Author Topic: MC34063 high voltage dc-dc boost converter  (Read 21437 times)

0 Members and 1 Guest are viewing this topic.

Online magic

  • Super Contributor
  • ***
  • Posts: 7458
  • Country: pl
Re: MC34063 high voltage dc-dc boost converter
« Reply #25 on: January 01, 2020, 05:00:56 pm »
Re-run the simulation and see if the predicted fall time really is in tens of nanoseconds.

My guess is that it's several milliseconds a microsecord ;)
Because it should coincide with the flat part in the middle of gate waveform falling edge.

You really need that turn-off speedup before you can start to worry about FET specs. And even then there is nothing to worry about at such low frequencies. 100ns repeated at a few tens kHz is only a few ms total per each second.
« Last Edit: January 01, 2020, 05:04:32 pm by magic »
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: MC34063 high voltage dc-dc boost converter
« Reply #26 on: January 01, 2020, 05:18:43 pm »
Datasheet rise/fall times aren't so much intrinsic to the device, they're typical of the drive conditions.  You'll see a lot of Fairchild datasheets with R_G = 25 or 50 ohms, even big transistors for some reason, which of course get laughable times like 400ns.  Typically you use a gate driver IC with an output resistance of a few ohms, and the rise time is in the 10s of ns.

Regarding the sim, the inductance looks awfully large.  Consider how much dI is obtained in the, what, 4us or so expected on-time, at 24V and L.  dI = V dt / L

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline dazzTopic starter

  • Frequent Contributor
  • **
  • Posts: 304
  • Country: es
Re: MC34063 high voltage dc-dc boost converter
« Reply #27 on: January 01, 2020, 05:25:33 pm »
Re-run the simulation and see if the predicted fall time really is in tens of nanoseconds.

My guess is that it's several milliseconds a microsecord ;)
Because it should coincide with the flat part in the middle of gate waveform falling edge.

You really need that turn-off speedup before you can start to worry about FET specs. And even then there is nothing to worry about at such low frequencies. 100ns repeated at a few tens kHz is only a few ms total per each second.

Exactly right. 1.15us until it reaches the flat portion of the falling edge. What would be the explanation for this, please? Shouldn't it be closer to the 45ns that the datasheet indicates? I don't know, but looking at the wave form of the gate signal, it doesn't look to me like the time spent in the linear zone is negligible at all, if that's what you meant. What am I missing?
 

Offline dazzTopic starter

  • Frequent Contributor
  • **
  • Posts: 304
  • Country: es
Re: MC34063 high voltage dc-dc boost converter
« Reply #28 on: January 01, 2020, 05:29:26 pm »
Datasheet rise/fall times aren't so much intrinsic to the device, they're typical of the drive conditions.  You'll see a lot of Fairchild datasheets with R_G = 25 or 50 ohms, even big transistors for some reason, which of course get laughable times like 400ns.  Typically you use a gate driver IC with an output resistance of a few ohms, and the rise time is in the 10s of ns.

Regarding the sim, the inductance looks awfully large.  Consider how much dI is obtained in the, what, 4us or so expected on-time, at 24V and L.  dI = V dt / L

Tim

Oh, right. I totally forgot about the transformer inductances. LOL
The thing is that I've been foolishly trying to get the simulation working properly with the old 400V 10A mosfet. So I started trying stupid things like increasing the inductance of the transformer until I finally realised I needed another mosfet... and then I totally forgot about the transformer  :-DD I'll fix that now.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: MC34063 high voltage dc-dc boost converter
« Reply #29 on: January 01, 2020, 05:40:17 pm »
Exactly right. 1.15us until it reaches the flat portion of the falling edge. What would be the explanation for this, please? Shouldn't it be closer to the 45ns that the datasheet indicates? I don't know, but looking at the wave form of the gate signal, it doesn't look to me like the time spent in the linear zone is negligible at all, if that's what you meant. What am I missing?

What's the test condition in the datasheet?  R_G or R_S or something.

What is the resistance here?

What is the gate capacitance?  (Why is that the wrong question to ask?)

What is the gate charge?  Gate capacitance equivalent?

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Online magic

  • Super Contributor
  • ***
  • Posts: 7458
  • Country: pl
Re: MC34063 high voltage dc-dc boost converter
« Reply #30 on: January 01, 2020, 05:47:17 pm »
The flat part is called "Miller plateau". It occurs due to so-called Miller capacitance, which is capacitance from gate to drain. These parts are close to each other and they form a capacitor hidden inside the FET. It's also called "reverse transfer capacitance" and as such you will find it on the datasheet.

