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

[marcelomd]

Marcos. Just measure the output voltage and check for noise/oscillation. But do what Rob said. Use a 100x or 10x probe. Or, veeeery carefully, make a voltage divider so you get a level that won't ruin your scope.

Rob. That is a lot to digest. I was under the impression that CCM was simpler. Peak currents being lower (if I got that right).

My end goal is a generic PSU module I can use to play with tubes. But I'm only playing with the idea for now.

[Rob Strand]

That is a lot to digest. I was under the impression that CCM was simpler. Peak currents being lower (if I got that right).

It's definitely easier to understand switch-modes when you look at CCM.   DCM is much more messy.   However when it comes to building a device DCM is much more friendly (even more so with NE555 designs).

CCM does have lower peak currents however, depending on details, the smallest inductor for a given power occurs around the boundary of CCM and DCM.    Think about it like this: CCM needs a larger inductance in order for the current to not fall to zero.   That means more turns on the inductor.    More turns means more field for the same current.  So on one hand you have lower peak current but on the other you have more field, so it's a matter of who wins as far as the inductor is concerned.

On CCM there is a smaller peak to peak (AC) current swing and smaller peak current than DCM.   That results in two things.  lower ripple current on the input supply and lower changes in the field inside the inductor.     Both those features help reduce interference.   The lower AC field of CCM helps reduce losses in the core and the windings.   For those reasons higher power designs tend to use CCM.   Lower power designs you have the option of either CCM or DCM.

The problem with CCM is it brings up feedback control issues and also issues with sub-harmonics.    Some of the finer features in the SMPS controller chips help with these problems.

There's so many details in switch mode design, lots of problems and counter-measures.  It's very hard to tap into the subject and most of papers don't tell the whole story.   If I don't look at it for a while I start to forget the finer points.

My end goal is a generic PSU module I can use to play with tubes. But I'm only playing with the idea for now.

Something like Marcos's design should cover a lot of bases.    Trying to design "the best" switch-mode is going to be a difficult task.   You can tweak things to get the most efficiency or get the most out of an inductor.

FYI, those bobbin/drum inductors are cheap an readily available but they often don't give the best results.   They can also radiate a lot of intererence.  Shielded types help stop the interference.   You can get better performance from, gapped cores like this,
https://neosid.com.au/images/RM-Title-pic.jpg

However, the MOSFET is probably the weakest link in a large step-up so you might not see enormous improvements with a better inductor.

[Marcos - Munky]

My scope have a input of 50V pp max, and no attenuated probe. So I should go for a voltage divider. How should I connect it? Smps to load and the voltage divider in parallel with the load?

Btw, it isn't my design, but Transmogrifox's. I don't have the knowledge to even begin to understand a bit of what's going on lol. Transmogrifox is the man behind that awesome design.

And the drum inductor really radiate some interference. I noticed on a few circuits the position of the smps board related to the circuit board changed how much noise was induced. Probably a Faraday's cage around the smps will solve this issue.

[Rob Strand]

Smps to load and the voltage divider in parallel with the load?

That should work fine.

[marcelomd]

Well. In a past life I developed device drivers for telecom equipment and I worked closely with the hardware engineers. They spent 80% of their time on the power supplies. Now I know why =)

In an ideal world, the end product here would be a PSU module. 9-24V in, 150-250V, 50mA and 6.3V, 2A out. The numbers come from 8 3.5mA anodes and 4 450mA filaments. Which is maybe an exaggeration, but will cover everything I will ever want to try.

[Rob Strand]

In a past life I developed device drivers for telecom equipment and I worked closely with the hardware engineers.

Writing device drivers has a lot of similarities in that it requires a lot of obscure knowledge that often isn't put in the one place.   Back in the day you could pay for one-on-one sessions with Microsoft consultants to fill in the gaps.

[marcelomd]

I actually started as a hardware engineer and slowly moved to the software side. This week I realized I spent more time with javascript frontend frameworks than programming. What a sad state of affairs...

Sad like this oscillation here, which appears only in a range of output voltages:

[Marcos - Munky]

I can try to measure the ripple on my smps on saturday. But I have two questions before doing it.
1- what should be the power rates of the resistors for the voltage divider? I assume if I use big values for the divider (>1M) I can use 1/4W resistors, since the load resistors are smaller. Is that correct?
2- should I measure the ripple for a fixed load/current, or for a range of loads/currents?

[Rob Strand]

Sad like this oscillation here, which appears only in a range of output voltages:

The problem is subharmonics.    That's caused by current-mode controller + CCM + duty > 50%.
If for example you had an output transformer (actually a split inductor) you could choose the winding ratio to give 125V out yet still have < 50% duty.   With a non-isolated boost converter you are going to be running with
high duty cycle so that's going to cause problems for CCM.

We need to changed to  DCM by using a smaller inductor.

