Using Nixie tube DC high voltage power supply for tube based pedal builds?

[Marcos - Munky]

Got it. Well, part of it . I understood the theory but don't know how to change Ton and frequency.

So, assuming Ton is 17us and f is 31kHz, just to follow from your maths. If I half the inductance, I'd have:
Ipk = 1.87*2 = 3.74A
E = (1/2) * 3.74^2 * 50uH = 349.69uJ
P = E * f = 10.84W, so 5.42W assuming 50% efficiency, and that makes around 27mA @ 200V

That's a nice high voltage current.

For voltages, I think a range like 200V to 250V is more than enough for lots of stuff. Some power amps would work better on higher voltages, but the current won't be enough.

So, another question. I assume Ton, Toff and f can be changed by changing values on the circuit, but I also assume they're very limited to the 555 itself. Would the UC2842 or MC34063 be a better choice?

[duck_arse]

the 34063 has a timing cap, but to use it correctly, you should go thu all the design formulas in the datasheet.

my method with unknown junked ferrites - wind on some turns 15, 20, whatever fits neatly. then using a transistor L//C oscillator on the breadboard and a few values of C, I get the operating f from the oscilloscope, and work the freq formula to get inductance. then use the inductor equations to get the A_L for that core, the turns-per-inductance number. then it's easy to work how many turns for how much inductance, then back into the test osc, rinse, repeat.

not terribly accurate, very long winded, pages and pages of scribble, but what else would I do with my time?

[Marcos - Munky]

Seems complicated. And I'm kinda like PinkJimi when it comes to theory

I have a UC2842 board based on a Transmogrifox design, which according to him it switches at around 50kHz. I'll see if I can measure the max current it can handle with the inductor I've used, then will try to make my own and see if there's any improvement. If I get improvement, then I'll try your way.

I'm working on 4-5 projects that I'm delaying to finish, if I start to read on L/C oscillators and equations I'll probably do the tests in 2021 lol.

[Rob Strand]

So, another question. I assume Ton, Toff and f can be changed by changing values on the circuit, but I also assume they're very limited to the 555 itself.

Upfront I should mention when you mess with these things there's plenty of opportunity for things to go wrong.  The current design is not optimized and it was set-up so you can adjust the output voltage and it probably works OK with no load.   By optimizing the circuit it might not work well at low voltages (or at low currents).  In fact it could cook something.    At high output voltages it should work fine.   Just keep that in mind.

FYI, the calculations I gave only work for high voltages.   At low voltages you get a bit more power output than the calculations show.



It's easy to adjust Ton and Toff using the 555.  The thing is what to set these to get best performance - that's not that straight forward. 

On the current design Toff is quite long.  This lets you adjust the output to low voltages.  If you give that up to some degree you can reduce Toff and that will make frequency go up without changing Ton, which will give you a little more output power - pretty much for free.   The long Toff means you can ignore how fast the MOSFET takes to turn off.  If you make Toff too small you will end-up with trouble.

Under light load you might find Ton gets reduced to a small value and Toff is the thing that sets the frequency.  It depends on the losses.   So while making Toff shorter will help get more output power it might also mean you need a larger minimum load.

One important thing about the NE555 design is the *only* thing limiting the current is the on-time Ton.   You need Ton long enough that the inductor can built-up to full current so the inductor is fully utilized *but* you don't want to push it too close as that will surely cause something to burn out.    True SMPS IC's usually have a specific current sense to limit the current so you don't have to be so fussy about getting Ton right.

So how to set the values: 

Ton is set-up using R1 *and* R2.   Toff is set-up using only R2.  C affects both but keeps the ratio of Toff and Ton the same.

Ton  = (R1 + R2) C  * ln(2) = (R1 + R2) C * 0.69
Toff = R2 C ln(2) = R2 C * 0.69

Ton is not fixed because the voltage regulation look adjust pin 5 which allows Ton to be reduced from the maximum value.   Also, on your schematic in reply#2,  resistor R1 (56k) on pin 5 stretches the maximum Ton a bit from the common NE555 formulas, perhaps 4% longer.   Toff is fixed and largely unaffected by pin 5.