As drain voltage raises, Cgd charges up and sends current into the gate. The rate of drain voltage raise is limited by the ability of your 200Ω resistor to pull that current out of the gate. In fact, you will see a perfectly linear ramp when you look at drain voltage over time.
 

Offline dazzTopic starter

  • Frequent Contributor
  • **
  • Posts: 304
  • Country: es
Re: MC34063 high voltage dc-dc boost converter
« Reply #31 on: January 01, 2020, 05:53:21 pm »
So if I got it right, it's some sort of stray capacitance opposing voltage change, hence slowing down the falling edge of the signal. I've read the pcb design is critical in switching regulators and power supplies, I guess stray capacitance is one of the main concerns for these reasons?
 

Offline dazzTopic starter

  • Frequent Contributor
  • **
  • Posts: 304
  • Country: es
Re: MC34063 high voltage dc-dc boost converter
« Reply #32 on: January 01, 2020, 06:02:40 pm »
Regarding the sim, the inductance looks awfully large.  Consider how much dI is obtained in the, what, 4us or so expected on-time, at 24V and L.  dI = V dt / L

Tim

Something's amiss here. I used the calculator for a flyback converter that you suggested yesterday: http://schmidt-walter-schaltnetzteile.de/smps_e/spw_smps_e.html
With 24V input, 300V and 200mA output, the calculator suggests 120uH in the primary and a turn ratio of 80... but as far as I know that would mean 80^2 = 6400 times more inductance in the secondary. That would be 7.68H in the secondary... can't be right. I must be making one of my stupid mistakes
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: MC34063 high voltage dc-dc boost converter
« Reply #33 on: January 02, 2020, 02:16:24 am »
The suggested values are for, I think, ripple fraction 10 or 20%.  You can always set it lower, and in this case you need to, with a peak current mode controller which will not behave well with a high inductance.

You misread the scientific notation: it's 79.81e-3 or 0.07981.  It's also labeled N1/N2 so your misread would imply the secondary is 80 times shorter! ;)

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline dazzTopic starter

  • Frequent Contributor
  • **
  • Posts: 304
  • Country: es
Re: MC34063 high voltage dc-dc boost converter
« Reply #34 on: January 02, 2020, 02:56:53 am »
The suggested values are for, I think, ripple fraction 10 or 20%.  You can always set it lower, and in this case you need to, with a peak current mode controller which will not behave well with a high inductance

Ripple fraction.. just another term that goes down on my list of things I need to research. Transformers are a world of their own apparently. Jeez, why does this stuff have to be so complicated?

You misread the scientific notation: it's 79.81e-3 or 0.07981.  It's also labeled N1/N2 so your misread would imply the secondary is 80 times shorter! ;)

Tim

Damn it! That was a clear case of confirmation bias. I was expecting a ratio > 1 so I missed the exponential term  |O
But I'm not sure I understand what you mean. If N1/N2 = Nprimary/Nsecondary = 79.81e-3, then the N2/N1 = 0.07981^-1=12.53
So if the suggested primary inductance is 120uH, the secondary would be (12.53^2)= 157 times that, hence 120uH x 157 = 18.8mH... assuming I'm not having another brain fart, that is
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: MC34063 high voltage dc-dc boost converter
« Reply #35 on: January 02, 2020, 04:55:07 am »
Correct! ;D

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline dazzTopic starter

  • Frequent Contributor
  • **
  • Posts: 304
  • Country: es
Re: MC34063 high voltage dc-dc boost converter
« Reply #36 on: January 02, 2020, 01:35:28 pm »
Correct! ;D

Tim

Bah, I just noticed I misread you too. Anyway, all clear now.  :-+

I'm getting some mixed results in the simulation now. Interestingly, the active pull down on the gate that Jack suggested worked great in the boost (no transformer) version, boosting the efficiency to a very nice 92%. OTOH I just can't seem to get the flyback version anywhere near as efficient. The active pull down doesn't work unless I reduce the current limiting resistor, which in turn increases the power dissipation in the mosfet, and the efficiency tanks as a result. Just to be sure, does the IRFZ44N look like a proper choice for this application?
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
Re: MC34063 high voltage dc-dc boost converter
« Reply #37 on: January 02, 2020, 01:46:49 pm »
I'm getting some mixed results in the simulation now. Interestingly, the active pull down on the gate that Jack suggested worked great in the boost (no transformer) version, boosting the efficiency to a very nice 92%. OTOH I just can't seem to get the flyback version anywhere near as efficient. The active pull down doesn't work unless I reduce the current limiting resistor, which in turn increases the power dissipation in the mosfet, and the efficiency tanks as a result. Just to be sure, does the IRFZ44N look like a proper choice for this application?

