Breaktru Forum
eCigarette Forum => Modding => Topic started by: CraigHB on December 07, 2013, 05:51:24 AM
-
So I got this wild hair to do a buck-boost mod. I've only built boosters so far. I've been working out a design on paper and through circuit simulation. I just finished the drawing board design and have something that simulates well. Wow, what a challenge that was.
I'm basing this design on the Linear LTC3785 buck-boost controller. I've had my eye on that controller for quite some time, but never got into it because of the high part count. I find it to be a pretty amazing chip after working with it on the sim and finding what it's capable of. According to simulation it will have no issue supplying 40 Watts output across the full voltage range with the LiPo I'm going to use.
This design is way more complicated than what I've done in the past. There's so many challenges because I have do deal with switching noise on both ends. Input is noisy when it's in buck mode and output is noisy when it's in boost mode. Then it has this trippy in between mode where it's switching with all four MOSFETs and both ends are noisy.
I have to use a boatload of filtering and decoupling. I'm going to use a sixpack of 10uF MLCC caps on both input and output. There's no other way I could figure out how to keep the noise down and get the capacitance I need to stabilize the controller over the full range. The high value MLCC caps are too noisy and have too much DC bias. The smaller tantalum polymers can't handle the ripple at those high power levels and the ones that can are too big.
Using such a large array of MLCC caps like that is a bit odd, but that's what I have to do. On my boosters all I need is a couple high value MLCC caps on the output and a couple low value ones on the input, not a big deal.
The LTC3785 can run 40W output up to 10V without issue, but I'm going to cap it at 8V for my design. One of the challenges was finding a 10 bit digital pot (1024 taps) to support such a wide voltage range. You don't see many of those. They typically only go as high as 8 bits (256 taps). Maxim came through on that one. It's uncanny how they always have the most obscure stuff when I need it, though it costs. Their stuff is expensive so I usually try to avoid it.
Now I'm working on the PCB design. It's going to be quite a challenge as well. I have a lot more parts to fit on that little circuit board. This thing makes my booster design look primitive. It will be cool though, a real beast with a max current output of 14 Amps at 3V and a range 2.7 to 8.0 Volts.
-
Wow Craig that sounds spectacular. Can't wait to see it.
Damm, more complicated than your boost circuit. This will be mind blowing for sure.
-
Sounds like a monster there. I can't wait to see it.
-
I won't pretend that I understand all that you talked about in your post Craig, but I have learned so much from you on this forum that it inspires me to keep track of your developments and learn more. It sounds like it's going to be very interesting indeed!
-
wowOwow!! This is going to be one spectacular module and a total game changer, Craig!
Looking forward to reading your journey and progress.
-
It does take time to do this stuff. Funny how the gears are turning even when I'm not actually working on it. Just now I thought of an optimization to do on the bias resistance (involves the pot). Wasn't even thinking about it directly. Anyway, it took me like two months just to draw up the design and successfully simulate it. It will probably take that much again to do the PCB design at the pace I work, it's a hobby for fun after all. Not fun if you get stressed over it.
I can't count how many times I've looked at an enginnering problem with this design and thought, "no way this is possible." For example, I use a detection algorithm on power up to differentiate an atomizer short from an overload. It uses a combination of voltage detection and current detection. It's easy to do with the booster since the controller I use for that has a hard coded soft start that initiates from battery voltage. The LTC3785 has a programmable soft start and initiates from zero voltage. It makes detection very tricky. I really though it was not going to be possible. But after endless iterations on the simulator, I got it figured out. I was like, "holy crap that was difficult." That's just one of serveral design problems I ran into.
-
Sounds great Craig! The kind of things that you and BT do are above my head, but it's really cool to hear about it. Can't wait to see the finished product!
-
I can be the adopted beta tester for you, so you will know for yourself if its awesome
as you will think that when upon finishing and vaping it.
I can assure you that, that emote, relieve you of its thought misery of it being awesome..
:popcorn:
-
Sounds awesome Craig, can't wait to see it finished. This is something I would definitely want in my collection, but you already knew that ;) Definitely sounds even more complicated than the original Powerblok.