When regulating, the design may drop the pin 5 voltage to an unusually low value and it might disagree with the calculated values.    With that in mind it would be wise to measure before and after timing changes

For the current design R1 = 1k, R2 = 10k, C=2.2n, so
Ton_max = 16.8us  or Ton_max = 17.5us after 4% adjustment ,   
Toff = 15.2us.
Tmax = Ton + Toff = 16.8us + 15.2us  = 32us    ; switching period
fmin = 1/Tmax = 31kHz                                          ; switching frequency
fmax = 1/Toff = 66kHz

Suppose we reduce Toff to 7.5us (I don't know if that is too small)
We calculate R2 = 4.9k, use say 4.7k giving Toff = 7.2us.
Set R1 = 1k + (10k-4.9k) = 6.1k,  to keep Ton the same (use 5k6).
Now,
Tmax = 16.8us + 7.2us = 24us
fmin = 1/Tmax = 41.7kHz
fmax = 1/Toff =  138kHz         ; pretty darn high and no doubt the losses go up with light loads.

These are no means a "best" design.   In your case you have the option of playing with the inductor winding and the NE555 timing.   To find the magic combination that gives maximum output power for a given output voltage isn't straight forward.  You would have to play around with the parameters.

Would the UC2842 or MC34063 be a better choice?

The MC34063 with an active pull-down will probably be better than the NE555.   The UC2842 and many others in that UC28xx and UC38xx series are probably better still but I think they only work down to 12V.  They have low voltage shut off etc.
The thing about the purpose built chips is they generally have support for current limiting.  The output stages are designed to drive MOSFETs.    The control loop is designed to work correctly,  the NE555 works but it's really pushing the device.  That's not necessarily bad but I does mean the you need to measure stuff because it is likely the calculated timings will be a bit off.    In short the purpose-built devices will handle higher switching frequencies.

Whatever you use the whole process of optimizing the design to get maximum output power or maximum efficiency is definitely not simple.   That paper I posted before shows a lot of calculations and you can see that in many cases, despite all those calculations, it still doesn't match-up with reality.

[jonny.reckless]

I've used these cheap boost converters from eBay for several tube guitar preamps:
https://www.ebay.com/itm/DC-DC-Boost-Converter-8-32V-12V-to-45V-390V-High-Voltage-Capacitor-Charging/133442139391?hash=item1f11c574ff:g:wdQAAOSwocFe6yAj

You see them from several sellers but they are all the same manufacturer as far as I can tell. Some have bipolar output which might be useful if you need to go above 350V HT for some circuits.

I just add a bit of smoothing with a 1k series resistor and 10uF cap to remove some of the residual switching noise. There is about 2Vpp of 60kHz sawtooth (with light loading, maybe 10mA) on the output which is easily removed. With the resistor and cap this goes down to about 10 - 20 mVpp. The ripple gets a lot worse under heavier loading.

It's cheap, compact, efficient and works fine for low power use. I've also built a couple of 2 watt tube output stages from the same module. I believe you could probably reliably get 5 watts out using it with a pair of EL84s. The PSU seller quotes 40W output but I don't believe that, I blew one up with 20W @ 250VDC load. Made a hell of a bang, blew a hole in the MOSFET and let out the magic smoke. The unipolar version is less than $5 each in quantity so I don't bother building my own HT power supplies  Use a cheap laptop power supply for the input side and you're good to go.

[Marcos - Munky]

Rob, got it, I think lol. So, if I got it correctly, we can say that every load have an optimal setup for best performance. So a bench-like adjustable 555 charge pump isn't possible, because what works best for a low-ish voltage doesn't works best for a high voltage, the same goes for current.

For the UC28XX and UC38XX series, they have a max input supply of 30V, but the thing is they do require a turn-on voltage and do have a minimal voltage to work. The start threshold is higher than minimal voltage to work. I just copied the UC2842 from Transmogrifox's schematic, and now I recall what really happens. The X842 and X844 have a minimal voltage of 9V to 11.5V to work after the turn-on, but it requires 15V to 17.5V to start working. So they can't be powered by let's say a 12V power supply, because you don't have enough voltage to turn them on, but you can use a laptop power supply to power them. While Transmogrifox was answering me some questions, he noticed that and checked the datasheet again. The X843 and X845 have a start threshold of 7.8V to 9V, and then requires 7V to 8.2V to keep working, so they can be used with a 9V/12V power supply. I tested both UC2842 (laptop power supply) and UC2843 (9V power supply).

