amplifiers so I know myself that tubes are basically long obsolete and transistors are what it is todays and class D amplifier achieveving the same sound quality as any AB or even class A amps can get with the power rating of a D amp. The point wasnt really about the tube amplifier itself. Its more like a showpiece. But HVDC can be usefull in a lot of applications, HIFI guys are surely not gona agree but some quitar guys want a  portable solution for tube amps and really replacing the transformer by this solution would basically half the weight of the thing. The HVDC supply does not have to be in a audio specific application. Transmitters use high voltage, and many other devices.

[SK_Caterpilar_SK]

Heres another look at one of the power pentodes.

6CA7 / EL34. Mmm...

Yes correct but its the EL-34L. The thing is made universal so it can use 6L6s, KT88 and such oktal power pentodes. The real advantage that would be given is by the KT88. You can upp the voltage on the anodes and get more power out of the thing.

[SK_Caterpilar_SK]

So over at the other forum people relly harrased me for wanting to make my own HV converter. Those people treated my work as a copy and paste, and were pushing their shitty chinese crap sinchronous inverters. Those converters do require quite the modifications to work as intended, but those supplies cant achieve what mine can, for example just directly connect to the output and call it a day instead of screwing arround with bs LCLCLCLCLCLCLCLCLCLCLCLCLCLCLCLCLCLC filters... cause the end of the day you end up with no power and thats basically geetting you nowhere. Also it gets no output regulation so in order to achieve your voltage you got to make a doubler for it. While mine just goes directly from the transformer to the diode and a cap and done. No aditional filtering required-small light simple. But considering the really crap response from those guys, I might just give up entire tube amp project and leave it as is. Afterall, who am I to create something new, a 17 year old student? Must be dumb as shit. I should just go get drunk all day night like my other classmates do and everyone in my age category. So yeah I give up. Its just another    situation.

[daedalux]

Never give up, just aim what you want to do in the good direction.

What I wanted to warn you is that with the voltages and kind of equipment you are working with you should setup a isolation transformer for power, fuses and a 15mA residual current breaker that you'll set in the isolated zone, if you can't find one at least put a 30mA one that you'll find anywhere.  I'm talking about your test set not what you design. This is very useful and needed for mains work, the transformer is usually multitapped which is useful and the breaker and fuses protect not only you but your test gigs and the parts you are working with.

[SK_Caterpilar_SK]

Never give up, just aim what you want to do in the good direction.

What I wanted to warn you is that with the voltages and kind of equipment you are working with you should setup a isolation transformer for power, fuses and a 15mA residual current breaker that you'll set in the isolated zone, if you can't find one at least put a 30mA one that you'll find anywhere.  I'm talking about your test set not what you design. This is very useful and needed for mains work, the transformer is usually multitapped which is useful and the breaker and fuses protect not only you but your test gigs and the parts you are working with.

My bench is isolated and has that breaker..I thought of that little earlier, not cheap to get that residual breaker but I have it. The bench has its own breaker, and right at the bench there are two more low amp quick breakers one is for the variac. and the variac on the secondary side of the isolating transformer is wired trought the GFCI or whats it called. The isolation transformer is not multitapped, I had it made specifically for me on order. I did not need multi taps because why would I and the voltage arround here does not fluctuate. Living right next to the transformer station. My equipment is old id say very retro, for high voltage measurements I use a VTVM, you cant just about destroy one of those. All other measurements were done with my OWON multimeter but the turning knob is dodgy and it falied and it was brand new. So now I use my old nixie bench meter, works good, im happy with it, just a little space consuming but very nicely readable.

But I might just not continue with the project I wanted to do, because of the negative feedback I got.

[TERRA Operative]

Nah, forget the haters, you have my interest and support.

I am planning on doing the sxact same thing as a friend wants to make a lightweight, ultra portable guitar amp for taking to gigs.
A heavy traditional amp is a bit hard to take on the Tokyo subway..... small and light is the key.