-
My original plan was for a buck-boost but I shelved that idea due to the complexity...
What software are you doing your sims in?
-
I use LTSpice which works out really well when using Linear parts since they're all modeled natively. I love the crap out of LTSpice, not just because it's free, but it works really well and it's easy to use.
Yes, there's a bunch of different types of buck-boost converters so it depends on which one. At one point I was thinking of an inverting buck-boost converter which would be quite easy to implement, however, there are no controllers with the range we need for an e-cig mod. At one point I tried a SEPIC converter which is not really classified as buck-boost, but it accomplishes the same thing. I actually built that one. The efficiency was crap on that so I scrapped the design.
The 4 switch design of the LTC3785 is quite unique. I've not seen another controller like it. It's a true hybrid of the traditional buck and boost topologies. But yes, it's quite complicated. I believe there's a number of 4 switch controllers with internal switches, but I never look at those since output is limited by the size of the controller's package. For high outputs, external switches are pretty much a must.
One of the fun things about doing these projects is building the test fixtures. Here's one I just did to check transient response on my converters. It switches the load to force a transient on the controller.
(http://webpages.charter.net/chblock/loader1.jpg)
My PCB making skills are pretty good now. I've made so many for little fixtures like this. That conroller I'm using has a .5mm pin pitch and the traces came out perfectly. I have to say my trick is the Elle magazine paper I use for toner transfer. The stuff is really amazing, it just lifts right off when soaked in water leaving a perfect transfer. It has to be Elle magazine paper, others don't work nearly as well. I also use a laminator for heating the toner onto the clad board. I actually took the guts out of a laminator getting rid of the controls and wiring it for maximal heat on the rollers. It works really good.
(http://webpages.charter.net/chblock/loader2.jpg)
The end result is the ability to see transient response for the circuit on the scope. The simulator gets you close, but it's never exact. Here's how it looks. Right now it's a little ringy, but okay. I'm actually going to play with the feedback compensation a bit to see if I can get it tightened up. I'm going to need it for that crazy wide voltage range.
(http://webpages.charter.net/chblock/loader3.jpg)
-
I didn't realize that you are back to making your own PCB's again. Wowee!
Are you using the GBC H-220 Laminator ?
-
Well, some I make some I send out. The simple ones for little one offs I do myself, but the complicated ones for actual projects I always send out. You need those vias and multiple layers sometimes. It's possible to make two sided boards yourself, but it gets to a point where it's counterproductive. You can spend way more time than it's worth to save what it costs to have them made. The simple single sided ones like the one pictured I can make myself pretty quick.
I actually picked up one of these (http://www.ebay.com/itm/Fellowes-EXL-45-2-Desktop-Portable-Pouch-Laminator-/140872086103) laminators. Now they don't really get hot enough to do a solid transfer so I gutted it and wired up the laminating mechanism using the over-temp cutout switch as the thermostat. It gets up to about 190C which is right around the ideal temperature for toner transfer. It does stress the mechanism running that hot, but as cheap as they are, if and when it wears out, I'll just buy another one. Here's a pic;
(http://webpages.charter.net/chblock/laminator.jpg)
-
Craig, does LTspice have models for common dc to dc converters like the OKR's or PTR's? I've got LTspice and can't figure out if they are there or if you have to use one of the ltxxxx ones in the Power section of advanced components that are close.
-
OMG reading the OP was like reading porn, heh *reading* porn. ;hubba; Anyhow.. I haven't seen a lot, ok most, of those terms since I was in the Air Force back in 04-05 but jeeeez. Then I found out you own an Oscope freaked_out:
Don't forget you have to miniaturize the design after you get it perfected :laughing2: Keep it up
-
As an EE student I always appreciate the details of your posts CraigHB. ;cheers;
-
Craig, does LTspice have models for common dc to dc converters like the OKR's or PTR's?
No, LTSpice would not have models for regulators like that. LTSpice is intended to be used to design those kinds of products. It carries models of many active parts such as transistors and diodes from makers other than Linear, but when it comes to integrated circuits, they are limited to Linear parts.
