Seeking advice regarding structure of a DC-DC/PWM reg. for personal vaporizer.

[David_]

d123 no I meant Digital Power-Conversion for the Analog Engineer, though they seem to be talking about sort of the same things.

I have just begun to really grasp the extent of research and new learning that this project will require to reach the end result sought after, but that only means a longer time until the start of the real fun though more of it. Unless I would run into some really difficulties in learning things, but I think that I'll be able to make something work.


I think that in order to reach a desirable time frame for the procedure of getting the coils up to the set temperature one needs to apply more power than the power which would result in an equilibrium between the desired temperature and the power applied, at least this is how I see it but I may very well be wrong about this.

If we assume that the coils is at room temperature(≈25C°), what one... "I don't know the English word for one satisfying inhale of a cigarette/smoking device is" requires to happen is in my case the following:

I can with my device specify a temperature and a wattage value, the device is displaying also the resistance of the coils and the current supplied to them(though that value doesn't update fast enough to be able to tell anything of the actual procedure of the algorithm) though the temperature is displayed so that one can with a not very comfortable position watch the temperature rise/fall but that is also at such a slow rate that it doesn't revile anything really interesting.

So I set a temperature of 220C° and a wattage of 55W with coils adding up to 0,27Ω, when I press the switch the current is ramping up to 13,5A and then falls to around 13,1A(there is no way ow telling how accurate these values are, all I know is that while comparing to examples of this device side by side the values didn't agree between the two devices and relating to a single device the numbers doesn't always match with each other, some inconsistencies may also be present. When I first got this device I spent some time observing the values and then making my own calculations to check them and they weren't always correct... When buying rather cheap electronics from China you never know what the quality will be like), I can also see the wattage number climbing.

Then when the temperature is reached the current drops along with the wattage and then for the remainder of the inhale those numbers jump up and down.
Though with this particular setup with these settings the values is pretty consistent but if I for example lower the temperature setting then the values will first overshoot much more before the settle around a lower value(s).

I take this to point to the need for a higher and sometimes much higher initial power applied to the coils, followed by a control loop action applied to maintain the set temperature. And I feel rather confident that I can achieve a better loop than my current device has.
Because after the initial climb/shoot over when the values have settled to it's lower range they still jump up/down quite a bit and it is enough so that I can feel the vapour production increasing/decreasing with them, and I believe that with my own design I would have to begin with better capabilities to perform better, which is the main advantage I see my self having compared to these commercial produced devices.

Giving what have been written since my last post and my self pondering the ideas around this project I think that I will probably stay with XMEGA(the future is going to be interesting given Microchips purchase of Atmel), there is so much other stuff to learn and relating to ice-creams I think XMEGA will taste just fine. I wouldn't want to switch platform if I don't have to so I guess I will wait until XMEGA proves to be insufficient which will probably not happen.

 

[David_]

Hello.

I have a couple of things I'd like to talk about.

1, Gate driver arrangement.
The only solution to this that I have come up with is to use a CTL7660 based(or whatever those 7660 where named) IC, today there are lots of them out there that is pin-to-pin compatible with the original ones only operating at higher frequencies(often selectable) and far greater output current, up to 100mA.
So the idea is to use one of those to double the voltage and then switch that voltage with BJTs with the mcu PWM signals.

But there will always be a difference between the two MOSFETs as the high-side MOSFET will always suffer a decrease in voltage equal to the battery voltage or half the gate drive voltage after the charge-pump, which will mean that the two MOSFETs will display different gate charge characteristics, but does the gate charge influence anything other than the loading of the gate driver?
Datasheet's revile that gate charge increases quite dramatically with higher gate-source voltages,but as far as I understand it doesn't affect the other circuit parameters in any way.

