High voltage +170V based on NE555 design - short circuit

[matb]

Hi guys, first time poster here.

I love Nixie tubes so I decided to build a Nixie clock.

I am a beginner. I have one successful design based on through-hole components and an Arduino.

Now I decided to try a SMD design (and I think way more complicated than the previous one.)

This design is based on an NE555 as that is the simplest one I could find online. This PCB is just to test the power supply generation and the drive of one Nixie tube IN-14. It is not the final circuit.

Here is the design:

Input voltage +12V

+170V generation

PCB without ground plane

PCB with ground plane

Picture of PCBA


What I know / found out:

I miscalculated the +9V LM317 resistors, it should be 1620 ohm and I soldered 1500 ohm.
Footprint of the fuse is way off, I soldered a wire to short it
C12 should be 100uF as per datasheet recommanded circuit
C14 is a 4.7u 400V aluminium capacitor (should have been 470uF.)
What is happening:

My power supply is showing a short circuit. I wire +9V@500mA (current limited) on the +9V test points as the +9V generation doesn't work.

Using a very scientific method (my fingers) I feel that Q26 (IRF740) is getting very hot.

Checking with my multimeter shows no short on any of the components (continuity mode) +9V and ground are not shorted when the power is off.

Is the +9V generation a must or can I feed the +12V to the NE555 circuit?

Can you guys see a problem in the design, maybe the choice of the components?

Farnell codes :
RV2 : 2786536 https://fr.farnell.com/littelfuse/v240ch8t/varistance-mov-395v-3220/dp/2786536?ost=2786536
D1, D8, D10 ES3J : 2677398 https://fr.farnell.com/taiwan-semiconductor/es3j-r6g/diode-simple-3a-600v-do-214ab/dp/2677398?ost=2677398
F1 : 3701350 https://fr.farnell.com/aem/t0603ff1250tm/fusible-tres-rapide-1-25a-65vdc/dp/3701350?ost=3701350
C3, C5, C11, C12 : 3013452 https://fr.farnell.com/samsung-electro-mechanics/cl21a106kaynnne/condensateur-10uf-25v-mlcc-0805/dp/3013452?ost=3013452
C14 4.7u 400V RS code 725-6966 (not 470u) https://fr.rs-online.com/web/p/condensateurs-electrolytiques/7256966
C10 : 2812527 https://fr.farnell.com/wurth-elektronik/885342007006/cond-47pf-500v-5-c0g-np0-0805/dp/2812527?ost=2812527
C9 : 3489975 https://fr.farnell.com/tdk/c5750x7r2e105m230ke/condensateur-1uf-250v-mlcc-2220/dp/3489975?ost=3489975
L1 : 3370508 https://fr.farnell.com/multicomp-pro/mp002851/inductance-de-puiss-100uh-blind/dp/3370508?ost=3370508
Q26 : 8657815 https://fr.farnell.com/vishay/irf740aspbf/transistor-mosfet-canal-n-d2pak/dp/8657815?ost=8657815
Q25 : 2706710 https://fr.farnell.com/rohm/sst2907ahzgt116/transistor-pnp-60v-0-6a-sot-23/dp/2706710?ost=2706710
Q24 : 1081232 https://fr.farnell.com/nexperia/bc847b-215/transistor-npn-sot-23/dp/1081232?ost=1081232
U12 NE555DR : 3121192 https://fr.farnell.com/texas-instruments/ne555dr/timer-single-precision-soic8-555/dp/3121192?ost=3121192
U2 : 2383003 https://fr.farnell.com/texas-instruments/lm1117sx-5-0-nopb/800ma-low-dropout-linear-regulator/dp/2383003?ost=2383003%20*
D9 : 1081180 https://fr.farnell.com/nexperia/bas16-215/diode-de-commutation-100v-215ma/dp/1081180?st=1081180

Probe on pin 3 of NE555DR:


Probe before and after the diode D9, that's where the issue comes I guess:

Here is the probe before D9:



Here is the probe after D9:


Desoldered D9 and powered the circuit: no changes I'm losing the 555 frequency after the diode.