LTSpice can be used to model intergrated circuits from other makers, but you have to build the models yourself if not made available by the maker. Most makers make Spice models available for their parts and since LTSpice uses a standard modeling language similar to other versions of Spice, it's usually not a big deal to adapt a model as required.
Speaking of Oscopes, I actually need a better one. I've been getting by on an old 100mHz analog scope I've had for years, but I really need a storage scope. Sometimes there's just no way around that. I've had my eye on a Rigol one for some time now, price is right;
http://www.tequipment.net/RigolDS1102E.html
-
As an EE student I always appreciate the details of your posts
Good for you on picking the best vocational studdy there is in college (of course I'm biased having a BSEE myself). I actually started out in mechanical engineering because that's also an interest for me and I could get into the school right away, but I was always more interested in the electrical stuff so so I switched majors when I was able. The EE school was impacted when I started and it took me a while to get in. I've actually been an electronics hobbyist since I was a kid so studying it in college was right up my alley anyway.
In any case, as a junior or senior you do simulations and build circuits in your labs. I really enjoyed the hell out of those labs when I was in school. They were like the easy A for me since I had already been doing some of that stuff before college.
-
So I got the first draft done for the PCB on this project. Wow, what a job that was. Talk about stuffing 10 pounds of crap in a 5 pound bag, there it is. I had to drop down to an 0402 (1005 metric) component to fit all this stuff. Boy are those things small, 1 x .5mm. I can't even see them without some magnification. It's going to be fun soldering those.
Anyway, I'm fairly proud of this creation so I'm going to show it off a bit here. This board is going to fit in the same enclosure I used for the last design and it uses the same 1400mAh 10C LiPo flat cell. The enclosure dimensions are 19 x 38 x 81mm.
Here's an image of the PCB drawing for all the layers. This is a four layer board (the forum resizes the images, but they'll expand if you click on them);
(http://webpages.charter.net/chblock/PB1400_Eagle_PCB_all.png)
Here's annotated images of the top and bottom layers only, top in red, bottom in blue;
(http://webpages.charter.net/chblock/PB1400_Eagle_PCB_top_notes.png)
(http://webpages.charter.net/chblock/PB1400_Eagle_PCB_bottom_notes.png)
I still need to proof the drawing before I can send it out for fabrication. It takes some time to verify the the schematic connection by connection and then there's a couple three weeks turn-around for fab. I'm going to want to tidy up the drawing where I can as well. Maybe I'll be able to build an actual circuit in a month. This stuff doesn't happen quickly.
-
Amazing work. It's so tightly grouped and laid out and 4 layers, WOW. That is something to be proud of for sure.
Another amazing thing is you are hand soldering the components. Wow again.
Another ultimate mod by the Craig man. :rockin smiley:
-
Very impressive Craig!
I'm on the phone to Chine...should be a clone by next week :) lol, just kidding! freaked_out:
Can't wait to see what's next!
-
Amazing work Craig!!!! If you want to build a second one for me as a Xmas gift I would be honored. LoL
I hope to one day to be able to do half of what you do.
-
Amazing work Craig!!!! If you want to build a second one for me as a Xmas gift I would be honored. LoL
I hope to one day to be able to do half of what you do.
Too late buddy he already announced on twisterific that Visus is his favorite forum character and would be selected for beta testing etc..
@craighb: Visus is the greatest
It was after he announced his political campaign to eat all the wheaties and cream corn..
-
Looks awesome Craig! I definitely want one in my collection :) Love seeing the progress of your mods.
-
wowOwow Craig - simply amazing and awesome!
Secret ingredient? :laughing:
-
Thanks guys.
Have to go with the wife off to see the inlaws for several days over the holiday so I'll be chomping at the bit to get back and finish up this board. Bummer how the obligations in life get in the way of the really important things, like building mods.
I'm on the phone to Chine...should be a clone by next week :) lol, just kidding!
Hehe, there's no way they would build something like this, parts are way too expensive. In fact, I'm actually a little flabbergasted by the cha-ching on the parts list myself.