2, the control circuits supply...
I can't remember any specifics but I do know that is it far better to operate the MCU I am going to use at 3,3V rather than at 2,7V, and as the ADCs in a design like this will determine much of the converters resulting performance I want to operate the control circuits from a stable and clean supply. So ether I use a LDO to drop the minimum 3V from the battery and run it all on 2,7V or I look for a low-power fully integrated high frequency buck-boost regulator so that can at all battery voltages derive a stable clean 3,3V rail through a LDO which cleans u the buck-boost output.

Have any of you any thought on this?
I consider a buck-boost IC much because of the low current needs from the control circuits so such an arrangement aught to be fairly small, and being able to count on a 3,3V(or 3V) rail makes it all much simpler.

3, How to sense the input/output current?

In my MCU I have 2 separate ADC peripherals(as well as 2 separate DAC peripherals) so I am hoping those will be sufficient to accomplish my goals satisfactorily so that I don't need to add external data converters which I originally planned(since I wanted a 14-bit system and not a 12-bit system as the internal data converters present). But I have been thinking about how the actual current sensing is going to be done and I have been drawn to the idea of using integrated current sense amplifiers who's output voltages is then digitized by the ADC, this seems rally trivial as far as the input current is concerned(but I am not sure about what kind of ripple voltage/current to expect on the input yet). But the output current might be more tricky.

Do you think I should attempt to accomplish cycle-by-cycle current limiting?
If so maybe I should look into implementing an external sample-and-hold circuit to be able to sample exactly the right point in the current waveform...?


I know that some of my thoughts would be answered by a test circuit but I feel as I should go through it all and not build the test circuit before I can implement all the stages that would be required for an actual control loop, though I am not that far of I believe.
I think that this also displays one of the main problems with ADD when it comes to project management, in that ones focus and thoughts jump all over the place and I am acutely aware that such an approach(willingly or unwillingly implemented) takes longer time(sometimes far longer) to complete than an more structured approach,but I am hoping I will be able to break from this behaviour in time. I only bring it up because I feel as it may be easy to get an impression that I ask for advice, then are given advice only to appear to not follow that advice, but I really am trying to do so.

 

[d123]

Hi,

Useless answer (sorry) (I use the word "you" a lot, but it's referred to me, you, anyone - "you" as in "one"):

1) Why can't you drive the high-side MOSFETs from the stepped-up/charge-pumped voltage? Could you use an amplifying transistor to boost the MCU PWM signals? I admit I don't understand/see what you're explaining.

2) If the battery is only 3.7V, and you'll let it discharge to 3V at most, then I doubt there is - yet - in existance an LDO that can go that low r.e. drop-out voltage, so either 2.7V MCU supply with boosted outputs, or as you say a buck-boost through an LDO - I'd try to implement the latter, not sure how much important/problematic switching noise, if any, slips through the LDO 'though.

3) Current sense/shunt amplifiers make things a lot easier (so long as output offset isn't going to be higher than the lowest current you need to sense - I "got my fingers burnt" with that issue!).

A great deal of this, referring to question 3, without expert input to give sager and more useful advice, sounds like putting off starting, David, get on with it and find out what's right and what needs further development: you don't know till you try. Again, my test circuits have to be built in stages/subcircuits because each subcircuit is a world of learning that can be effortless and take a day or two to prototype and have correct quickly or more usually a slow, painful month of seeing I haven't understood some of the theory I thought I had, need to read up on a "tiny detail" that expands into a whole new area of needed knowledge and have to go back to the books for another month and keep trying until the first rubbishy mess works as required, but until I've actually tried, the theory always theoretically will work as expected...

Testing rough versions is good as DMMs and oscilloscopes can teach a lot quicker than too much research - real live number facts show where you need to review/re-design a lot quicker sometimes, as you know.

What is ADD? Anyway: Focus, dude - what is necessary that you must do now to get to the next place in the project? If you focus on everything at once without focussing on anything specific you will jump around and get nowhere and not know where to start. Choose a point and stick to that one, make a checklist of circuit parts and until each is "done" don't let yourself get distracted. I like reading, but I have to control: "Oooohhh, that's an interesting topic..." otherwise nothing gets finished, or finished several months later...