The IRFZ44N is a very old MOSFET so, no, not a good choice for anything these days.

The isolated flyback is somewhat tricky to design. Post your LTSpice .asc file here if you want some help with tuning it, but as I already warned, SMPS transformer design is an art unto itself, so expect a lot of pain and escaping magic smoke.

 

Offline dazzTopic starter

  • Frequent Contributor
  • **
  • Posts: 304
  • Country: es
Re: MC34063 high voltage dc-dc boost converter
« Reply #38 on: January 02, 2020, 02:00:28 pm »
I'm getting some mixed results in the simulation now. Interestingly, the active pull down on the gate that Jack suggested worked great in the boost (no transformer) version, boosting the efficiency to a very nice 92%. OTOH I just can't seem to get the flyback version anywhere near as efficient. The active pull down doesn't work unless I reduce the current limiting resistor, which in turn increases the power dissipation in the mosfet, and the efficiency tanks as a result. Just to be sure, does the IRFZ44N look like a proper choice for this application?

The IRFZ44N is a very old MOSFET so, no, not a good choice for anything these days.

The isolated flyback is somewhat tricky to design. Post your LTSpice .asc file here if you want some help with tuning it, but as I already warned, SMPS transformer design is an art unto itself, so expect a lot of pain and escaping magic smoke.

Yeah, I'm not going to build the flyback smps anytime soon. I'll make sure I do proper research before I do that.
The asc is attached, thanks
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
Re: MC34063 high voltage dc-dc boost converter
« Reply #39 on: January 02, 2020, 03:13:39 pm »
Something's amiss here. I used the calculator for a flyback converter that you suggested yesterday: http://schmidt-walter-schaltnetzteile.de/smps_e/spw_smps_e.html
With 24V input, 300V and 200mA output, the calculator suggests 120uH in the primary and a turn ratio of 80...

Hey, that website is one I often recommend to SMPS beginners but it is giving me a "bad request" error on most (but not all) of the sections. Is it just me, or is it broken for others?

Yeah, I'm not going to build the flyback smps anytime soon. I'll make sure I do proper research before I do that.
The asc is attached, thanks

Right, a bunch of things wrong just at first glance. The transformer phasing is wrong - for a flyback the dot ends are inverted w/r/t each other - and you don't include the H when specifying the inductance, so the primary should just be 120u, not 120uH (you did specify the secondary correctly).

Next big issue is there is no RCD or Zener clamp across the primary; switch destruction is guaranteed in the real world, but LTSpice apparently doesn't care...  :P

You also need to specify a diode on the secondary with an appropriate voltage rating; a 40V or whatever Schottky ain't gonna cut it with a 300V output.

The pot isn't a standard LTSpice component and simulating 3 seconds is about 2.99 seconds too much (quite literally: 10ms should be enough here). I like to set the simulation time to capture 100 to 1000 switching cycles, or 4-10 AC mains cycles, whichever applies.

Probably some other things but this should be enough to make your brain hurt for awhile.
 

Offline dazzTopic starter

  • Frequent Contributor
  • **
  • Posts: 304
  • Country: es
Re: MC34063 high voltage dc-dc boost converter
« Reply #40 on: January 02, 2020, 03:19:25 pm »
Damn, thanks for that, MagicSmoker. I'm hopelessly clueless,  :-DD

And http://schmidt-walter-schaltnetzteile.de/ seems to work just fine for me. I tried a few pages and got no errors
 

Offline dazzTopic starter

  • Frequent Contributor
  • **
  • Posts: 304
  • Country: es
Re: MC34063 high voltage dc-dc boost converter
« Reply #41 on: January 02, 2020, 03:24:43 pm »
The pot isn't a standard LTSpice component and simulating 3 seconds is about 2.99 seconds too much (quite literally: 10ms should be enough here). I like to set the simulation time to capture 100 to 1000 switching cycles, or 4-10 AC mains cycles, whichever applies.