Secret ingredient? :laughing:
I was wondering if anyone was going to comment on that. I'm not giving up my find on that one :laughing:
-
Sorry to say this one's a bust.
The LTC3785 looked great in simulation, but it didn't do so good once I built the circuit. It has nasty startup transients that cause problems. I burned out two of the chips taking them up to 8V output so there's a design weakness in the chip. Specs say it can run up to 10V, but it can't even handle 8V.
It worked well other than that, I was getting 40W+ over the full range, 2.7 to 8V, but those startup transients are a show stopper. They cause negative input voltage spikes that play hell with the logic circuits. Worst case, they fry the converter controller. In comprison, the startup for my TI TPS43000 based design is unconditionally a smooth uniform ramp up.
This is a big disappointment for me. The LTC3785 is just not robust enough and has other issues. I'm positive it's not an issue with my design. Excessive startup transients are generally caused by a poorly tuned converter, but output step response is right on the money so for sure the converter is tuned well. It can only be a design issue with the chip. Needless to say, I'll be avoiding Linear parts in the future.
Well, back to the drawing board.
Just for shits and grins, here's how the build looked;
(http://webpages.charter.net/chblock/PB1400_LTC3785_top.jpg)
(http://webpages.charter.net/chblock/PB1400_LTC3785_bottom.jpg)
-
How disappointing. Sorry to hear that Craig. It's a beautiful design. I'm sure you'll work out the kinks eventually and put it aside for awhile to clear your head.
-
Perhaps you had the wrong drip tip on while testing.
Bummers mahoney who always pokes badarse out there when he has no idea where he is.
You'll just knock him on the head and build another..
The data sheet claims, found to be less than working order, never happens to James Bond..
He has the best e cig ever I bet..
:Thinking:
-
Well that sucks, I was looking forward to seeing another awesome build from you, hopefully you can get everything sorted out.
-
That looks too nice not to work and definitely not something I'd be trying if it did lol
-
That sucks, sorry to hear it Craig. Was looking forward to your next creation.
-
Thanks for the condolences guys. It's hugely disappointing when a big project like this doesn't make the cut.
No matter how well you think something is going to work on paper, you have to build the thing to find out for sure. It's not the first time a project of mine has failed and probably won't be the last. I've scrapped at least three mod designs in the past due to some failure in performance somewhere.
A similar project I did that failed was a design that used a SEPIC type converter, efficiency was just not good enough. Then there was another where the converter controller had a nasty failure mode to a soft short. That one was a fire hazard, chip fails, gets red hot, burns hole through battery.
The gears are always turning, trying to come up with another way to put out high outputs above and below the voltage of a single cell. I'll keep my eyes peeled for a good controller. The Linear LTC3785 is just not a robust, well designed part, it's pretty weak really. The data sheet and simulator made it look like a highly innovative part with impressive performance, but that doesn't seem to be the case.
-
You can do it!
-
Thanks for the vote of confidence there Dave, I've got some ideas. Might have to invent my own topology, thinking of a bastardized flyback converter right now.
-
So, I had an epiphany last night and worked up a new topology. Never seen this one before so it may be something new;
(http://webpages.charter.net/chblock/topology.jpg)
It's a buck-boost topology that unlike the standard one, does not generate a negative voltage.
What it does is produce a positive level shift with respect to input positive. That can be seen by the orientation of the plus signs in the diagram. It generates output voltage equal to Vin*D/(1-D) like the standard buck-boost topology, but without the negative polarity. This allows me to use a standard boost controller with a single cell for an output voltage less than a Volt to whatever the controller can handle.
I've run this topology on the simulator and it works perfect. Though, the output level shift pushes voltage over tolerance for the TI controller I'm currently using. The controller has to deal with input voltage ~plus~ output voltage, though the load only sees output voltage. The TI one I'm currently using only goes up to 10V, not quite high enough, damn. If I can find a good boost controller with output tolerance up to 12V and an input voltage range suitable for a single Li-Ion cell, I can build it.