I admire you, you've taken on what looks like a little project, but it's obvious that it's actually a big project with several circuits and complex concepts rolled into one device, that you have to fit together, so I wouldn't worry if it's taking time for all the parts to fall into place - all good things in time, and fools rush in where angels fear to tread..., eh .

Sorry I can't offer any practical electrical advice.

 

[BradtheRad]

I know that some of my thoughts would be answered by a test circuit but I feel as I should go through it all and not build the test circuit before I can implement all the stages that would be required for an actual control loop, though I am not that far of I believe.

Certainly you will construct an excellent control loop! The amount of thought you're putting toward it is a sign of your ability to make the control loop a success. And it's okay for you to think the project through, in its concepts, in its flowchart, how to apply feedback, selection of devices, power source, the connections, the logic, the layout, etc. Good planning is all a part of getting the project to work properly. However experience shows that there's usually some unexpected behavior which pops up after construction. Of course that's part of the fun when we construct a project, to overcome the challenges. Right now you seem to sense that a test setup is a good idea. I agree this would give you a better idea whether the final result is possible, namely to push enough Amperes through a coil to heat it to a certain temperature, and in a package that fits your hand. Only tests on real components will reveal whether it's in reach.

 

[David_]

I had come to feel rather unsecure about the hole thing but I now feel quite happy and more confident

I have a fear I guess you could call it as well as a question or rather I am looking for your thoughts even if it is simply guessing if no one knows.

As I read about digital control of DC-DC converters I come across things that makes me concerned and sometimes fearful due to the fact that I have no idea about the things spoken of, but I have to imagine that there are techniques for implementing the control loop to regulate the converter that are far more complex than others. What is it I fear? I really don't mind finding out that I have a large amount of research to do or new subjects to learn, but what I fear is that sometimes I experience that although I have to will to learn more i can't find any resources to enable me to learn more.

The most recent such thing I came across was someone talking about using Laplace-transform in order to transfer the... values is it...? whatever is transferred it was from the time-domain into the Laplace-domain which apparently makes certain thing much easier. In that text Z-transform was also mentioned, which I don't know what it is but I suspect that it can be in some manner or another related to Fourier or Fourier transform. I have read about Fourier but I found in the end that I couldn't grasp it by reading online on my own and that I need to find someone to whom I can talk about that in person, but I do get some aspects about Fourier transform and I do know a little about what it is used for, as for Laplace I know less and all I really know is that is it used in circuit analysis.

But to the question, do you think that I need to keep reading(along side making other tests and measurements, I can't recall if I wrote about it or not but this time I actually have a reason for putting things of since I am in a prolonged process of switching living location and so my equipment isn't in order, usually though I do have tendencies to put of doing things for no reason) about the more complex things mentioned above in order to learn how the professionals are doing this or do you think that I should go ahead and try to implement my own naive control loop scheme and see what happens?

I am pretty sure I should go for my own naive implementation first but I still want to ask in case this would have a sort of definite answer like "that will never work out, you need to study up on how to do it correctly".

Sorry for the lengthy post but I have another matter that is more reading than answering I think.

I have come to know that usually what is done is that one builds a model of the converter using software's such as Matlab(which I do have a student license for, Mathworks was kind enough to allow me one even though I am not a student) by applying Kirchhoff's current/voltage laws to the schematic of the converter to derive equations that are used to derive the transfer functions or something like that.

What I don't understand is if that is done in order so simulate the converter so that problems can be observed and component values adjusted to give the desired response or does that model once constructed have anything to do with writing the software to control it(the software that is to be running in the µC, more specifically the algorithm that interprets the feedback and regulates the output of the converter?

 

[BradtheRad]

It looks as though you're forgetting to find the fun element, and instead finding the fearful. It's okay to be leery of not getting it right the first time, or of blowing up components, or failing to achieve a perfect blend of software and hardware. Still, these should be outdone by your own vision of yourself making a successful project, and saying it's yours and you built it.