Resoldered D9, desoldered Q25 and powered the circuit: no changes, same as before.

When I power the power supply at 9V and allow a little bit of current it instantly goes to something like 5V (instead of 9V) so I guess that's another proof that something is wrong somewhere.

The LTSpice simulation works so maybe I have a bad design or a bad component ?

[Ian.M]

The whole 9V regulator is *STUPID*.  The NE555 can tolerate 12V supply, and the MOSFET will work better with 10V or so gate drive and can withstand up to abs. max. +/-20V on its gate, so there is no reason to limit it if 15V or under is expected at its gate.

You are only getting about 4V drive (with leading edge spikes to 5V) out of the 555, and the other side of D9 you have a 4.68V DC level with a bit of noise on it.   That suggests Q25 is probably bad, as the MOSFET gate can charge via D9, but never discharges.   Also, *SOMETHING* is clamping the output as it should be swinging to about +7.5V, 1.5V below the nom. 9V VCC.  It cant go higher due to the high-side NPN Darlington in a bipolar 555's output stage.  Its possible the MOSFET cooked itself to death due to inadequate gate drive.

High ratio simple boost converters generally have poor efficiency - as a rule of thumb its best to avoid stepping up by more than a factor of 10, so for 170V out, you'd definitely want 12V in and may well want either a flyback circuit, or a voltage doubling rectifier circuit.   

Also the choice of a NE555 to control a boost converter is sub-optimal.  Why aren't you using a dedicated switching regulator controller chip, that can do cycle by cycle current limiting to protect the MOSFET?

[Siwastaja]

555 is not a switch mode controller. Get any actual controller IC for the project. Current limiting in one way or another is pretty much mandatory for reliable operation. Integrated ICs with integrated MOSFETs can do that by sensing the Vds, for example, maximizing simplicity and minimizing cost.

[matb]

Thank you for your replies

Why aren't you using a dedicated switching regulator controller chip, that can do cycle by cycle current limiting to protect the MOSFET?

Well I never designed a power supply before so I wanted to use the 555 as the design seemed fairly simple. I don't want to buy a pre-built high voltage module as I want to design one myself.
What IC should I consider then ?
I saw the Pile O'poo https://threeneurons.wordpress.com/nixie-power-supply/hv-supply-kit/ one uses a MC34063 should I try to use this one ?
I'm more afraid of the whole design, with ICs like that you really need precise layout etc

[matb]

That suggests Q25 is probably bad, as the MOSFET gate can charge via D9, but never discharges.   Also, *SOMETHING* is clamping the output as it should be swinging to about +7.5V, 1.5V below the nom. 9V VCC.  It cant go higher due to the high-side NPN Darlington in a bipolar 555's output stage.  Its possible the MOSFET cooked itself to death due to inadequate gate drive.

What do you recommand, replacing Q25 and Q26 ?

[Siwastaja]

Switch mode converter is switch mode converter. Using an unsuitable part does not make it simpler, it does the opposite, making it more complex to get working. I mean, you can do it with a 555, but that ain't simple. Or reliable. The core issue in 555-based circuit is the lack of current limit, saturating the inductor core and destroying the MOSFETs. It also lacks voltage reference and feedback amplifier, needed to regulate the voltage. You can get all this integrated in a single easy-to-use chip today.

I can't recommend anything because I don't personally use "go to" parts; I use what is available, modern, and seems good at the time of purchase. So just go to a distributor site, pick boost converter ICs section, choose suitable parameters and most importantly in today's situation, buy what is available. Follow the datasheet instructions regarding component values and layout. You can do it, it's important to start somewhere.

Yes, layout is important but that's not because of the IC, it's because how switch mode converters work. This is not explained in the 555 datasheet because the part is not designed to do it. Similarly, if you think operating manual for Airbus is too difficult, getting a bicycle instead seems easier, but that's only because it doesn't actually fly - unless heavily modified, which again isn't easy! In this case, get a part with a datasheet explaining inductor selection, capacitor selection, and layout. All this information may seem overwhelming at first, but it's finally helpful.