-
There ya go, Craig.
Sometimes disappointments/failures turn out to be a blessing in disguise and lead to something much much better that would never have happened if not for the original disappointment/failure. :yes"
-
Thanks mamu. Yeah, you never know. This one does the same job with a lot less parts, that's always good.
-
woot finally you can use breaks Ti4050c. It is badarse Craig I have no complaints and mine has been rigorous tested. It continues to lay the vapor just chuckin it out there.
-
Vis, You can't compare Craigs creation to a PTN04050c.
Did you see his video?
http://www.youtube.com/watch?feature=player_embedded&v=9QIB2jazLAA (http://www.youtube.com/watch?feature=player_embedded&v=9QIB2jazLAA)
http://youtu.be/6JHMm3cCdow (http://youtu.be/6JHMm3cCdow)
-
I saw the vids but he is looking for a booster chip that he can blow brains out of, the 4050c is maybe the one he will like most..
If I can find a good boost controller with output tolerance up to 12V and an input voltage range suitable for a single Li-Ion cell, I can build it.
-
I could possibly use some of the power modules you find. Some are designed to be surface mounted like any other electronic component. However, part of the fun is designing the power systems myself and I can incorporate features you don't find in off the shelf products. If I can build that buck-boost converter that I outlined a few posts up, it will be something really unique, not just in terms of an e-cig, but for any electronic device.
-
Ya, the grand innovations guy that builds the GI unit is having a fit tuning each unit independent and just as you said, if you mass produced your build, its a pain in his rear...
I wish you well, that R&D is brain boggling and specifically why are you not building a robot army...
The country of CraigHB. :laughing:
-
Oh yeah, calibration is a big time consumer. Each build has to be individually calibrated which takes some time. I don't have to "tune" them since the converter is unconditionally stable for every build, but I do have to calibrate the gauges. I'm working on a better calibration mode for the software that will allow me to speed up that process somewhat. There's no way around it. Anything that measures stuff has to be calibrated manually to achieve any real accuracy.
-
Anything that measures stuff has to be calibrated manually to achieve any real accuracy.
Yes indeed. I agree.
Do you use an algorithm to average out the readings or take a sampling after a set time? I find that real-time readings fluctuate.
-
Interesting and relevant question. There's two approaches to this problem, and it's always a problem. You can filter in software of filter in hardware. I opt to filter in hardware using an RC circuit. The other option is to filter in software using an averaging algorithm. It's pretty much like six of one, half dozen of the other. Personally, I think it's easier to use a cap and resistor than write a bunch of code to average samples. It's really up to the engineer. You could use both, but it's quite redundant, you only need one or the other.
-
So...some interesting things went down with this.
I found a really cool boost controller to implement the topology I posted in reply #35, the TI TPS43061 which is a really amazing chip. It's so cool because it's current mode and uses inductor DCR current sensing. Also, it uses an N-channel rectifier instead of a P-channel. That's good because N-channel MOSFETs perform better than P-channel MOSFETs in switching converters.
Let me explain why inductor DCR sensing in a current mode controller is such a great feature. Generally speaking, the big down side with current mode control is it requires inductor current sensing which costs considerable efficiency. A current sense resistor is normally required inline with the the inductor which wastes power. However, the big advantage of current mode control is it's really easy to tune requiring only a simple RC circuit for feedback compensation (resistor and capacitor). Voltage mode requires a Type III compenstion network which is a hairy 3 resistor, 3 capacitor circuit that is hell to tune. My previous design uses a voltage mode controller and I spent *a lot* of time stabilizing it. Inductor DCR sensing uses the inductor's DCR to measure current which is free so you get the advantages of both types, higher efficiency and easy design.
I was able to simulate the topology in #35 using that controller and it works wonderfully, but with a caveat. I get a 40W output with a range of 2.5 to 8 Volts, but that particular buck-boost topology doubles inductor current compared to a booster. In doing that, it wastes more power since double current means 4 times the losses in the components that make up the converter. The efficiency hit is about 8 points maximally. At max load, I can get about 93% efficiency with a booster using the same components, but it's all the way down to 85% using the buck-boost topology. Those big inductor currents are a bitch with the PCB design as well, again more losses over the PCB connections.
I don't think I'm willing to give up that much efficiency to get those voltages below battery voltage so I'm going to have to give up on the idea of a high output buck-boost mod. In the meantime, I'm going to rework my previous boost design with that new controller. It allows me to use some better parts that get me an efficiency bump a couple points and also raises my ouput cap to 40W over my current 36 Watts. I could actually take it up to 50W with that controller, but I don't want to give up the ammeter resolution for it. That's another advantage of current mode control, better stability with higher loads.
-
I admire your sticktoitness, Craig. Can't wait to see your new creation!
Most of us don't vape below 4v or do uber low ohms. I prefer 1.2 - 1.5 ohm range with 15W, so a boost without the buck is aok for vapers like me.
It's funny, but just last year I considered myself a high watts vaper. What I see going on nowadays makes me feel like a lightweight. :laughing:
-
True, when I started vaping it was with a 3 Ohm atomizer on an eGo, what's that, like 4 watts. Though as soon as I built my first mod, I was at 5V with those so that's more like 8 Watts. Now I typically run a 1.2 Ohm atomizer at 5V for around 20 Watts so that'a pretty big difference, but the atomizers I use now are hugely better than the 3 Ohm cartomizers I was using before that.
In any case, I usually run around 5V so I don't really ~need~ those low voltages. It's more for the challenge of doing it and to build a device that's more of a one size fits all type of thing. My wife could actually use those lower voltage settings. Right now she's running the same atomizer builds as me at 4V which is still a bit much for her. It would have been nice to have a mod she can use with a lower bottom end on the voltage range. I could still build something like that, but if I'm going to do it, I want to build something that meets my own high standards. Otherwise, my wife can just live with 4V.
-
We all new that you would find a solution Craig. Fantastic work my friend :thumbsup:
-
On voltages
I have had a heated :laughing:, sorta argument, with people who think a 3 ohm 11/10 nano wrapped wick cannot beat or match a .4 slr ohm coil with vapor.
I posted the heat chart and at 2amps 32 gauge kanthal is almost 1600 degrees it takes almost 4 amps for #28 to do the same.
I have made both coils and there is not much difference except the 3 ohm is hotter..
They cannot believe the power of regulated mods versus their mechanical its funny..
Woot power and current is your vaping friend
Power it up Craig
Vape chart just for knowledge if you did not know how hot when vaping, what it approaches when wick goes dry..
(http://hotwirefoamcutterinfo.com/_NiChromeData_files/1_Amperage.jpg)
-
The thing is, there are a lot of variables involved with heating the wick of an atomizer. First there's power which only represents the total heat generated. Then the way in which the total heat is distributed has a significant effect. Then there's the actual temperature of the wire. My feeling is that it's more about wire temperature than power. As as extreme example, you could use a 10 foot long heating wire and run 100 Watts through it and it would not get hot enough to make vapor. So, just thinking in terms of power leaves out a lot of important considerations.
There's no real science in determining the best gauge, length, and power to apply in a heating coil. It's rather subjective so I'd have to say it comes down to art more than science. Certainly a number of characteristics can be sited in theory, but it really comes down to the feel of an atomizer's performance when the rubber hits the road.
-
I don't think I'm willing to give up that much efficiency to get those voltages below battery voltage so I'm going to have to give up on the idea of a high output buck-boost mod.
Looks like I spoke to soon. I thought of yet another way to implement a buck-boost mod. So, I guess I'm just determined to come up with something that works to my satisfaction here.
I thought in terms of using two controllers. It's a buck-boost converter that is literally a buck converter and a boost converter on one board. I had to figure out a way to disable one controller or the other without actually using the chip's shutdown mode. Shutting down a controller shuts down the whole circuit. I have to fool the idle controller into keeping the current path fully open. Pulling the feedback up for the boost controller or down for the buck controller does that. It makes the controller think it's out of its regulation range so it goes idle without shutting down. I tried this dual controller idea in the simulator and it works.