To tell the truth one thing that I would be nervous about is handling 40A. It's a jump to a different realm, not that easy to manage. A big demand on your batteries. Therefore I would take gradual steps up to it. Experiment with small models.
Say 40 gauge copper wire (or nichrome wire if you have it.).
Run a few A through it, see how hot it gets.
Develop a way to measure its temperature.
How to switch current with a transistor/mosfet.
Add software control.
Etc.

 

[David_]

Yes that is a good thought to keep in mind, today is actually the first day that I feel as I am capable of deriving a model of my converter with Matlab/Simulink with the resources online(although with a lot of work but after having overcome the biggest difficulties that task aught to be great fun), and I want to both experiment hands on with the circuit as well as deriving a model for it(the modelling comes second).

This is a wheel I feel is well worth re-inventing since it would put me in touch with the very fundamentals of a switching converter.

However I find it difficult to split this project into separate stages.

1.
While looking for current sense amplifiers I could not avoid finding my self at LT's site since there current sense amplifiers go down to 10µV Offset voltage, but the ones that has such low offset mainly LTC6102 if I recall correctly only has a bandwidth of 200kHz which would not be enough if I would switch my converter at 100kHz. The ones with higher bandwidth(1Mhz) has around 200µV offset voltage but also has 1 or 2 comparators with an internal reference voltage which is suggested to be use for under-/over-current indicators by setting a over-current value and then have the comparator output notify the mcu through a external interrupt. Which sounds useful, I just haven't determined if it is useful in my application, the over-current value is set by a voltage divider on the output of the current sense amplifier(which outputs a current) and I haven't found a way yet but it would be neat if one could find a solution for how to adjust the over-current value, I guess a digital potentiometer could do the job.

2.
I am also considering changing my circuit to a buck converter and limit the range of coil resistances I can use, I would need to order SS316L wire from the UK since I can't get thick enough SS316L in my own country(max 0,5mm) and I need 0,7mm if I was to wind coils that add up to a maximum of 0,12Ω if I recall correctly. But I am still unsure about this because on one hand it would make the hole design easier but also limit it's usable range if one want high power.
But I am still not entirely onboard on this idea, since I will have to discover as much as I will need to about switching converters I don't really see that reducing the circuit to a buck-converter would change it all that much, but I might be wrong about that.

3.
A new situation has come up, while visiting my local vape shop I got to try another kind of atomizer which is very different from the ones I usually use. It had an incredible amount of taste, so much more that I need to get me one of those. However such an atomizer is made to be fired with a maximum of 17W and with such a large decrease in power I am just thinking that there might be a case to design for two ranges of current sensing in order to increase the control at both high currents and low currents. Since I don't have a selected solution for sensing current done yet I can't check the values but as soon as I have done so I should know.

Also I would actual not consider my self as a person whom should be messing around with a 40A converter, but although I am designing it all for use at 40A I will not even go close to that value before I have thoroughly verified the design at lower currents.
If I had the design completed today and could start to use it I would use settings that would result in an output current of an average 14A, and I don't know if I will ever use the device at 40A but the 40A limit could be said to be the metaphorical limit on the amounts of fun the device can deliver.

By which I mean that if I where to want to have fun with this device I would play around with new types of coils and the more fun coils are always lower resistance, so the maximum current limits the resistances that can be used, that and the minimum voltage but that isn't as big a problem as the high currents.

My currently used device can only use coils above 0,1Ω(a coil build that is 0,1Ω does not work) and I want to enable the use of lower resistance coil build such as 0.025Ω(1V at 40A). That fact together with the battery configuration options made me choose 40A since two 18650 batteries in parallel should be able to deliver 50A, so then I set it to 40A to be safe.

 

[David_]

So I am having some concerns about my circuit, I generated the following picture with a LTspice simulation but I simply can't believe that it is correct, take a look at this:


It was a while since I took that picture but I ran a simulation based upon the target, which is an output current of 40A...

There is something weird going on isn't it or does this topology create such insane current spikes?