Jonny, those modules seems nice enough. For $5, I woudn't think twice and order a bunch, but for the link I found the seller doesn't ship to Brazil. What's the mosfet on this one?

[Rob Strand]

So, if I got it correctly, we can say that every load have an optimal setup for best performance. So a bench-like adjustable 555 charge pump isn't possible, because what works best for a low-ish voltage doesn't works best for a high voltage, the same goes for current.

That's pretty much it.   It's not that you can't get adjustable voltages, it's more there is a penalty of poor performance.

While Transmogrifox was answering me some questions, he noticed that and checked the datasheet again. The X843 and X845 have a start threshold of 7.8V to 9V, and then requires 7V to 8.2V to keep working, so they can be used with a 9V/12V power supply. I tested both UC2842 (laptop power supply) and UC2843 (9V power supply).

Well spotted, so the UC2843 might be the one.  I skimmed over the datasheet again and the UCx845's have a duty cycle limit.    There's actually a lot of chips in that series.

[Marcos - Munky]

Question before I make the inductor. Do I need to left any part of the rod without windings, or can I wind the wires from one end to the other?

I mean, on the calculator page, there's this: "Winding length should be less than 3/4 length of the rod;" as a "calculating limitation. Is it a limitation of the calculator itself, or is this how I should wind the wires?

[Rob Strand]

Question before I make the inductor. Do I need to left any part of the rod without windings, or can I wind the wires from one end to the other?

I mean, on the calculator page, there's this: "Winding length should be less than 3/4 length of the rod;" as a "calculating limitation. Is it a limitation of the calculator itself, or is this how I should wind the wires?

It's a limitation of the calculator itself.   

The calculations for rods inductors are quite inaccurate.   People make measurements of a few coils then try to come up with formulas that match the measurements.   It's hard to get enough measurements to cover all the rod and coil shapes and after that it's hard to come-up with a formula that matches.  If you take one guys measurements and plug them into another guys formula they often don't match  - I've had a few attempts at this myself.

It's best to measure the inductor, there's plenty of methods.  The one duck_arse mentioned works well.

It's not uncommon to wind the coil to the ends of the rod.   There might be some advantage leaving each end of the rod unwound, say one diameter or at least half a diameter.    The calculator restriction of the coil length being 3/4 the length of the rod isn't a good thing to follow if you have to squash-up the coil.   It would be different if you cut the ferrite to match the coil.

[Rob Strand]

Couple of other things that came to mind are:

It's not recommended using more than two layers at high frequencies.

IIRC: When the rod length is close to the coil length the saturation current of the coil will be a less than my formula predicts.   However that doesn't mean you should wind over a shorter length in the case where the rod length is fixed (as is your case).  In this case you will find nothing improves.  In fact all you do is increase the coil resistance because you will need to use a thinner wire to fit the coil over a shorter portion of the rod.

[Marcos - Munky]

Too late I've already made it... using three layers and winding up to almost the entire length of the rod. . So it's a very poor made inductor.

So, I tested it on those cheap parts testers. The resistance is 0.4r, less than the drum inductor (which is 0.5r). For the inductance, I got 70uH. Later I discovered I made a(nother) mistake while using the calculator.

Well, let's see how this one goes with all those mistakes . I left my soldering iron at work, so hopefully I can test the inductor tomorrow. It will be excellent if I can get more current using this inductor, but to tell the truth I'll be happy if it at least works and I can get current enough to power one or two 12AX7. Inductors are very hard to find here so, if this one works, I'll have a spare inductor to use to build one more smps.

[Rob Strand]

Too late I've already made it... using three layers and winding up to almost the entire length of the rod. . So it's a very poor made inductor.

Not necessarily poor.  The inductors' AC losses rise quite quickly with the number of layers.   Inductors are a little weird because the resistance at high frequencies is higher than the resistance at DC.      Going to two layers adds a small amount of loss but with three and four layers it can increase quite a bit.   Letting the rod extend past the coils also helps reduce the AC losses.    It depends on the wire size and there's more messy and inaccurate calculations for all that.   It might not even be a problem.