And sorry about that, forgot about the pot model. I got it somewhere on the web and tweaked it to my liking, but I think it may slow down simulation runs a bit too much, not sure.
I can post the model if needed

ETA: The reason for the long simulation time is because with large filter cap values it takes a while to reach 300V in the output, not sure if there's anything I can do to speed it up
« Last Edit: January 02, 2020, 03:30:35 pm by dazz »
 

Offline dazzTopic starter

  • Frequent Contributor
  • **
  • Posts: 304
  • Country: es
Re: MC34063 high voltage dc-dc boost converter
« Reply #42 on: January 02, 2020, 04:36:15 pm »
Now we're talking. Fixed the tranny's polarity and increased the gate pull down resistor and got a 96% efficiency already. I still need to find a fast switching diode that will take 500V plus peak reverse voltage and perhaps a better mosfet.

ETA: OK, I believe a UF4007 should do for the diode
« Last Edit: January 02, 2020, 05:39:19 pm by dazz »
 

Offline dazzTopic starter

  • Frequent Contributor
  • **
  • Posts: 304
  • Country: es
Re: MC34063 high voltage dc-dc boost converter
« Reply #43 on: January 03, 2020, 03:28:40 am »
By using a transformer, you get all the ratio with none of the wasted capacity.  You get an additional wrinkle of leakage inductance (and also stray capacitance, at high voltages), but this isn't new as such, it was always present (stray inductance between the transistor, diode and cap); it's only a matter of correct transformer design, and snubbing if applicable.

Next big issue is there is no RCD or Zener clamp across the primary; switch destruction is guaranteed in the real world, but LTSpice apparently doesn't care...  :P

Learning about this stuff right now. Already understand what you mean by leakage inductance and why I need a snubber network, either RCD or Zener, and I'm in the process of learning how to pick the right clamping voltage to have the best compromise between limiting the voltage overshoot in the mosfet and the power losses in the snubber itself. I can see in LTS the massive spikes in the mosfet's drain voltage (and current). Let's fix that
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
Re: MC34063 high voltage dc-dc boost converter
« Reply #44 on: January 03, 2020, 11:14:23 am »
Learning about this stuff right now. Already understand what you mean by leakage inductance and why I need a snubber network, either RCD or Zener, and I'm in the process of learning how to pick the right clamping voltage to have the best compromise between limiting the voltage overshoot in the mosfet and the power losses in the snubber itself. I can see in LTS the massive spikes in the mosfet's drain voltage (and current). Let's fix that

In the real world - especially if input voltage range is wide - the RCD clamp is usually the better choice (and is more forgiving of sloppy component selection) but the Zener clamp seems to be easier for beginners to understand and is an excellent choice at lower power levels and/or when the input voltage is more or less fixed.

The basic idea is to pick a Zener voltage that is around 1.5x to 2x the reflected voltage from the secondary, or the secondary voltage divided by the turns ratio going from secondary to primary. For example, the voltage reflected back to a 1 turn primary from a 300V secondary that has 12 turns is 25V (300 / 12); a Zener (or TVS diode) rated for 36V to 51V (using standard values) would be appropriate. You want the clamping voltage to be as high as possible to both reset the leakage inductance quickly and minimize the loss in the clamp - current will not flow out of the secondary until the leakage inductance is reset - but not so high that you have to increase the voltage rating of the switch so much that efficiency suffers from the inevitably higher Rds[on] (keeping price constant). Note that the Zener clamp doesn't help snub any high frequency ringing - it can make it worse, in fact - so you might need a light RC damper across the primary, too (but this is getting into one of motivations for using an RCD clamp... a bit advanced of a topic for the moment).
 

Offline dazzTopic starter

  • Frequent Contributor
  • **
  • Posts: 304
  • Country: es
Re: MC34063 high voltage dc-dc boost converter
« Reply #45 on: January 03, 2020, 11:40:30 am »
Learning about this stuff right now. Already understand what you mean by leakage inductance and why I need a snubber network, either RCD or Zener, and I'm in the process of learning how to pick the right clamping voltage to have the best compromise between limiting the voltage overshoot in the mosfet and the power losses in the snubber itself. I can see in LTS the massive spikes in the mosfet's drain voltage (and current). Let's fix that

In the real world - especially if input voltage range is wide - the RCD clamp is usually the better choice (and is more forgiving of sloppy component selection) but the Zener clamp seems to be easier for beginners to understand and is an excellent choice at lower power levels and/or when the input voltage is more or less fixed.