This simply has to be a simulation quirk, good luck finding the Schottky diodes for that supply...

- - - Updated - - -

M1 is the buck MOSFET and M2 is the BOOST MOSFET, although now it occurs to me that these values could be real since apparently in a non-inverting buck-boost converter the power components are experiencing I_IN+I_OUT.
This could perhaps indicate that I really need to avoid operating in buck-boost mode and restrict the operation to buck mode or boost mode, however I am pretty sure this picture was taken during a boost-mode simulation.

But still 180A-200A...!

I'd might as well add that I believe that I am closing in on being able to implement the power converter stage of the system to find out what it will do in the real world.

- - - Updated - - -

Ok, now I see some short comings of that picture as it is apparently ramping up but is not allowed to peak.
I'll try to make a new one showing the complete process.

 

[FvM]

Posting simulation waveforms without the simulation circuit, or better the complete simulation files seems useless to me.

 

[treez]

There is something weird going on isn't it or does this topology create such insane current spikes?

Are you sure you select correct voltage rated diodes and fets?.....ltspice used to let you get away with this but not any more

 

[David_]

Posting simulation waveforms without the simulation circuit, or better the complete simulation files seems useless to me.

Yes of course your right, I would have made a new post already with the circuit schematic and the simulation file but I seem to have confused my self with a few similar simulations, but I am working on reproducing it so that it is possible to discuss it properly.

I am pretty sure that I have made proper component selections but I will double check that and return when I have a new simulation.
One thing though that I'd like to add right now is that I have had problems in getting the responses I have been expecting when I vary the buck-MOSFET and the boost-MOSFETs duty cycle and I couldn't understand why, but I hope I will return soon with all the info.

 

[David_]

I thought I would do the simulation as realistic as possible and as such select diodes and MOSFETs that have available SPICE models but I don't understand the current ratings of schottky diodes, I am looking for diodes that are in stock at mouser.com but I'm quite confused about which current rating to use.

On one hand the current rating can appear to be 90A but when I open the datasheet the
"Forward rms current" is = 90A while the
"Avarage forward current, δ = 0.5" = 30A

So what is it?

Further more sometimes the rated current that they use to advertise the diode on mouser or other sites is sometimes the current that 1 out of the 2 diodes in one package can handle while other times the apparent rated current is the combined current of the 2 diodes in one package, this isn't the first time I have tried to find a properly rated diode but then failed because I become so confused about all this stuff.

- - - Updated - - -

What I need is a diode that can manage to pass 40A in a converter with an 5V output.

- - - Updated - - -

And what is the deal with the reverse current, I have read documents that claims that the reverse current rating needs to be weight against the forward voltage drop but I find that the ranges of reverse current is puzzling since some have ratings of over 1000mA, others have between 100mA and 600mA and then there are those that have values such as 165uA...
Which makes me to look for the uA range but that is perhaps not necessary but when I can apparently choose between 1000mA reverse current and 200uA reverse current then why would I not go for the 200uA reverse current diode.?

 

[treez]

because you are looking for very high current diodes, the rms rating is coming into it aswell. You have to keep to the average and the rms rating.
Reverse current of schottkys is worse than other diodes.....so as you say, choose the lowest. Reverse current varies with reverse voltage, so if you dont have much reverse voltage then ot wont be too much pf a problem. Rev currnt also gets worse with temp, but it sounds like you are looking at pretty monstrous diodes, so things
may be a little different.
i wonder if you woudl be better paralleling sic diodes, (sic can be paralleled), for your high current rating.

 

[David_]

I see, but about the paralleling.

Reading online shows a pretty divided opinion, some say that as long the two diodes are in the same package then the temperature difference which might set of the thermal run away while others say that it can probably work but you shouldn't. And I have also read what treez is suggesting paralleling sic diodes(which is the option I will investigate next).

The only sure thing appears to be to parallel Schottky silicon carbide diodes, but one aspect remains unclear to me.
In discussions regarding paralleling ordinary schottkys or other diodes I have read opinions that each of the diodes used need to be able to handle the full current just in case.
But if SIC diodes would be used is it then valid to parallel 3 20A diodes to get one 60A diode(disregard margining)?