Suppose we keep f = 112kHz and Imax=2.5A

Choose say L = 33uH to give DCM  and still give output power.
For Vout = 125V & 2.5A pk, need to ensure inductor fully discharges to maintain DCM so Toff > L dI / V  = 660ns

Currently the design has a small td.   The timing cap controls the td, so we increase the cap
and decrease the timing resistor to keep the frequency the same.

Choose Ct = 3.3n to set td > 660nS. Rt = 4.7k  to give about 100kHz.
Then from formulas in the datasheet.
td = 820ns
tr = 9.0us
Tosc = td + tr = 9.82us
fosc = 101kHz

Next step would be to tune R6 and C6 for stability and best dynamic response.

None of the above is optimal.  I've only kicked the tires a bit to get DCM mode and (hopefully) stop the sub-harmonic oscillations.

I actually started as a hardware engineer and slowly moved to the software side. This week I realized I spent more time with javascript frontend frameworks than programming. What a sad state of affairs...

Overall hardware has moved in a similar direction.   It's more like lego now.

[Rob Strand]

I can try to measure the ripple on my smps on saturday. But I have two questions before doing it.
1- what should be the power rates of the resistors for the voltage divider? I assume if I use big values for the divider (>1M) I can use 1/4W resistors, since the load resistors are smaller. Is that correct?
2- should I measure the ripple for a fixed load/current, or for a range of loads/currents?

Large value are OK.   If you use 1M and 125k, then with the extra 1M from the oscilloscope you will get a 1/10 divider.    Use a 120k.

It's a good idea to measure the ripple at full-load, half-load.  Also check for weird output at minimum load (oscillations and pulsing).  If you see weird outputs then increase the minimum load until it goes away.

[Rob Strand]

That's caused by current-mode controller + CCM + duty > 50%.

The sub-harmonics problem is mentioned in this video,

https://www.youtube.com/watch?v=gIZW6jJe8Tc

The PDF application note for the same video is,
https://pdfserv.maximintegrated.com/en/an/AN6609.pdf

See the part on current-mode control

Forgot to mentioned, there's a thing called slope-compensation which is a counter-measure for sub-harmonic oscillations.   Some controllers have it, others require you to add more parts.   It affects the small-signal stability a bit.

[Marcos - Munky]

Large value are OK.   If you use 1M and 125k, then with the extra 1M from the oscilloscope you will get a 1/10 divider.    Use a 120k.

That's something new to me. I was gonna do 1M and 1M, or something like that.

It's a good idea to measure the ripple at full-load, half-load.  Also check for weird output at minimum load (oscillations and pulsing).  If you see weird outputs then increase the minimum load until it goes away.

I'll try to do a few measures then. I'll start with something like 1mA and see how it goes, then increase the load.

[Rob Strand]

That's something new to me. I was gonna do 1M and 1M, or something like that.

The idea is if you connect to 200V 400V and it drops it down to 20V to 40V (< 50V CRO limit).

One oversight on my part is we are dealing with high-ish frequencies.

The capacitance of the oscilloscope input, typically 20pF to 30pF, will act like a low-pass filter.    The 1M + 120k + 1M oscilloscope impedance works out to be R = 97k ohm,  so the 20pF to 30pF is going to low-pass at 55kHz to 82kHz depending on the capacitance.    Just enough to have the potential to stuff-up the measurements.

The way that is solved is by adding a cap across the 1M divider resistor.   A rough value would be 2.2pF (yes very small).    Oscilloscope probes have this cap built-in.       The cap needs to be rated at a voltage greater than the maximum DC voltage.    You can tune the capacitor so a square-wave (often the "cal" waveform on the oscilloscope) has a nice flat top.    If you don't have a small cap like that, get two solid core wires and twist them together.   Tune the cap by chopping the open end of the wire.  When you chop disconnect the source as the cutters will short out the capacitor.

http://akavalve.blogspot.com/2009/05/thats-not-capacitor-thats-just-two.html


EDIT:
This gives you an idea of the length of wire,
https://en.wikipedia.org/wiki/Gimmick_capacitor

[Marcos - Munky]

I don't know if I have a capacitor that small. I do believe I have a 4.7pF cap, but with some wild tolerance. That will work?

So, to be clear: output to load. In parallel with the load, a voltage divider of 1M+120K, with a very small cap in parallel with the 1M resistor?

[Rob Strand]

I don't know if I have a capacitor that small. I do believe I have a 4.7pF cap, but with some wild tolerance. That will work?

Two 4.7pF in series would be better if you have them.   If you have a probe "cal" output on your oscilloscope (just a square-wave) you can see if it's OK.  See wave forms at the bottom of this page,

http://electronics-diy.com/electronic_schematic.php?id=1037


I built a 2pF "gimick capacitor" using the solid core wire from my breadboard; like Cat5 wire.    Take two wires each 43mm long strip about 5mm of one end of each.  Put about 4 *full twists* in the wire.  A full twist means the spiral ends-up on the same side again -  so one twist is sort of like twisting twice with your hands.    It's about 30mm tall and you can bend it in half.