So far, I plan to build two supplies, one HV PSU and one filament PSU, so it is more modular for reuse in other projects.
I was thinking of adding a status output from the filament PSU that would go high once the filaments warmed up (current monitoring, time limit e5c) which would feed into an input on the HV PSU to only bring it online when the tubes are ready.
This will also allow for the usual 'standby' switch you se on a lot of amps to be able to be used still.

I'm just at the concept stage of my project, so I'm just starting my design at the moment.


Keep at it, and don't let the haters get you down.

[schmitt trigger]

@SK_Caterpillar_SK

Haters will always hate. Most so if they are shielded from you by the web's anonymity. Don't pay attention to them.

I once got flamed big time because I dared to publish the design of a switching filament regulator! Calling me ignorant was the tamest insult I received.

Your project looks very nice to me. Keep on experimenting, and be sure continue to post your results here.

The only suggestion I would provide you, is that after thoroughly debugging them, clean them very well with IPA and then proceed with a spray-on acrylic conformal coat.
That will prevent pesky humidity and dust from depositing across traces and creating leakage paths.

[Circlotron]

SK_Caterpillar_SK,
We actually understand what you are doing
We are with you all the way
The opinions of people that have no understanding of what you are doing are worthless.

[theworldbuilder]

Hey SK_Caterpillar_SK,

Just wanted to say that I'd be really interested in seeing this project progress.
I'd definitely be interested in a smps for tube applications.

I've been tinkering with my own SMPS for tubes for the same reason; I can't afford power transformers. But I've only ever built low power MC34063 based supplies which have tons of audio level buzz.
I'd love to see what you've come up with so far and to see you develop this into a fully functional system! Keep it up!


Never mind the audio snobs on the other board. I understand their wish to keep it transformer only, but it's just not practical. It's too expensive as you said.

[Doctorandus_P]

If you are intesting with non-mainstream audio amplifiers then consider a voltage to current conveyor instead of a voltage amplifier.
For more background info, read:

[strawberry]

LLC SMPS topology is the way to go in audio . you can find this kinda power supply on ebay aliexpress

[SK_Caterpilar_SK]

Hey SK_Caterpillar_SK,

Just wanted to say that I'd be really interested in seeing this project progress.....

Progress so far:

The noise is caused by the filament heater supply being a crappy chinese PSU. I bought a 5V 8A supply that I modified the feedback to give 6,3V exactly, and some movearround. The entire thing is a Flyback supply. Its putting out 56W at the voltage shown (measured the anode current round about 130mA). The switching transformer for the mains version arrived but I dont know if I will ever use it. I hit so many roadblocks and solved one by other and so far the biggest problem is isolating the output from the input because you dont want mains refferenced output which ground can be either live or not. A second less of a problem is having ground loops. But the main concern is safety which I cannot reach currently. I need to figure out how to isolate the ouput voltage with a opto isolator from the input, and also figure out how am I going to power the TL494 on the mains side, so far I have came up with a capacitive dropper, would work but not ideal. The biggest thing slowing me down right now is the isolation, a problem that would truly need help with someone who knows how that stuff actually works.

Yet still I get told to buy some ebay powersupplies just as on the other forum .

[SK_Caterpilar_SK]

So here is my problem. Opto feedback for the TL494. Really the only issue is how do I choose the values for the components. The original values are for 5Voutput voltage. I need it to work in the range of 400 to about 630V. Or I can just have them a fixed a fixed value and solder the components accordingly for the voltages. I would really appreaciate if someone would rather show me a way how to choose these values instead of just giving me the values,  if someone could tell me what values to use.

My idea was to simply lift the entire FB circuit with a resistor, so instead of grounding it directly I would put a resistor where the ground was originally and lift that entire circuit above ground. I just dont have an idea if that would work, and the most criticall of all is that its best to not load the output but very lightly. Small mA currents in a 5V supply dont matter but 1mA at 500V is HALF A WATT. Not like its a big deal considering the output power of the circuit but its there just generating heat.