[Andy Watson]

I believe you have base and emitter swapped on the footprint of Q25.

[nctnico]

Thank you for your replies

Why aren't you using a dedicated switching regulator controller chip, that can do cycle by cycle current limiting to protect the MOSFET?

Well I never designed a power supply before so I wanted to use the 555 as the design seemed fairly simple. I don't want to buy a pre-built high voltage module as I want to design one myself.
What IC should I consider then ?

The NE555 has a lot of the parts needed to make a DC-DC converter so the idea is not outlandish. Especially if you have a fixed input voltage so the PWM ratio can remain constant. Many years ago I have designed the circuit in the attached PDF (from -48V to +120V). I had to resort to parts that where available so not every component is properly dimensioned. It has cycle-by-cycle current limit and output overvoltage limiting.

[Someone]

I mean, you can do it with a 555, but that ain't simple. Or reliable. The core issue in 555-based circuit is the lack of current limit, saturating the inductor core and destroying the MOSFETs. It also lacks voltage reference and feedback amplifier, needed to regulate the voltage. You can get all this integrated in a single easy-to-use chip today.

Current limit? yep.
Voltage feedback? yep.
Just a 555 and a fet? yep, single chip
https://www.eevblog.com/forum/projects/555-timer-boost-converter-(and-buck-converter)-switching-power-regulator/
Easy to use with idiot proof design steps in the application note? nope.
Cycle by cycle current limiting, nope.

Also the choice of a NE555 to control a boost converter is sub-optimal.  Why aren't you using a dedicated switching regulator controller chip, that can do cycle by cycle current limiting to protect the MOSFET?

Yes, switching regulators are much more robust when the switch is protected by immediate current limiting.

[Ian.M]

Cycle by cycle current limiting, nope.

Also the choice of a NE555 to control a boost converter is sub-optimal.  Why aren't you using a dedicated switching regulator controller chip, that can do cycle by cycle current limiting to protect the MOSFET?

Yes, switching regulators are much more robust when the switch is protected by immediate current limiting.

Cycle by cycle current limiting with a 555 is certainly possible and Nctnico's 555 boost circuit does it by resetting the 555 if/when the current limit trips.  The Reset pin only has a single internal transistor between it and the flipflop, and there's one external transistor pulling it down against a 10K pullup, so we can be fairly certain its got less propagation delay than Trigger to Output, which is about 300ns worst case for a high amplitude trigger pulse.  The IRF730 datasheet claims 38nc total gate charge with a Miller plateau at approx 6V, so the 555 output is going to take charge out of the gate at approx 100mA.   The MOSFET should therefore turn off fully within 700ns of the over-current protection tripping, which should keep it well within its SOA, as  the trip point is under 5A and the SOA curves say it can withstand as much as 20A for a millisecond.

Of course the tradeoff is lower efficiency due to I²R losses in the 0.3 ohm sense resistor.  A purpose designed boost controller can have a far lower trip voltage for its Isense input so can use a lower value sense resistor, and get better efficiency.

[Someone]

Cycle by cycle current limiting, nope.

Also the choice of a NE555 to control a boost converter is sub-optimal.  Why aren't you using a dedicated switching regulator controller chip, that can do cycle by cycle current limiting to protect the MOSFET?

Yes, switching regulators are much more robust when the switch is protected by immediate current limiting.

Cycle by cycle current limiting with a 555 is certainly possible and Nctnico's 555 boost circuit does it by resetting the 555 if/when the current limit trips....

Using additional components.

I mean, you can do it with a 555, but that ain't simple. Or reliable. The core issue in 555-based circuit is the lack of current limit, saturating the inductor core and destroying the MOSFETs. It also lacks voltage reference and feedback amplifier, needed to regulate the voltage. You can get all this integrated in a single easy-to-use chip today.

