Digital sync. buck-converter project

Go To Last Post
5 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hello.

 

I am developing a project, I was going to write an introduction post but I will post that later due to me having written the hole introduction but then loosing it in a browser crash and I simply don't have the attention to re write it right now.

 

But what I am developing is a synchronous buck-converter controlled by a XMEGA128A1U, possibly a smaller XMEGA A series depending on whether or not it is to my advantage or not to have access to two separate ADCs and DACs. It is powered by two series 20700A Li-Ion batteries(they are 1 or 2 millimeters larger than 18650 batteries), however I have not yet determined if I need to measure the voltage over both batteries or if it is sufficient to measure the resulting 8,4V - 6,4V(they are charged with a external charger).

 

I am trying to determine which pins of the MCU to tie to which part of the PCB so that I can make the prototype in order to start developing the software, and I wonder about the following.

 

I have the need to measure the:

  • battery voltage/buck input-voltage(ether the voltage across both batteries individually or across both combined)
  • buck-converter output-voltage
  • buck-converter output-current
  • the voltage across the load as part of a load-resistance measurement(this might be done with an external more accurate ADC)
  • 1 or 2 voltages representing the temperature of the circuits

 

The buck-converter output-voltage and output-current will be used to regulate the converter so they will be done most often and I wonder if it could be beneficial to measure each with separate ADCs in order to be sampling those two values at the same time?

 

Would there be any difference at all compared to measure them with two ADC channels on one ADC, in succession?

 

When I try to make these plans I first considered using ADC A for measuring the values relating to the converter regulation and using ADC B for measuring non-time critical values such as battery voltage and the temperature voltages, but then I thought maybe I should use ADC A for measuring the buck output-current and ADC B for measuring the buck output voltage and the remaining measurements would then be shared amongst ADC A &  B...

 

The converter is switching at 100kHz by the way, there is clearly differences between the requirements of the time in between measurements comparing the output voltage/current and the battery voltage. It would perhaps be convenient to use DMAs for making the time-critical measurement with ADC A and then make the rest of the measurements with ADC B.

 

Though I have forgotten whether or not I had determined how often I should sample the buck output voltage/current, I'll look that up.

 

Any thoughts on this would be interesting to read.

 

Regards

 

 

  •  
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Is the converter output voltage somehow different to the voltage across the load?

Where are you measuring the output current? Through the inductor or through the load?

Your control loop frequency is much lower than your switching frequency, so if you want to do cycle by cycle current limit, you'll need to sample the inductor current at 100kHz in a tight window of time. Normally one would use a comparator.

If you want to measure voltage and current at your loop frequency, then consider the error the difference in time might make. You can always use sample and holds to sample a number of values simultaneously then convert the values at a later time. Then you have a snapshot of the values at the sample time.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I guess it shows that this is my first experience with these sort of things, I have not been able to think straight lately so I will probably be delayed even more addressing your response. But I will for sure be back as soon as I can.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I think I was hoping that the situation would be much more easily answerable, but your questions really shows the necessity for posting a hole description. To avoid loosing the text again I am writing this in a text editor but any pictures/schematics mentioned is attached if not directly included in the post.

 

This will be one hell of a post(lenght), which I apologise for but if read you will be presented with the hole picture of this project and it's goals end requirements, in any case the first segment will deal with describing the device and it's operation while the second segment deals with the hardware and schematic.

 

Operation:

The characteristics of the power supply is best described by describing it's purpose which is to be the driving circuit in a electronic vaporizer(aka e-cigg), without going into too much details such a vaporizer consists of two separate parts/products. 1, the so called MOD which is the power supply and control device and 2, a atomizer, which is a metal structure to which hand made coils is attached(there are many sorts of atomizer, called things like RBA, RDA, RDTA, clearomizer, tank etc), then cotton is placed through the coil(s) which are then soaked in a liquid containing amongst other things nicotine.

 

So this digital converter needs to operate in one out of 2 user selectable ways, ether it is used in Variable Wattage(VW) mode or in Temperature Control(TC) mode.
In VW mode the user sets a value for the output wattage and then when the user presses and holds down the activation button the converter shall supply that amount of power to the coil(s)(through the atomizer). whatever mode is used there needs to be a idiot proof 10 seconds maximum activation duration in order to prevent things like accidental activation in pockets or bags or whatever.

 

The TC mode is for sure the more interesting mode of operation from a design point of view because in TC the user shall be able to specify a temperature which depending on another user define parameter shall be maintained, that other parameter determines the ramp up behaviour so that one may make the device to slowly ramp up and not overshoot or quickly ramp up while tolerating temp over shoot in some manner.
To be honest I haven't thought through that other parameter yet but it will exist some parameter users needs to define other than temperature.
The temperature is only approximated through an algorithm that first needs to measure the coil(s) resistance at room temperature which is stored for later comparison to the new resistance of the coil(s) as it is rising due to the coil(s) metal rising temperature. It happens to be such that I only use and am only concerned about the materials SS316/SS316L, but unless I am mistaken this algorithm would be adaptable to any material through adapting the coefficient for that particular metal assuming its temp coefficient is large enough for the resistance measurement circuit to notice the changes.
This is also the largest task for me to figure out but I am sure I will be able to make it at least sort of work... This will definitely be a very interesting code to write and I have begun to research temperature regulation(digital) and have started to try and make sense out of Laplace-, Z- and Fourier-transform.

 

Here is a picture of a commercial device and an atomizer attached:

 

As a user interface I will use a 0,96" OLED display 128*64 pixels used standing on it's side so that the display is 64 dots wide and 128 dots high, which would look kind of like the picture above.