We don't know the input capacitance of your oscilloscope, so we don't know the capacitor value required.

So, to be clear: output to load. In parallel with the load, a voltage divider of 1M+120K, with a very small cap in parallel with the 1M resistor?

Yes that's right - - the 1M goes to the HV side of the divider and the 120k to the ground side.   You want to make sure the connection to the ground and 120k don't come off by accident.

[Marcos - Munky]

Sorry it took me so long, some things come in the way. So, here are the results.

This is the "cal" waveform. Looks good enough, so I didn't used a compensating cap.


Then I picked resistors for a low current (1mA), something in the middle (about 20mA) and something near the max (about 40mA). 1/10 voltage divider, só I got an 20V-ish output. This is when I noticed a problem: this cheap scope have so little resolution options I coudn't see the waveform! With the 5V resolution, the wave was so small. The next resolution is 2V, but then the waveform was too big to fit the small screen.

I took pictures anyway, maybe the info shown can be useful. If not, I can set the voltage divider to
something like 1/20 so I get a voltage of like 10V.

[Rob Strand]

This is the "cal" waveform. Looks good enough, so I didn't used a compensating cap.

Looks OK.

This is when I noticed a problem: this cheap scope have so little resolution options I coudn't see the waveform! With the 5V resolution, the wave was so small. The next resolution is 2V, but then the waveform was too big to fit the small screen.

At least the output looks relatively clean.

Normally you would AC couple the input so you can crank the sensitivity up.

If your oscilloscope doesn't have an input AC coupling mode then you should be able to put a 100nF cap between the divider output and the oscilloscope input.

You could get a 10x the sensitivity with a completely different set-up:

200V DC --->  10k resistor  --->  100nF 400V cap  --- 2x 1N4148 (reverse paralleled) clipping diodes to ground to protect the Oscilloscope input.

The 10k and the diodes stop the voltage pulse getting to the CRO when you first connect the 10k resistor.   The 10k resistor limits the current through the diodes.

Warning:  When you finish testing the cap with be charged to 200V.   You should be able to discharge it by grounding the 10k resistor.

[Marcos - Munky]

So I got the AC coupled pictures, but I think the result is not better that the previous one. At least it's not a big waveform, but it's a different one and may tell something useful. Probably the scope limitations are disturbing the measurements. Or it's the operator (me) who's doing a bad job - which I think is more likely .

Measures for 1mA, 20mA and 40mA:

[Rob Strand]

So I got the AC coupled pictures, but I think the result is not better that the previous one. At least it's not a big waveform, but it's a different one and may tell something useful. Probably the scope limitations are disturbing the measurements. Or it's the operator (me) who's doing a bad job - which I think is more likely

It looks like the ripple is a bit irregular.  It could be the oscilloscope display mode.

The size of the ripple is largely due to the ESR of the 47uF output cap.   

Now I think about it a 47uF high voltage cap is probably going to have a ripple spec of 1A at 100kHz.  So a high current boost converter is going to be pushing the cap rating.   The only way out is a large cap value which will have a higher ripple current rating.

I plugged your circuit into LTSPICE and it looks like the inductor current is in CCM mode and that you are getting some sub-harmonic oscillations.  That would cause the ripple to be irregular.   It's the same complaint marcelomd had in his sim but not as pronounced.

I wasn't sure what inductor value and sense resistor you used so I took the 100uH and 0.2ohm from transmogrifox's design.

It looks like transmogrifox made an effort to put in some slope-compensation to prevent the sub-harmonic oscillations.  That was some of the stuff you removed.    It also looks like he made a good effort to get a CCM  boost converter going.

I modded marcellomd's circuit to operate in DCM mode like I mentioned in reply #68 and it doesn't show any signs of sub-harmonic oscillations.   It also reacts quite fast.    I had a clock of 100kHz and when the 12V was stepped up to 13V the circuit only produced a 14mV blip  on the output and recovered in 6ms [edit].    Your current design (which is 50kHz) produced a blip of about 780mV and took 25ms to recover.    I didn't try transmogrifox's design to see how that performed.

My sims were only rough so I wouldn't read too much into it.
------------
EDIT:  changed recovery from 2ms to 6ms, 2ms was time to peak of blip.

[Marcos - Munky]

I used a 100uH inductor. For the resistor, I'm using 0.25r. The smaller resistor I could find was 0.5r, so I used two in parallel.

Transmogrifox indeed made a good effort to make a very nice circuit. I removed some stuff to size it to be "stompbox-friendly", so the performance dropped a lot for higher loads. But since a stompbox usually uses a tube or two, the smps can handle it pretty well.

ESR is something local sellers doesn't know a thing. Even online is kinda hard to find low ESR caps. So I had to use a regular cap. Also, my cap bank is way smaller than Transmogrifox's for the "stompbox-friendly" reasons, so the ripple is even bigger.

I think I should build the complete circuit, as Transmogrifox designed. Even with regular high ESR caps, it should have a way better performance just by adding the skipped parts and by using a decent cap bank.