Current limit? yep.
Voltage feedback? yep.
Just a 555 and a fet? yep, single chip
https://www.eevblog.com/forum/projects/555-timer-boost-converter-(and-buck-converter)-switching-power-regulator/
Easy to use with idiot proof design steps in the application note? nope.
Cycle by cycle current limiting, nope.

Also the choice of a NE555 to control a boost converter is sub-optimal.  Why aren't you using a dedicated switching regulator controller chip, that can do cycle by cycle current limiting to protect the MOSFET?

Yes, switching regulators are much more robust when the switch is protected by immediate current limiting.

Different requirements, different solutions. Of course a 555 could do all sorts of things if you build a Rube Goldberg contraption around it with discrete transistors/fets, at some point its silly (projects/designs/people/situations will put that at different points). If you want to poke holes that were never there, the zener string coming back to the same inhibit/reset doing the current control in Nctnico's posted example is a very wobbly voltage feedback method, swapping to current signalling through Q502. So it would fail on the criteria of having a "feedback amplifier" by most peoples definition of feedback.

[nctnico]

Well, sometimes you don't need a feedback amplifier. Remember that the ratio between the input and output of a buck / boost converter is set by the PWM duty cycle when in continuous mode. In my circuit the zeners are there to prevent the output voltage from running away in discontinuous mode; not to provide regulation.

A feedback amplifier basically compensates for the diode /inductor voltage drop in a buck/boost converter but that is only necessary when the voltage drop over the diode and/or inductor is significant compared to the output voltage OR when the input voltage fluctuates a lot.

[aargee]

MC34063 works pretty well for the purpose, I have one built on Veroboard for a 'one day' Nixie clock. There are plenty of reference circuits for Nixie applications.
For my head, there are too many gaps to fill in pushing a 555 into service for this sort of work.

[matb]

What do you think of a design like this one would that be better ?
@7:01

[not1xor1]

555 is not a switch mode controller. Get any actual controller IC for the project. Current limiting in one way or another is pretty much mandatory for reliable operation. Integrated ICs with integrated MOSFETs can do that by sensing the Vds, for example, maximizing simplicity and minimizing cost.

one might also buy a ready made module... but there is no fun...
one can even make a current mode boost with a few common BJTs and a power MOSFET...
no need to worry about chip shortage

here is a simulation of a single Li-ion cell booster for 6 LEDs (stil untested) with more than 90% of efficiency

[not1xor1]

Even 1 BJT + 1 NMOS work... it is noisy and I've not measured the efficiency. A quick modification of another circuit...
I just wanted to see what might be achieved with a minimal count of common parts.
Note: I've not checked if the absolute maximum ratings of any part have been exceeded.

[xavier60]

This is was I used 14 years ago and still works.
The MOSFET could do with an upgrade.
The UC3843 comes out of UVLO at about 8V.

[Circlotron]

I believe you have base and emitter swapped on the footprint of Q25.

Andy Watson is on the money.
https://www.farnell.com/datasheets/2198030.pdf

[matb]

Wow just saw that I'm really off on that footprint.
I will change the mosfet and that transistor and I will keep you informed if that fixes the problem.

[nctnico]

BTW: one thing to be aware off is that many SMT inductors are rated for 50V maximum. For reliability you will want to use inductors that can deal with high voltages.

[poorchava]

I wonder why most of those designs use stuff like IRF740 and NE555. I mean there are so many better parts....

I mean lol, just take a pic10f200 or some crap arsuino and generate a PWM at least. Just one sot23 instead of loads of passives.

[nctnico]

I wonder why most of those designs use stuff like IRF740 and NE555. I mean there are so many better parts....

The design I posted is over 2 decades old 

[pardo-bsso]

I wonder why most of those designs use stuff like IRF740 and NE555. I mean there are so many better parts....

I mean lol, just take a pic10f200 or some crap arsuino and generate a PWM at least. Just one sot23 instead of loads of passives.

In many forgotten places (like where I am now...) those are components that you can still buy at a brick store for a not unreasonable amount of money. And also copy - paste engineering