The basic idea is to pick a Zener voltage that is around 1.5x to 2x the reflected voltage from the secondary, or the secondary voltage divided by the turns ratio going from secondary to primary. For example, the voltage reflected back to a 1 turn primary from a 300V secondary that has 12 turns is 25V (300 / 12); a Zener (or TVS diode) rated for 36V to 51V (using standard values) would be appropriate. You want the clamping voltage to be as high as possible to both reset the leakage inductance quickly and minimize the loss in the clamp - current will not flow out of the secondary until the leakage inductance is reset - but not so high that you have to increase the voltage rating of the switch so much that efficiency suffers from the inevitably higher Rds[on] (keeping price constant). Note that the Zener clamp doesn't help snub any high frequency ringing - it can make it worse, in fact - so you might need a light RC damper across the primary, too (but this is getting into one of motivations for using an RCD clamp... a bit advanced of a topic for the moment).

Awesome, thanks. I've been trying zeners in that range of voltage, and sure enough, they don't do anything for the large, almost instant spikes in the rising edge of the waveform.
I'm currently trying this:



What I don't understand is why the clamp is connected to the positive line instead of GND, wouldn't that derive the spikes right back into the tranny?

I will also try the RCD clamp, good to know it's a better option here
« Last Edit: January 03, 2020, 11:45:23 am by dazz »
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
Re: MC34063 high voltage dc-dc boost converter
« Reply #46 on: January 03, 2020, 12:09:19 pm »
What I don't understand is why the clamp is connected to the positive line instead of GND, wouldn't that derive the spikes right back into the tranny?

I will also try the RCD clamp, good to know it's a better option here

You can return the clamp to ground - ie, wire it across the switch - but then it has to withstand the sum of the input voltage + the reflected voltage (ie - same as the switch). The transformer (and switch) don't really care one way or the other. The same applies to the RC damper, if used. BTW - a good rule of thumb for the RC damper is to make C about 3x the output capacitance of the switch and R somewhere around 1x to 3x the characteristic impedance of the LC network formed between the total capacitance of switch and damper and the leakage inductance. For example, if there is 2.4uH of leakage and the switch output capacitance is 50pF then a damper comprised of 150pF and 100-330R will likely clean up the highest frequency ringing (the lower frequency ringing in the flyback is between the magnetizing inductance and the lumped capacitance and can't really be suppressed as it is invariably too close to the switching frequency).

One other thing is that 1% leakage is a more realistic minimum for a typical E-core design (and even that requires considerable care in winding geometry), so set the coupling coefficient, K1 (2, 3, etc.) to 0.995 (leakage factor is 1-K2).
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: MC34063 high voltage dc-dc boost converter
« Reply #47 on: January 03, 2020, 12:29:15 pm »
As long as C2 is close to the snubber and transistor, it acts to turn the power rail into a supernode, i.e., in AC terms, equivalent to GND.  In an AC-equivalent circuit, all rails are drawn as GND, which looks very strange the first few times you see it, but helps illustrate how things work.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline dazzTopic starter

  • Frequent Contributor
  • **
  • Posts: 304
  • Country: es
Re: MC34063 high voltage dc-dc boost converter
« Reply #48 on: January 03, 2020, 12:41:29 pm »
Thanks guys. Plenty of things to google in those two last posts. If you find it takes a while to load google.com, bare with me, this is important  :-DD
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 17429
  • Country: us
  • DavidH
Re: MC34063 high voltage dc-dc boost converter
« Reply #49 on: January 03, 2020, 02:31:46 pm »
What I don't understand is why the clamp is connected to the positive line instead of GND, wouldn't that derive the spikes right back into the tranny?

As is usually the case, the schematic does not reflect what is actually happening with the circuit layout.  (1) The leakage inductance spike which needs to be snubbed comes from the transformer primary so minimizing the loop area means placing the snubber electrically close to the primary connections.

As T3sl4co1l points out, C2 makes the positive and negative input the same AC ground but besides the issue MagicSmoker identified, this also increases the loop area to extend through C2 lowering the effectiveness of the snubber.

(1) There is a way to indicate proper layout on the schematic with 45 degree connections but hardly anybody bothers.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf