To Boost Or To Buck?

[iusedtoanalog]

I have been building mods for personal use now for a little over a year. My first mod was built with a Ti PTR08100, Followed by OKR T6 buck modules. My first unit (which wife still uses everyday) is running 14500 batteries, My mod is running imr 18490 batteries. I have a good handle on the complete run time of both units, mine runs from ~7:30am until about 10:00pm on a single pair of  aw imr 1100mAh batteries.

Now I am ok with the overall size of my unit, but I would like another three to four hours away from the charger.

So Now enough rambling. I am going to build another unit. I am going to include on-board charging, and a hobby cell with 1100-2000mAh capacity. My question to all of you who have used both the boost and the buck chips is this. If I typically get 12-14hrs on the okr-t6 with the 18490 imr cells, would I do better with the Ti 04050c chip or should I stick with the okr-t6 in relation to runtime? I have heard the boost chips are typically harder on batteries than the buck chip, so how drastically does this affect the use time of the battery?

I typically vape between 5-5.6volts, on a two ohm coil, Or there about lately.  Thank you for any input you may be able to provide to help me make up my mind. 

[CraigHB]

Buck is better than boost.  The only real advantage to boost is that it allows you to use a single cell and easily incorporate USB charging.  Buck modules are more widely available, more efficient, have higher output capability, and have a wider voltage range.  Buck reduces the current load on the battery where boost increases the current load on the battery.  That being the case, you're limited to a "lower energy density" unprotected high drain battery with a booster where you can use "high energy density" protected cells with a buck converter.

Anyway, the only drop-in boost module I've seen that's usable for en e-cig is the TI one and it's pretty limited.  You have to use a hack on the feedback circuit that Breaktru came up with to get the right voltage range and it can not output voltage lower than input voltage.  That's normal for a booster.  Plus, output is limited to 12W, at least according to the data sheet.  You can get well over 30W output with an OKR-T/10 using a wide range of battery options, even with a couple protected 18650s.  A 36 Watt mod with 6 amp-hours is nothing to sneeze at.  If you keep output under 15W, you can use a couple protected 1200mAh 18350s for a total of 2400mAh which aint bad at all for the size.

If you're intent on plug-in charging, it's actually pretty easy to incorporate into a dual cell mod.  It's just a matter of using a balancing connector instead of a USB charging module and a balancing charger instead of a USB adapter.  Here's one for $12.

Granted, I only build booster mods myself, but I also design the whole thing from scratch so my boosters don't have all the limitations of the TI module.  I do single board designs and they're actually easier with only one cell.   I'm pretty intent on using a USB charging connector as well.  If I did a dual cell mod, it would have a balancing connector for charging and non-removable cells.  I really dislike removing and replacing batteries.

[redwolfe]

In my personal experience, and from what I have learned, the cell run time will come from the amp rate being pulled from the battery and the ohm load of the atomizer.

For example, on my Darwin, which uses 2 900mah cells in parallel and uses a boost circuit rather than buck, I typically get 15-16 hours with a LR 1.5ohm 306 running at 10.8 watts, which according to the Darwin is 4.3 volts, and pulls approximately 2.86 amps.

On the other hand, if I used a 2.9 ohm 510 at that same wattage--10.8--I get 5.7 volts, which pulls approximately 1.9 amps. I seem to get around 24-26 hours with that setup.

Hope this helps out in any way!

[iusedtoanalog]

Thanks Craig, And Redwolfe for the replies. Deep down I already suspected the boost modules where just not exactly what I was looking for. I may build one or two of these just to use as backups or just to see if I can get away with using them. At least I dont have to keep searching for a super badazz case to house them in. Looks like a regular plastic box is in their future(if I even build them into anything at all). Thanks again guys.

[CraigHB]

If you really want to try using a booster that can compete with a buck module in output, you'll need to build it with one of the myriad of controllers available.  The selection is pretty wide for high output boost controllers.  For some reason, modules are limited in selection, but not the controllers for them.

Unfortunately, DC-DC converters built from scratch can be difficult to get running well right out of the gate.  They can be tricky things.  There is a number of architecture options with some being easier and more forgiving in design than others.  Generally, the higher the efficiency, the harder it is to design. 

If you don't care all that much about efficiency, you can go with an "asynchronous current mode" controller that's easy to build, easy to get working well, and does not require a tight circuit board layout.  The other end of the spectrum is the "synchronous voltage mode" controller.  Very difficult to design, but very high efficiency.

[iusedtoanalog]

Hello Again. Sooo...... If I where to entertain the idea put forth by Mr. CraigHB, Would I be looking for a controller like the http://www.linear.com/product/LTC1700 from Linear Technology? Or is that outside of spec from what I would need. I have a fair amount of electronics knowledge, and from what I can tell this controller is intended to run an external mosfet, which can switch very high amerage, but the missing link(for me) is this,  does this device simply control the switching frequency of the mosfet to allow the mosfet to carry the load? This device specs on their product comparator as being capable of 6volts & 6amperes of output, but in the datasheet everything appears ~5v....

Or I could have misinterpreted your post ,or maybe Im not entirely certain of what product would fit the bill of materials for a build like this but I would appreciate any reputable advice..... I may still want a single battery device if the limitations can be similar to what I am accustomed to... Thanks Again.

[CraigHB]

That's a good controller to use generally speaking.  I'd have to study the data sheet to determine maximum current output, but it should be able to achieve a reasonably high output.  The limit on switch voltage is 6.5V so it sounds like it can run at 6V without issue.

A DC-DC converter based on that controller is going to be a bit more complex than a traditional current mode controller.  It's good that it uses the voltage drop across the conduction MOSFET to sense current.  That eliminates the need for a current sense resistor which improves efficiency.  However, this complicates MOSFET selection since R_DS(on) must be low enough to ensure the converter can reach the desired current limit.  That along with the requirement for a low enough V_GS(th) and gate charge is going to make selection somewhat of a chore.

Another consideration is that controller is synchronous which means it uses a MOSFET rectifier instead of a diode rectifier.  That's good because it's much more efficient, but also requires careful MOSFET selection.  The MOSFET's body diode must be able to handle startup currents and gate charge must be low enough to avoid lengthy turn on/off times.  A slow MOSFET can cause excessive heating and timing issues for the controller.  Also, R_DS(on) needs to be low as low as possible to minimize heating and losses.

The controller switches both MOSFETs at a duty cycle required to maintain 1.205V at the V_FB pin.  The resistor values of the voltage divider driving the V_FB pin determine output voltage.  You'll vary the value of one of the resistors to adjust it.   

Switching occurs at a default frequency of 530kHz for that controller which is a good default.  My own boosters run at 550kHz.  It's a good tradeoff for ripple current, switching losses, and inductor size.  The MOSFETs operate in a complimentary fashion.  When the conduction MOSFET is on, the rectifier MOSFET is off.  When the conduction MOSFET is off the rectifier MOSFET is on.  The switching speed occurs at the controller frequency which would be 530,000 times a second if using the default.

Here's one that exemplifies what I was talking about before, really easy to design and build with a high current output capability and wide output voltage range;

http://www.linear.com/product/LTC1872

[iusedtoanalog]

I think I have found a great solution for the first option with the synchronous controller. Looks within spec of the controller, and its capable of 16a  according to vishay. http://www.newark.com/vishay-siliconix/siz700dt-t1-ge3/mosfet-dual-n-ch-20v-16a-powerpair/dp/83T3535 plus its a pair, high and low side packaged together.

[CraigHB]

It's dual N-channel.  For a synchronous booster, you need a P-channel and an N-channel.  Also, the P-channel is hooked up in reverse when used as a rectifier.

[utak3r]

Another problem I ran into, when I was trying to design my own booster, is the inductor - its size. In my application I needed 47uH inductor, capable of at least 3.5A current. If you want to buy it, the smallest one I found was ca. 5-7mm in each side. That's quite a big inductor... I have no idea how the hell they managed to make something like in PTN0504 thing, it's tiny...

[CraigHB]

With a switching frequency over 500kHz, a 1uH inductor can be used without issue  The bigger value inductors are better because there's less ripple current and the DCM/CCM  (discontinuous/continuous conduction mode) threshold is much lower.  For an e-cig, high ripple currents are not as much of a problem and you normally run with heavy loads (the atomizer) so you don't need a very low DCM/CCM threshold.

[utak3r]

Yep, my design was working with around 100kHz frequency... So, that's the way to improve   ok, thanks for the info.

[CraigHB]

The lower frequencies are actually better to reduce switching losses, but of course as you found, it requires a higher inductance value.  500kHz is a good trade off for inductor size and switching losses.  As with everything in electronics design, it's all about trade offs.

[miskol]

Another consideration is that controller is synchronous which means it uses a MOSFET rectifier instead of a diode rectifier.  That's good because it's much more efficient, but also requires careful MOSFET selection.  The MOSFET's body diode must be able to handle startup currents and gate charge must be low enough to avoid lengthy turn on/off times.  A slow MOSFET can cause excessive heating and timing issues for the controller.  Also, R_DS(on) needs to be low as low as possible to minimize heating and losses.

The MOSFETs operate in a complimentary fashion.  When the conduction MOSFET is on, the rectifier MOSFET is off.  When the conduction MOSFET is off the rectifier MOSFET is on. 

Hi Craig,

thanks for this info, i'm getting more understand of my current Boost circuit.

i believe it is an asynchronous voltage mode Boost circuit because it uses an Inductor, a switching MOSFET and a diode.

i think i want to build both version which is the synchronous voltage mode Boost circuit utilising an Inductor, a switching MOSFET (N-ch) and another MOSFET(P-ch) to replace the diode where both MOSFETs operate in a complimentary fashion. to produce the complimentary control signals for the MOSFETs, i will be using LM5106. as you mentioned, the P-channel is hooked up in reverse when used as a rectifier. up to this point, LM5106 is a possible solution to drive both the MOSFETs, am i correct Craig?

i'm hoping to build both circuits to the see the real difference in their performance and also as backup strategy if one of them fails ;p


Please help advise, TQ!

[CraigHB]

Yes, you should be able to use that part for buck or boost.  The difference between buck and boost in terms of the switches is they are on the opposite side of the inductor and the high and low positions of the switches are swapped.  The high side switch becomes the low side switch and the low side switch becomes the high side switch.  I've not actually used the LM5106 so you would have to test it on the bench to say for sure, but I don't see any reason why it would not work for boost.

A gate driver like the LM5106 is not really required for an e-cig power converter since voltages are low and switches with low gate charge can be used.  There's more space requirement on your PCB for both a controller and a gate driver.  PCB space is usually a valuable commodity for an e-cig.

I would suggest you look at something like the TPS43061.  This is a current mode synchronous boost controller from TI that is easy to work with.  You can simulate it easily using the PSpice model provided by TI.  It has fairly powerful gate drivers on-chip.

You can alternatively go with an asynchronous controller if you're not particulary concerned with the higher loss of diode rectifier instead of a MOSFET rectifier.  There's generally a wider selection of those controllers and they are easier to work with.  Though, a diode rectifier becomes a big source of power loss at higher outputs.

Even though you can utilize a switch with higher gate charge using a stand-alone gate driver, gate charge is big source of power loss so it should be minimized as much as possible.  That kind of defeats the point of using a separate gate driver since the only reason to use one in this case is to allow higher gate charge.  That's something to avoid for optimal efficiency.

I would recommend you use switches like the TI CSD17501Q5A or any other switch with low gate charge and low "on"  resistance.  It's best to minimize gate charge as much as possible without adding too much "on" resistance.  There's a trade-off there, lower gate charge, higher "on" resistance, lower "on" resistance, higher gate charge.  There's a balance depending on output requirements.  For higher ouputs you can lose more power in switch resistance.  For lower outputs you can lose more power driving the switches.

Current mode has that extra sense resistor which makes things easier in terms of tuning the converter, but also introduces an extra point of loss.  You can use inductor DCR sensing with that TI controller I mentioned.  It's a bit tricky, but easy to tune in the simulator.  There's a paper on that topic in the listing for the controller on the TI web site.  For optimal efficiency, inductor DCR sensing is something you want to use in lieu of a sense resistor whenever you can.

Voltage mode is real PITA to tune so I would actually not recommend that type of controller.  It uses Type III feedback compensation which is an RC network of 3 resistors and 3 capacitors.  There's way too many poles and zeros to deal with mathematically or by brute force in the simulator.  Current mode just has one pole and one zero in the feedback compensation, way easier.

The TPS43061 is a QFN type package so it's hard to mount on a PCB.  Not an issue using reflow, but difficult to solder by hand.  TI doesn't have a real wide range of boost controllers.  Linear has a wider selection of boost controllers and you can use their circuit simulator with their parts.  Though I think the TI stuff is better engineered.

You can use that LM5106 gate driver if you really want, but it's not an optimal solution given the range of chips available for this kind of thing.

[miskol]

I would suggest you look at something like the TPS43061.  This is a current mode synchronous boost controller from TI that is easy to work with.  You can simulate it easily using the PSpice model provided by TI.  It has fairly powerful gate drivers on-chip.

i had been working with the LM5106 for a few months, it was working nicely up until from learning much from you i've gotten more understanding now that my previous design is a buck converter (because of the Inductor's position)... maybe my next design is a closure on the use of LM5106. as you said, it is much easier to used market-ready Controllers such as suggested by you, the TPS43061 and one sample i just ordered, LMZ31710 (have built-in MOSFETs, inductor).

thanks for the suggestion Craig. TPS43061 is a nice, simple and possible for a small footprint design. unfortunately TPS43061's min input voltage requires two-cell batteries.

Voltage mode is real PITA to tune so I would actually not recommend that type of controller.  It uses Type III feedback compensation which is an RC network of 3 resistors and 3 capacitors.  There's way too many poles and zeros to deal with mathematically or by brute force in the simulator.  Current mode just has one pole and one zero in the feedback compensation, way easier.

regarding the Type III feedback compensation which is an RC network that you mentioned, this is feedback for output voltage? so that we can program the MCU to fine tune the boost output?

as per the OSC image, i'm not happy with the voltage signal, this is using 30kHz. the output signal will be better if i either increase the freq to 500kHz OR increase the output capacitors to 800uF. i'm sure the latter option is not recommended.

[CraigHB]

regarding the Type III feedback compensation which is an RC network that you mentioned, this is feedback for output voltage? so that we can program the MCU to fine tune the boost output?

Indirectly, yes.  Since you would use an MCU to control a digital pot, you can set up an outer control loop where the MCU samples output voltage and adjusts in code sending commands to a digital pot.  This is normally required anyway.

All converter controllers use a closed loop feedback system to regulate output voltage.  You always need some type of compensation in the feedback.  This is due to the phase shift that occurs in the LCR network formed by the converter's inductor and output capacitors.  It's a naturally unstable system and can resonate like a bridge in the wind.  Current mode requires only a simple type I network (one pole one zero).  Voltage mode requires the much more involved type III network (two poles two zeros).

My boost design uses a voltage mode controller (TPS43000) and I spent way too much time tuning that thing, though it works really well.  In the future, I'll try to stick to the current mode controllers. 

Using a current mode controller with inductor or switch DCR current sensing introduces no additional loss and it's much easier to tune.  I took me a few hours on the simulator to tune the TPS43061 converter I built recently and it took me probably weeks on the bench to tune the TPS43000 based converter I've been using for some time.  It was way too time consuming.

[Visus]

Craig

Wondering if that new simulator that helps with that tuning at Ti would have helped..  Unfortunately it didn't come out until last week but did ya take a look at it post build and did it have what you came up with?   

[CraigHB]

If you're talking about TI's Web Bench, I've used that before and it's helpful for zeroing-in on designs and also the new reference design feature is nice.

Web Bench does not actually simulate circuits in real time.  It can chart some characteristics and provide lots of pretty graphs, but you need real time response for all nodal voltages and currents to handle the design of these converters.  Only a SPICE simulator can do that.  TI does provide models for PSpice and also another SPICE simulator called TINA.  That's what you have to use to tune component values. 

It's actually possible to design this stuff all on paper just using the math, but you damn near need a PHD to do it.  The math is ridiculously convoluted and even if you can do the math, it's ridiculously time consuming.

TI does have tons of reference designs and in some cases, you can simply build the design as shown with little to no modification.  It's like having your hand held the whole way.  You don't even need to know much about circuit design to build those, just good PCB making skills.

[Visus]

Yeah I wondered how much intuit it was into it, as it says, "a huge time saver."  I like that holding my hand part.  Dang was hopes that it would save ya some time  to tune, but makes sense,  simulate and  tune on the fly perhaps using their suggestions but for different applications would need real world apply.  Never as simple as want, sometimes it is..

Im running linux badly and having issues running sims/cads,  I tried when you first posted the free utility sims ltspice etc,  my windows side of puter is xp and it keeps locking up..  lol  One day ill buy another puter and play with the sims and such.  I refuse to buy a puter after spending huge for em in the past and they being worth notta today, until absolutely necessary.  I was a test mule for windows based studio programs, they were trash at first then blossomed.  Was stubborn not to buy a mac and use protools cause the d/a converter interface I needed with my setup was 10k alone for the cheap one lol.  Its still hella expensive today..

One day soon bigger brains 

[CraigHB]

Hey, I know what you mean.  I have a Windows 7 installation all ready to go, but I'm not going to use it until I have to.  At this point all my stuff runs on XP SP3 so until something comes along that forces me to use a newer version of Windows, I'm going to keep using XP.  I'm sure people might find that ridiculous, but I'm so sick and tired of the forced upgrade merry-go-round these software makers put you on I'm going to fight it as much as possible.

In any case, I have no issue with LTSpice on XP SP3.  I like that one a lot.  It's really good for quickly checking simple circuits.  It's the fastest and easiest simulator I've used.  Though it has an unusual interface, but once you get used to that it's super fast and easy.

I have a licensed version of OrCAD PSpice I use, but that' not something a hobbyist would normally have access to, it's pretty expensive.  I got it through work. 

TINA-TI is free and unlimited so that can work out well.  If you look on TI's website, they have a link to it, let me look, here it is;  http://www.ti.com/tool/tina-ti

If you have no issue running Linux there's lots of free tools available for that.  Pretty sure there's a good SPICE program you can use, but I haven't used a Linux based SPICE myself.  There's also gEDA which is a good PCB design suite.