Test it as is.     It would be best to put a reasonable load on when you power it up (maybe 0.5W, 2.5mA @ 200V, RL= 80k) .  I don't know how it will behave with light loads.  Set the output voltage and measure the input current.   Write those values down.   After that do the same thing with a few different loads.   Write the measurements down somewhere.  If you make changes then you can see if they are step in the right direction.   

You can always unwind some windings.

[Marcos - Munky]

Just to be clear, should I use resistors as the load? If so, what's the suggested wattage for them? For a 0.5W as you suggested, I believe I should use at least 1W resistors to be on the safe side, correct?

[Rob Strand]

Just to be clear, should I use resistors as the load? If so, what's the suggested wattage for them? For a 0.5W as you suggested, I believe I should use at least 1W resistors to be on the safe side, correct?

It's a good idea to use the high rating 1W resistor.   

You can put resistors in parallel or series to get more power.   The parallel method is easier to work out the power in each resistor when the values are different.   Using more than one resistor lets you play around with the load.

[Marcos - Munky]

So, today I got some free time to test the UC2843 smps and the inductor I bought. Didn't tested the one I made.

This is the original smps. Mine is way more simplified.
http://www.cackleberrypines.net/transmogrifox/BoostConverter/250V_SMPS.png

The output voltage was set to 220V using fixed value resistors instead of a trimpot. Pout was calculatted assuming the output voltage didn't dropped as the current raised. Here are the results:

Wow, I wasn't expecting it to go past 20mA, and was very surprised when I got 41mA and nothing exploded!

On the efficiency, it is really is better with a heavier load. While this isn't an awesome efficient circuit, I can live with that numbers. But since Pout was calculated assuming there was a constant 220V voltage output (no voltage drop), those values on the table are actually a bit wrong.

Since the output voltage was supposed to be fixed because I used fixed value resistors, I didn't measured it, I just assumed it was 220V all the time. From P=R*i², I got P=8.92W for the last row, confirming the real output power is indeed different from the ones on the table. From P=V*i, I got V=212.8V, so there was indeed a bit of voltage drop. Nothing to really worry about.

[Rob Strand]

time to test the UC2843 smps and the inductor I bought.

Having an inductor which can handle the current is a big help.

Wow, I wasn't expecting it to go past 20mA, and was very surprised when I got 41mA and nothing exploded!

   Looks good.

On the efficiency, it is really is better with a heavier load. While this isn't an awesome efficient circuit, I can live with that numbers.

It's normal for the efficiency to rise but eventually it drops off; like the graph in Reply #10.   Your efficiency is still going up but that doesn't mean something won't smoke with more load.

At full load you are dissipating 14.06W  - 9.02W = 5.04W.   You should be able to find which part is getting warm at this level.   The MOSFET is a likely cause.   The inductor might get a little warm if it's a bobbin/drum type. 

From P=V*i, I got V=212.8V, so the;re was indeed a bit of voltage drop. Nothing to really worry about.

Yes the regulation looks pretty good.   V=212.8 looks about right.    I cross-checked the output voltage using V = IR.  Since the R is known we can estimated Vguess = 220V * (Imeasured / Iexpected).   There's only two decimal points in the current measurements which limits the accuracy.    Using the currents in the last two entries of your table, I calculated Vguess and took the average and I got a similar number to your 212.8V.

[marcelomd]

Marcos, your measurements look great so far. Do you have access to an oscilloscope, or something, to check output ripple?

After my last post here I spent a few hours simulating circuits with UCx84x controllers and I either get too low voltages or huge oscillations on the output.

This is my best try:


My goal was 250V, 50mA, at +-100kHz. This looks like it'd work. I'm guessing your design is similar. Switching frequency is 110kHz. Inductor current is limited to 2.5A.

Looking at the voltage across the switch, maybe there's potential for sub harmonic noise.

1- Seems like everyone has a different method for calculating components. Using the same set of specifications (250V, 50mA, 100kHz) I got wildly different inductor values, from 100uH to 470uH depending on the application note/datasheet formulas.

2- 12->250V seems like too big of a boost for a single stage. The switch is at full duty cycle all the time (only off for the mininum deadtime set by the timing capacitor). Someone correct me, but in this mode, this is working pretty much the same as the 555 circuit plus current limit.

2.1- Going 12->125V or lower and using a voltage doubler/tripler looks like a viable alternative. Any downsides?