Edit- the pics are fixed yes I noticed sorry my mistake

[TimNJ]

Popular ways to get a low-voltage Vcc rail for powering  your primary side ICs are...

1.) Add another winding to your transformer. Usually called 'aux winding' or similar. You'll still need some sort of "bootstrap" supply to get the thing up at startup.
2.) Use a small offline switcher. Power Integrations makes controllers with integrated MOSFETs. Those are nice.

Regarding the output error amplifier. There are rigorous ways to determine those values, and not-so-rigorous ways to determine them. Ideally, you'd like to know the gain/phase response of your entire system. Then, you apply compensation to your feedback network to ensure that the system is unconditionally stable, under all conditions. You can look up feedback network compensation with TL431.

You can basically change:
1.) DC Gain
2.) Mid-band gain
3.) Frequency of pole
4.) Frequency of zero

[SK_Caterpilar_SK]

Popular ways to get a low-voltage Vcc rail for powering  your primary side ICs are...

1.) Add another winding to your transformer. Usually called 'aux winding' or similar. You'll still need some sort of "bootstrap" supply to get the thing up at startup.
2.) Use a small offline switcher. Power Integrations makes controllers with integrated MOSFETs. Those are nice.

Regarding the output error amplifier. There are rigorous ways to determine those values, and not-so-rigorous ways to determine them. Ideally, you'd like to know the gain/phase response of your entire system. Then, you apply compensation to your feedback network to ensure that the system is unconditionally stable, under all conditions. You can look up feedback network compensation with TL431.

You can basically change:
1.) DC Gain
2.) Mid-band gain
3.) Frequency of pole
4.) Frequency of zero

Thanks for the reply. I forgot to mention that I dont have acess to custom wound switching transformers. I also have 0 idea how to calculate them, what core materials to use and everything. It is much easier to just get a part thats already available. I plan to power the controller with a capacitive divider, for now. Later I might go about creating a specific transformer for this purpose but I need bigger minds to help me do that. The transformer part is "Coilcraft HA4060".

I simply have no patience of custom winding the thing. This way I get known numbers of the switching transformers like working frequency and peak currents and all the other stuff, that I would have had to measure in the first place. Maybe in the future I might end up teaming up with some switching transformer manufacturer, but that wont be any time soon since I really just want one or two pieces, no company in their right mind would support that.

And I have no clue how am I going to solve the issue with the feedback. I dont know what values to choose and or how to do measurements on the original circuit so I can determine the values based on measurements..

[T3sl4co1l]

Yeah, lifting ground won't work because you need a change in LED current to transmit a signal, and that current has to flow out the ground, so the voltage sensed by the error amp changes with load...

What you can do, is make a little local supply, say 10V or so, with a zener shunt regulator, or an LR8 or something like that.  This supplies the error amp, which should be TLV431 (note the voltage limit!) for lower current consumption.  The "cathode" current can be cascoded into a pass transistor that drives the opto, or the whole thing supplied from the low voltage supply (either way, you need to handle the dissipation caused by the couple mA maximum consumption).

BTW, the opto symbol is upside down?

Tim

[SK_Caterpilar_SK]

Yeah, lifting ground won't work because you need a change in LED current to transmit a signal, and that current has to flow out the ground, so the voltage sensed by the error amp changes with load...

What you can do, is make a little local supply, say 10V or so, with a zener shunt regulator, or an LR8 or something like that.  This supplies the error amp, which should be TLV431 (note the voltage limit!) for lower current consumption.  The "cathode" current can be cascoded into a pass transistor that drives the opto, or the whole thing supplied from the low voltage supply (either way, you need to handle the dissipation caused by the couple mA maximum consumption).

BTW, the opto symbol is upside down?