- - - Updated - - -

I have from the beginning decided against using two more MOSFETs instead of the two diodes, but it is starting to look as going with MOSFETs might be a good idea.
I didn't want to do that since it would make things more complex.

- - - Updated - - -

It wouldn't be space effective but three of these could then be used.

 

[treez]

But if SIC diodes would be used is it then valid to parallel 3 20A diodes to get one 60A diode(disregard margining)?

yes if they have a positive temperatue coefficient, and i am sure sic schottky diodes do....avoiding getting into margining as you said

- - - Updated - - -

are you sure that SBR20E100 diode is a fast diode?......its rated 100v, and high current, and at that voltage and current, fast diodes usually have a trr rating...even schottkys have a trr rating when they are 100v ones....there is no trr quoted in that datasheet, so i suspect its a slow diode.

By the way, if you buy a diode which is said to be two-in-parallel internally then you can assume they share current equally, because the doping process in the fab house will have assured this.

 

[David_]

I thought I posted a post asking about whether or not I should start considering using MOSFETs instead of diodes and if anyone have any resources to learn about the important aspects of using such circuits(but it was never actually posted), but then I found some information that indicate that even if I use a MOSFET instead of a diode a diode is still needed in the circuit.

Anyway I hope that those thoughts is irrelevant because I think I have found a diode which could work, but I would like to ask for your opinion if it is viable.
The datasheet is here, and it looks really good. It's a diode designed for switching power supplies, has 100A rating and around 320mA reverse current and it can withstand 45V DC.

- - - Updated - - -

And a forward voltage of 0,71V which is higher than I had hoped I could find but it is at least better than the approximately 1,3V that most SIC diodes seem to present.

- - - Updated - - -

Maybe it is unnecessary to ask if it can work since obviously it is made for this sort of application.

 

[treez]

looks good, as you know its just a case of checking if it will overheat or be overvoltaged, as you know, schottkys dont like any overvoltage, even a bit of overvoltage ringing will kill them.

 

[David_]

I have found a better diode but it's datasheet says that it is optimized for OR-ing of parallel power supplies, though it also says it is for "High frequency operation". I have looked through the datasheet but can't find anything else that seems different with it but I don't understand the difference between a diode optimized for high frequency switching power supplies and diodes optimized for OR-ing parallel power supplies.

Is there really a distinction between them?

 

[treez]

yes, for a high switching freq diode you would want it to either be a "schottky", or have a trr less then 50ns or so.

If it didnt state either then i would be suspicious and back off away from it.

 

[David_]

I see, well it is called a high performance schottky diode so if I understand you correctly it might suffice. But I feel as it is safer to go with the other one specifically specified for a switching power supply application, although the power dissipation quite a bit larger something like 18W for the one specified for OR-ing application versus 22W for the one specified for SMPS.

But onto a very different subject, earlier we have talked about MOSFET gate-charge and the current required to drive particular values of gate-charge.
And the equation for calculating this involves the transition time, which I took as to mean 1/f_SW(switching freq) but I question that now.

I fear that my calculated current might be quite a bit lower than the actual current required based upon the idea that transition time might not be equal to 1/f_SW, which I think might be the case since wouldn't the transition time need to be quite a bit faster than the switching frequency?

the equation is:

I_G = Q_G/t_transition

And if I am using a switching frequency of 100kHz which equals 1/100000 = 10µS then wouldn't the gate voltage during those 10µS need to first rise quickly in order to turn the FET on, and then stay on a while and then go back to 0V or below the gate threshold voltage to turn the FET off again?

So if the transition time was based upon the current that is available to supply the gate, which was determined based upon turning the gate on in 10µS, couldn't that make a 100kHz switching frequency impossible to attain?

- - - Updated - - -

I haven't been able to find a clear definition of "transition time" which makes me wonder this.