High voltage boost converter design (Nixie)

[bpye]

Hey,

I'm trying to design a boost converter for some nixie tubes, aiming to get around 200V at 15mA, which leaves some headroom as actually 180V @ 15mA should be all that's needed for the 6xIN-12As I'm trying to drive. I have based this off of the Microchip app note ww1.microchip.com/downloads/en/AppNotes/91053b.pdf .



I'm trying to get this from [email protected], something which a lot of USB chargers can do. I have designed most of the boost mode converter in LTSpice, it is designed to operate in discontinuous mode and will be controlled by an STM32's PWM output, using the ADC for feedback (through a voltage divider).

I have a couple of issues to overcome however. I need to add some sort of gate driver for the BSZ16DN25NS3 MOSFET. I have opted for a totem pole configuration with BC847C/BC857C BJTs and I have also used a BC847C as a logic level converter as the STM32 IO is 3.3V but the maximum threshold voltage for the FET is 4V, so I need to increase the voltage slightly.

I have a couple of unresolved issues:

• The snubber (R1/C4) is based upon http://www.cde.com/resources/technical-papers/design.pdf from page 4. I can't take the optimisation steps as I lack PCBs, but does this seem reasonable?
• I'm not sure the BC847C/BC857C are a good choice, but I can't find anything better either. I suspect it's definitely overkill for the logic level converter but I figured 1 less part in the BOM is probably more important.
• I'm not sure my modelling for V1 is reasonable. I figured that USB is 4.5-5.5V so I have modelled with a 5V source that should experience .5V drop at 1.5A by setting the series resistance to .33ohm, is this reasonable?
• I have kinda guessed the values for R3 and especially R4, any suggestions?
• Should I just use an off the shelf MOSFET gate driver? It seems the passives/discretes are cheaper but is it worthwhile?
• Should I add a cap across the totem pole driver? Some literature appears to indicate I should ( www.ti.com/lit/ml/slua618/slua618.pdf )

As it stands it does appear to meet my requirements but I definitely could have missed something. I know I need to add some sort of switch to the 5V input as a USB device is only allowed up to 10u of capacitance until it enumerates.

I have attached the LTSpice model for my circuit and the diode I have chosen. I appreciate any advice! I'm a CS major so this is a little different for me, I certainly haven't touched power electronics at all.

Thanks!

Ben.

[T3sl4co1l]

It's better(?) with a 555,
https://www.seventransistorlabs.com/Images/555%20Boost.pdf
but there are more concise ways to realize the same system.

The first priority is to note why this circuit works: what is it switching based on, and why?  The rest (namely, getting a regulated output voltage) isn't as important.

For high voltage, do consider an extended winding.  You can wind your own (use a gapped ferrite core, or powdered iron), or buy an off-the-shelf multi-winding inductor (a 1:3 to 1:10 ratio would be perfect), or DC-DC transformer (like a 120-240V to 12V flyback transformer, wire it up in reverse).  Doing it in a single stage is just not a good idea, for a number of reasons.

Tim

[ebclr]

[bpye]

It's better(?) with a 555,
https://www.seventransistorlabs.com/Images/555%20Boost.pdf
but there are more concise ways to realize the same system.

The first priority is to note why this circuit works: what is it switching based on, and why?  The rest (namely, getting a regulated output voltage) isn't as important.

For high voltage, do consider an extended winding.  You can wind your own (use a gapped ferrite core, or powdered iron), or buy an off-the-shelf multi-winding inductor (a 1:3 to 1:10 ratio would be perfect), or DC-DC transformer (like a 120-240V to 12V flyback transformer, wire it up in reverse).  Doing it in a single stage is just not a good idea, for a number of reasons.

Tim

Your 555 based converter is actually pretty neat, I haven't seen any 555 based boost converters with feedback before. The reason I'm using the STM32 is I already have one on board for controlling the displays so it doesn't increase my part count too much to use it to drive the SMPS too.

I haven't considered using an inductor with multiple windings, I'm not too sure of the advantages, is there anything you suggest I read?

[T3sl4co1l]

This sort of thing: https://www.eetimes.com/author.asp?doc_id=1319222

Tim

[bpye]

I realised I had the PNP flipped on the previous schematic.



I have looked a little at having a multiple winding inductor, though the eetimes article did suggest a DCM boost converter as an alternative which is what this would be. The switch is needed as the 470uF cap would otherwise violate the USB spec.

[anderoo]

Hi bpye. A concern with this approach is that a software error can leave the N-ch FET stuck on which will cause destructive current to flow through the inductor.

So I guess you have accepted this possibility of software error and have chosen to live with it.

To prevent the N-ch FET turning on due to SMPS_SW becoming Hi-Z when the MCU is in RESET you can add a pull up resistor to 3.3V at the base of the BC847 that is controlled by SMPS_SW. This will cause that BC847 to be on by default and force the totem pole to pull low.

To protect the inductor against a short to GND at the HV pin, you should include a series resistor for each nixie anode as part of this power supply circuit. These anode resistors will be several kOhms for your application so current pulled from the 5V rail due to a short to GND on HV will be tiny.

[Buriedcode]

Every time I see 'boost', 'flyback' or 'nixie', I link this webpage:

http://www.dos4ever.com/flyback/flyback.html

It has a boost converter design using the venerable 555, and more importantly, explains exactly how it works.  I regularly see posts from people who knock up circuits without knowing/caring how they work (although the fact you've simulated one in LTspice means you must have some idea!).

As someone else mentioned, the trouble with using a MCU in the feedback loop is, you're at the mercy the firmware.  With a fixed inductor and constant input voltage you have a fixed maximum ON-time before the core saturates.  And regardless of how fast the MCU is, it will almost always be significantly slower to react than a half decent closed loop analogue system (even just a comparator with hysteresis).  Sure, it has been done many times, and if the load doesn't change quickly then it shouldn't have to react quickly, but it still means you MCU has to pretty much constantly update the PWM.

If you can find a dedicated current-mode boost converter chip that uses an external switch/MOSFET, use it.  It will be much better than a 555, which itself is much better than using an MCU.

You can still use the MCU's PWM to modulate the feedback, and therefore, setting/changing the output voltage digitally but I don't think it should be used as the actual switching controller.

If you build the above boost conerter, or design your own and can't get anything to work, I have several, very basic 555 current-mode boost converter designs from way way way back.

[flowib]

You have got quite a tall order building a supply that works from 5V the switch will see quite high peak inductor currents so will have a relatively high loss over the switch.  (My calculations show 0.5A- 1.5A peak inductor current @ 180V out, 10-30mA out..) Given that its an high voltage device a mosfet will have 1-2 ohms RsdOn (Except for BIG mosfets that need a beefy driver to switch on and off)  A  HV BJT will saturate at ~500mV 

One idea would be to boost the +5V to say 15V or use an external power adaptor..

Perhaps you would be better off using a simple switchmode driver IC such as the 34063

[ogden]

I'm trying to design a boost converter for some nixie tubes, aiming to get around 200V at 15mA

I would just use xenon flash charge controller in continuous mode. One of possible choices is MAX8622. Sure, 1:10 to 1:15 transformer needed, but at least you don't have to invent the wheel.

[bpye]

Hi bpye. A concern with this approach is that a software error can leave the N-ch FET stuck on which will cause destructive current to flow through the inductor.

So I guess you have accepted this possibility of software error and have chosen to live with it.

To prevent the N-ch FET turning on due to SMPS_SW becoming Hi-Z when the MCU is in RESET you can add a pull up resistor to 3.3V at the base of the BC847 that is controlled by SMPS_SW. This will cause that BC847 to be on by default and force the totem pole to pull low.

To protect the inductor against a short to GND at the HV pin, you should include a series resistor for each nixie anode as part of this power supply circuit. These anode resistors will be several kOhms for your application so current pulled from the 5V rail due to a short to GND on HV will be tiny.

SMPS_PWREN is actually to disconnect the load from the 5V rail when the device is first connected to USB. The USB spec states the maximum capacitive load is 10uF, which the 470uF cap exceeds.

Adding a pull-up does however seem a sensible change. I do also have a 2A PTC on the 5V input, the maximum current I want to draw is 1.5A so that seems reasonable. I do already have the anode series resistors.

Every time I see 'boost', 'flyback' or 'nixie', I link this webpage:

http://www.dos4ever.com/flyback/flyback.html

It has a boost converter design using the venerable 555, and more importantly, explains exactly how it works.  I regularly see posts from people who knock up circuits without knowing/caring how they work (although the fact you've simulated one in LTspice means you must have some idea!).

As someone else mentioned, the trouble with using a MCU in the feedback loop is, you're at the mercy the firmware.  With a fixed inductor and constant input voltage you have a fixed maximum ON-time before the core saturates.  And regardless of how fast the MCU is, it will almost always be significantly slower to react than a half decent closed loop analogue system (even just a comparator with hysteresis).  Sure, it has been done many times, and if the load doesn't change quickly then it shouldn't have to react quickly, but it still means you MCU has to pretty much constantly update the PWM.

If you can find a dedicated current-mode boost converter chip that uses an external switch/MOSFET, use it.  It will be much better than a 555, which itself is much better than using an MCU.

You can still use the MCU's PWM to modulate the feedback, and therefore, setting/changing the output voltage digitally but I don't think it should be used as the actual switching controller.