3- This implementation is from Blackstar's HT distortion. It uses voltage mode control:


Valeu!

[Marcos - Munky]

Marcos, your measurements look great so far. Do you have access to an oscilloscope, or something, to check output ripple?

Well, I have one of those cheap pocket oscilloscopes from Aliexpress. If it can be used, then I can check the output ripple. I just need to be told what to do, since I know nothing on how to use an scope.

Transmogrifox's original circuit (http://www.cackleberrypines.net/transmogrifox/BoostConverter/250V_SMPS.png) was designed to have switching frequency at 50kHz. Mine is a stripped down version of his:

Basically, I removed any protection circuit and some other suff lol. The goal at the time was to build a small as possible smps to power tubes and fit small enclosures. Also, since my output cap bank is very small, compared to the original design, the output will probably have some ripple. I've used this smps to power a few tube circuits, one I can remember I surely used it was a rectifier preamp with a class D power amp, both power amp and smps powered by a laptop power supply. It was surprisingly quiet.

[Rob Strand]

After my last post here I spent a few hours simulating circuits with UCx84x controllers and I either get too low voltages or huge oscillations on the output.

Sometimes you have to be careful that things aren't caused by spice.   Some tips to help things are:  put a load on the circuit.   Using the initial conditions set the output capacitor voltage to be at a reasonable voltage, perhaps just under what you expect.  If you set a high voltage and light load you can cause the feedback to shut off.   If there's more than one cap you need to set them all up.  If you get the sim settle down at the end of the spice run, measure the final cap voltages and then use them for the initial voltages - that gives spice a head-start and you don't have to wait for it to settle each time.   Note also for the cap initial voltages you need to get the sign right.

Seems like everyone has a different method for calculating components.

Some calculators are based on continuous conduction mode (CCM). This is where the current in the inductor never gets to zero.   This is not recommended for 555 designs and many other designs.   Simple designs use discontinuous conduction mode (DCM).   The DCM design are much easier to stabilize the feedback control and you will get it working without frying parts much more easily.

12->250V seems like too big of a boost for a single stage. The switch is at full duty cycle all the time (only off for the mininum deadtime set by the timing capacitor). Someone correct me, but in this mode, this is working pretty much the same as the 555 circuit plus current limit.

It's pretty normal for it to work like that.    Yes, it's quite close to the design concept of the NE555 circuit except using proper chips helps.  The current limit also helps because you aren't relying on limiting Ton to set the limit the current.  It will also handle light loads a little more gracefuly.

The penalty in using a big voltage step is the circuit is less efficient.   If the losses are too high we can get poor efficiency and low output power.    One issue is the MOSFET: it needs to be rated for the current of the input but the voltage of the output.   High voltage MOSFETs inherently have higher on resistance Rds_on.  So the high voltage step is rigging the whole circuit for poor efficiency.     You can see some more discussion here,

https://www.analog.com/media/en/technical-documentation/application-notes/AN-1126.pdf

The reason you see the large step booster is because it is a simple design with a low parts count.   It doesn't require any special or custom parts.   You just throw a big enough MOSFET and inductor at it and it works.   All off the shelf inductors, so no inductors to design.  The penalty is the efficiency can be poor.  Some care can improve things.   There's plenty of options to improve efficiency.   Designs with 90%+ efficiency are possible.  The thing is these are very complicated designs and use a lot of parts, many mosfets, and often custom inductors or transformers.    If you can ignore efficiency it simplifies things a lot.

Another point is coming up with a design capable of 20W is unlikely to perform that well at 2W.   SMPS design is the ultimate in forcing you to pin down what you really want and making trade-offs along the way.

If you look at Marcos's original design and try to get the most output with an inductor you already have on hand, it's quite a different problem than designing from scratch.    That involves tweaking the circuit to get the best performance without changing any major parts.

[Rob Strand]

Well, I have one of those cheap pocket oscilloscopes from Aliexpress. If it can be used, then I can check the output ripple. I just need to be told what to do, since I know nothing on how to use an scope.

You really need x100 or x10 probes to make sure you don't fry the front-end of your Oscilloscope.
Double check all the probe settings and Oscilloscope settings before connecting.

I don't have a lot of faith in the input protection of simpler oscilloscopes.