Tim

Thank you Tim, in the ideal world I would have had a tap at the tranformer for the 431 and the opto which would have been propbably the best but overkill solution. If I would have had a custom tranformer that would have another secondary for the tube filaments (6,3V) then I would have used that. But I think this might be the best solution for my current setup. Simply just a shunt regulator.. now just to measure the current and choose the proper value for R6.

[T3sl4co1l]

Yeah, heater winding would do perfectly... ah well.

That's why I added positions for an aux winding on this general-purpose power supply board: https://www.seventransistorlabs.com/Images/100WPowerSupplies.jpg they've mostly gone unused, but are there for when it's needed.   (I did make one, a 150-300V 100W model, which uses this aux secondary winding.)

Also what's nice about custom transformers, and having a bunch of bobbins in stock.

Tim

[SK_Caterpilar_SK]

Yeah, heater winding would do perfectly... ah well.

That's why I added positions for an aux winding on this general-purpose power supply board: https://www.seventransistorlabs.com/Images/100WPowerSupplies.jpg they've mostly gone unused, but are there for when it's needed.   (I did make one, a 150-300V 100W model, which uses this aux secondary winding.)

Also what's nice about custom transformers, and having a bunch of bobbins in stock.

Tim

That looks hella professional. The big deal with the heater winding if there would be any, is that for an average amplifier like mine for example mine, two power tubes and two pream tubes, It requires at least 4A of output current with absolutely no headroom, 3A for the power tubes and 900mA for both preamp tubes. So id say a 30W winding would be ideal for an average amp (6,3V 5A). Maybe sometime in the future if I get a chance to work on the transformer with a professional SMPS transformer manufacturer. But the big deal is I got no idea how to calculate the transformers at all. I may ask Coilcraft if they would be willing to help out but I doubt any of that will happen.

So I guess its only left to do measurements and go ahead build the thing. for the FB. I just hope it will be sub 5mA cause then I can simplify things and not waste incredible amounths of power lol.

[TimNJ]

I think you are smart to not inflate the scope of your project by using an off-the-shelf transformer. My experience in designing SMPS transformers is fairly limited, but I can tell you a few things.

1.) Getting a first “working” prototype is not that hard, even if you have little experience. But, optimizing the transformer can be quite a thought experiment, forcing you to think in terms of magnetic/electric fields, circulating currents and so on.

2.) Hand winding a transformer is not hard at all! Getting your hands on all the supplies can be a bit challenging, but these days, you can buy sample kits on eBay inexpensively. For example, you might want various diameters of enameled wire, litz wire, triple insulated wire (TIW), TIW litz, poly tape (the yellow stuff), Teflon tube etc.

It’s probably not worth the trouble at the moment. But sometime in the future, it may be worth it to give custom a shot.

[SK_Caterpilar_SK]

I think you are smart to not inflate the scope of your project by using an off-the-shelf transformer. My experience in designing SMPS transformers is fairly limited, but I can tell you a few things.

1.) Getting a first “working” prototype is not that hard, even if you have little experience. But, optimizing the transformer can be quite a thought experiment, forcing you to think in terms of magnetic/electric fields, circulating currents and so on.

2.) Hand winding a transformer is not hard at all! Getting your hands on all the supplies can be a bit challenging, but these days, you can buy sample kits on eBay inexpensively. For example, you might want various diameters of enameled wire, litz wire, triple insulated wire (TIW), TIW litz, poly tape (the yellow stuff), Teflon tube etc.

It’s probably not worth the trouble at the moment. But sometime in the future, it may be worth it to give custom a shot.

Very well said, to say my first ever HV experiment was a hand wound transformer and it was working better than expected but it definitely wasnt powerfull enought- 10W is not really useable for anything other than preamps and the ferrite was not glued and it was screming soo..but it worked. I much rather use the off the shelf stuff because the specifications are already there measured and I only work with those value instead of guesstimating and I really cant calculate that stuff, wire width inductances coupling coeficient and other stuff. So I did make a hand wound one but it was less than ideal. And also the enameled wires are not cheap. Their maybe look cheap for you but for some reason all the wires I would like to buy have either crazy postage and or crazy price.