I am not yet certain about how I will make the buttons work, most commercial devices have 3 buttons, one for activating the device or to make choices in the menu along with two other buttons that are used for navigating the menu or increasing/decreasing settings.

And I have planned to use such an arrangement my self but I might add more buttons, I might use capacitive buttons as such buttons can be very neat as I will 3D-print the case for the device.

 

So now you know what the design is supposed to function like in broad terms, one of the main aims of this design is to give me a TC mode that is superior to the TC of commercial devices as I am not pleased with them, and I believe that I am in a fairly good position to be able to accomplish that since I believe that the main factor degrading the TC of the commercial devices is the fact that they are performing all the measurements at the lowest possible cost in order to make the device profitable and commercially viable.
I however do not have that restraint and by allowing my resistance measurement circuit to cost more that might present me with a much better starting point for regulating the temperature. While using the device on the picture I can clearly feel the device ramp up and then start to cut the power on/off to keep the temperature close to the set value, but it does so in a very slow and uncomfortable way and I have to believe that it could be done much more comfortable if the required circuitry is present. I am also using a higher switching frequency, although I might be wrong about all this.

 

 

Hardware:
This device is powered with 2 series Samsung INR21700-30T Li-Ion batteries(charged externally), they have 3000mAh and there is a e-cigg forum where a guy who have a long professional experience with batteries and a serious setup have evaluated and published data of a lot of different batteries and in his opinion these batteries are safely rated as able to deliver 40A and assuming the batteries are drained as in an e-cigg application then 60A is safe. Link to results for INR21700-30T.
And it's discharge voltage is specified as 2,5V so the available voltage will start out at 8,4V and the device shall stop operating when the supply voltage reaches 5V

The batteries(cells) voltage is measured to keep track of the remaining charge however I have not yet determined if I need to measure both batteries individually or if it is sufficient to measure the combined voltage given that they are both charged and discharged together(external charger).

 

Anyway they are powering two synchronous buck-converters, one low-current buck(TPS62135) which is followed by a 3,3V linear regulator(REG101)(it will be explained), these are powering all the control circuits including the power stage of the other to be described converter.

Then there are the high-current "main" buck-converter which is powering the atomizer, I am going to make good use of a IC called CSD95472, it's a "Synchronous Buck NexFET™ Smart Power Stage" containing both the high-side and low-side(synchronous) N-channal MOSFETs and the MOSFET drivers, requiring only 1 single PWM signal and it takes care of the dead-time and everything.

 

"assuming the batteries are drained as in an e-cigg application" refers to the fact that a vaporizer like this is only operated for about 2-8 seconds a couple of times each hour, it will of course vary a lot depending on the user but I will be able to violate rules concerning to thermal management as they pertain to ordinary DC-DC converter designing as this high-current buck converter is never going to operate continuously. If compared to a continuously operated converter this one will hardly be operating at all.

There are precious little information about this device(due to some legal stuff TI is tied to with some other company and I am not able to sign the required documents that would allow me to see that information) but there are a device called CSD95373 that appears to be more or less identical to the CSD95472 except that it doesn't contain the current sensing circuit that the CSD95472 does, otherwise it is the same package and the little information that is present in the pubic datasheet for the CSD95472 matches the information in the datasheet for CSD95373. Since I won't use that internal current sensing then why wouldn't I use the somewhat cheaper CSD95373 instead? Well according the the datasheet CSD95472 have quite a lot better thermal resistance values for some reason. But the datasheet of CSD95373 may be used to get more descriptions about some of CSD95472's functionality.

As for the low-current buck-converter followed by a linear regulator.
The CSD95472(power stage) requires a supply voltage between 4,5V and 5,5V while the XMEGA and any other additional circuits run of of 3,3V, so two voltage rails is necessary. But the TPS62135/TPS62136 have a cool feature. It has an extra feedback pin, a 3rd feedback resistor is connected from the ordinary feedback pin to this new feedback pin which is internally connected to GND through a FET, and that FET is controlled by yet another pin.

 

So then I thought that I would choose resistor values such as the MCU can assert control over the low-current buck and make it output 4,5V or 5V when the power stage is required to function and during all other times have it instead output something like 3,5V, so the power dissipated by that linear regulator is minimized. First I thought that it might not make any notable difference in the battery time due to the insignificance of the lost power compared to the power drawn when the device is firing but the fact is that the device is spending the wast majority of the time not using that high-current buck-converter. Commercial devices have a shut-down function but I don't think it is a shut-down function as much as a power-down function which I see as useless if there are another simple "button-lock" feature since isn't it simply logical to implement the most extensive power-down state possible to be used during all possible instances of time, due to no feature being time critical in anything than in a human time perspective?

 

As to what cartman asked.
I initially counted upon there being a point in measuring the voltage across the atomizer as close as possible to the atomizer connector apart from the converter output voltage measurement but now I don't believe there is a point in using two voltage measurements and I will plan for measuring 1 output voltage and 1 output current.

I have never done anything like calibrating out the impact of parasitic resistance but since the parasitic resistance inside of the atomizer might vary between different atomizers perhaps I could include some calibration feature in which the atomizer would be shorted with what is assumed to be a 0 ohm connection so as to revile the internal resistance of the atomizer and then subtract that value from the resistance measurement?

 

I have a couple of issues with the schematic that I would like to address, but I can't think straight about that stuff after having written this but I will return shortly. One of which is about the reverse battery polarity protection which I think is insufficient in the schematic I will include in this post.

But the schematic is a work in progress so there may very well be a few weird things and faults in it.

 

Regards

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Why the complexity for a simple application?  Surely you can just pwm a mosfet - the thermal inertia of the heater obviates the need for an inductor.