If you build the above boost conerter, or design your own and can't get anything to work, I have several, very basic 555 current-mode boost converter designs from way way way back.

I have considered flyback or tapped inductor topologies, my issue is finding a suitable inductor. There are some that would work but it seems the £1 FET I'm using for the simple boost converter is cheaper than the inductors/transformers I'd need for a reasonable flyback/tapped inductor topology. There are a lot with a 1:1 turn ratio but 125V MOSFETs aren't much cheaper, if I could find a good 1:4+ ratio inductor then this would be a nice idea. Coilcraft have a cheap(ish) 1:4 inductor, but, it's saturation current is only 1.3A and I seem to need about 2A to get the 180+V@15mA. For reference, https://www.coilcraft.com/NA5880.CFM , is the part that very nearly works.

I guess I figured the load shouldn't change too much so having the switcher controlled by the micro wouldn't be an issue, I might take another look at dedicated switcher ICs though. The STM32 should be able to sample the ADC input every 1uS or so, and I'm using the hardware PWM peripheral. I could use it's comparator but I don't think that's really an improvement?

You have got quite a tall order building a supply that works from 5V the switch will see quite high peak inductor currents so will have a relatively high loss over the switch.  (My calculations show 0.5A- 1.5A peak inductor current @ 180V out, 10-30mA out..) Given that its an high voltage device a mosfet will have 1-2 ohms RsdOn (Except for BIG mosfets that need a beefy driver to switch on and off)  A  HV BJT will saturate at ~500mV 

One idea would be to boost the +5V to say 15V or use an external power adaptor..

Perhaps you would be better off using a simple switchmode driver IC such as the 34063

Yeah I do need a fairly big FET, as I mentioned I plan to use the BSZ16DN25NS3, it has an Rdson 165m which seems reasonable. This is the reason for the totem pole driver, the max gate threshold is 4V so the USB VBus should be okay. This is the reason for the logic level converter and totem pole output. I definitely want to avoid an external power supply, maybe two stages could work though.

I'm trying to design a boost converter for some nixie tubes, aiming to get around 200V at 15mA

I would just use xenon flash charge controller in continuous mode. One of possible choices is MAX8622. Sure, 1:10 to 1:15 transformer needed, but at least you don't have to invent the wheel.

I would rather avoid Maxim, but I haven't really looked at Xenon flash controllers. Same issue with finding a 1:N transformer as for the tapped boost converter though I suspect.

[Buriedcode]

Although that webpage I linked covers flybacks it also includes a boost converter.

You are correct in that its a trade-off between Ron (lower the better) and Maximum Vds (which itself tends to increase Ron) because the switch has to both conduct Ipeak, and cope with a peak voltage of ~Vout before breaking down.  That is essentially why flybacks exists, to reduce the Maximum voltage across the switch - that and you can have an isolated output, or an output that falls below the input voltage.

But these days you can get MOSFETs with relatively high breakdown voltage and reasonably low Ron, FCA47N60 is an example.  It does have a rather large gate charge though.

I'll agree that finding suitable transformers 'off the shelf' is very difficult.  Even those 1:10 flybacks designed for Xenon flashes are hard to find (also tend to be <10uH primary which eliminates the possibility of you driving it with a n MCU).  But note that for a boost converter with a very high output/input ratio, it wont' be the most efficient.

Edit: "reduce" changed to "increase"... dumb mistake.

[bpye]

Although that webpage I linked covers flybacks it also includes a boost converter.

You are correct in that its a trade-off between Ron (lower the better) and Maximum Vds (which itself tends to reduce Ron) because the switch has to both conduct Ipeak, and cope with a peak voltage of ~Vout before breaking down.  That is essentially why flybacks exists, to reduce the Maximum voltage across the switch - that and you can have an isolated output, or an output that falls below the input voltage.

But these days you can get MOSFETs with relatively high breakdown voltage and reasonably low Ron, FCA47N60 is an example.  It does have a rather large gate charge though.

I'll agree that finding suitable transformers 'off the shelf' is very difficult.  Even those 1:10 flybacks designed for Xenon flashes are hard to find (also tend to be <10uH primary which eliminates the possibility of you driving it with a n MCU).  But note that for a boost converter with a very high output/input ratio, it wont' be the most efficient.

Yeah I'm well aware effeciency suffers. It does seem that with a 1:4 inductor that I can't find, I would get significantly better effeciency (about 20-30% better or something), but given I can't find the part I may try and build the straight boost anyway. I have ordered some of that 1:4 inductor as I would like to try and build a tapped boost converter for a future project with lower power requirements but an even worse ratio (lithium polymer battery...) and some quick maths says it should work there.