About the overinflation- its MUCH cheaper for me to buy a switching transformer that was designed more or less for the application im using it at. And a whole lot simpler. The first transformer I got I payed for and then now looking at the transformers I realised I can ask coilcraft for samples so I did and hopefully, the transformers will arrive so I can have test PCBs to experiment with.

[T3sl4co1l]

XFMRS may be more likely to prototype small quantities, or others that I've since forgotten; and also, any local motor rewind shop will likely have the tools and materials to do the same.

Don't expect an amazing design from any of these sources, though -- to get that, you'll have to supply at least a basic drawing, preferably a drawing accurate down to the last turn, and designed based on transmission line transformer principles, and, probably, based on a lot of experience too.

Custom voltages don't preclude the use of commercial parts, but it does severely limit the choice and performance.  For multiple outputs, consider multiple single-output transformers wired in parallel (parallel primaries, of course).  Any extra aux windings, of course, just leave open.  (By the way, don't be tempted to use one as secondary-side aux: the pri-aux insulation is usually not full reinforced.  If the datasheet says it is reinforced, though, that's fair game!)

Tim

[SK_Caterpilar_SK]

XFMRS may be more likely to prototype small quantities, or others that I've since forgotten; and also, any local motor rewind shop will likely have the tools and materials to do the same.

Don't expect an amazing design from any of these sources, though -- to get that, you'll have to supply at least a basic drawing, preferably a drawing accurate down to the last turn, and designed based on transmission line transformer principles, and, probably, based on a lot of experience too.

Custom voltages don't preclude the use of commercial parts, but it does severely limit the choice and performance.  For multiple outputs, consider multiple single-output transformers wired in parallel (parallel primaries, of course).  Any extra aux windings, of course, just leave open.  (By the way, don't be tempted to use one as secondary-side aux: the pri-aux insulation is usually not full reinforced.  If the datasheet says it is reinforced, though, that's fair game!)

Tim

Yeah about the aux windings- no I only use the transofmers as specified in the datasheet. Now that im thinking about it, a all in one powersupply would be the ideal thing to do. 6,3V 35W for heaters, 400-600V for anodes about 80W, and another aux that would deliver almost no current , in some tube amps there is a -50V supply for biasing the output tubes. In cathode biased circuits you only need the heater and the HV, but cathode biased amps have their drawbacks, I would definitely want to try and make a fixed bias amp at some point. So yeah 6,3V 35W, -50V 5W, 400-600V anode voltage at about 80W of power would be ideal, a replacement compatible with most common amplifiers.

I would like to do that but I have no idea how to calculate SMPS transformers at all and I cant really find literature that would explain it to me at the level of my understanding. But I will very likely do something like that as a grand finale for the PSU.

[SK_Caterpilar_SK]

Since I found that blowing with a fan on my original HV PSU was UNACCEPTABLE I went and redesigned the whole thing. This time using the TL494 as the controller. Operating frequiency will be set to 50kHz. Also have a soft start function, NO OPTO ISOLATION YET. Im still working on the mains powered solution.

So basically I want to listen to my amplifier while I work on its powersupply some more without that fan blowing on the PCB lol. The reason the fan has to blow on the PCB is because i made the traces to thin (2mm) for the required power about 60W. It would require at least a 3,5mm trace with...but not to worry this time I bombproofed the thing and went 6mm trace width where I could and massive ground and power planes. And my pull is to tin it with liquid tin or the traditional handsoldering, depends how fast I can get it done

The TL494 should give a way more stable output that I could ever get with my hacked together quad op amp controller I designed previously.

The TL494 also extends the operating voltage up to 40V. About the efficiency- I have no idea what to expect so far, I will find out after loading and measuring the characteristics of the powersupply.

(The board layout is not done completely.)

EDIT: The IGBT will be mounted vertically to a heatsing not to the PCB (its also related to the not completeness of the board..)