Stuffing two controllers on my PCB is going to be quite a challenge. Though it should resolve my previous efficiency issue and provide the 2.5 to 8V range I want out of a single cell. Plus I can use two TI controllers and I love TI products, they make robust, really well designed parts. That Linear stuff is crap, I'll never use them again.
-
I did a little researching into a chips for ya wondered if this company ever you crossed paths with.
http://www.intersil.com/content/intersil/en/products/power-management/switching-controllers.html?pi_ad_id=41434550032&gclid=CMzDsL7V7LwCFSXl7AodnwYAqw
They do not cover your voltage range but looks like they integrate a lot into the chips and can stack them.
Costs are high was just wondering if they are legit chips..
-
I've looked at their stuff before. I haven't used any of it so I can't say one way or the other on the quality myself. I know they've been around a while and I've seen other products they make so I imagine quality should not be an issue with them. Thanks for linking them for me. I may revisit their product line-up to see if I missed something they offer.
Most of the converter controllers on the market are not designed for the relatively high outputs you see powering an atomizer. They're really designed only to serve as power supplies for other electronics that don't see those kinds of power demands. It seriously narrows the field when you start looking for stuff over 10 Watts. The more requirements you have, the harder it is to find something that can do the job.
It seems I'm trying to do the impossible here. Even when using two controllers, there's big problems I run into. For example, I need a buck controller that is 100% duty cycle capable otherwise there's a dead spot where I can't regulate voltage output. That hugely narrows the field for buck controllers.
Interestingly, the TPS43000 I'm using currently is 100% duty cycle capable and is a multi-topology controller (can be configured for either buck or boost). The down side with that controller is it uses a P-channel high side switch instead of the better N-channel and it's voltage mode control. That's actually not a problem since I've already stabilized it, but I'd really like to have those N-channel high side switches if possible. That's a really nice feature. Also the package for the TPS43000 is pretty big, it's a TSSOP. I need something in a 3x3 or 4x4 DFN. It's not a problem when there's only one on the board, but to put two on the board, the larger package is a problem.
-
So I've been looking at another buck-boost design for some time now. I mentioned before I was going to try a new topology that I've never seen used before. It allows me to use a standard boost controller to build a non-inverting buck-boost converter. It does that by making use of a virtual ground at battery positive. Buck controllers are sometimes used to build an inverting boost converter. This is sort of the same thing, but using a boost controller to build a non-inverting buck-boost converter.
I finally got a hold of the PSpice software I need to run the specific controller model TI provides for the TPS43061 (http://www.ti.com/product/tps43061) which is the controller I plan to use. Simulation of the circuit has shown it works. I can build a design running off a single cell for 40 Watts with a 2-8V range and maximal output current of 15A. I need to go through some fine tuning on the sim then I guess I'll be off to a circuit board design. Hopefully I'll have better luck with this than the last, but I have a lot more faith in TI to provide solid working parts. I'm going to use TI power MOSFETs as well.
I mentioned before I had efficiency concerns with this topology, but I think I've managed to get power loss within reason by using MOSFETs with very low RDS(on) and by dropping inductance to get a lower DCR on the inductor. Effeciency is not fantastic, but according to the sim, it does around 93% at moderate loads which is tolerable. It gets as low as 85% when pulling 15A maximal.
One interesting note is a little bird from TI told me they're going to be releasing a 4 switch buck-boost converter controller by the end of the year. This is the best topology for a buck-boost converter, highly efficient. A 4 switch buck-boost converter could probably make 90% at maximal output currents.
The Linear controller I used that failed miserably was a 4 switch topology, but the failure was Linear's fault for releasing a controller that fails to meet advertised specs. I'll for sure be doing controller board with the one from TI when it comes out. I'm sure it won't have the same problem of being total crap like the Linear part. TI makes really good stuff, the best.
-
Good to hear that Craig. I'm sure this one will work out better. Maybe a bit of hair pulling but I know you'll manage another fantastic result as your powerblok was/is.
-
Thanks Dave,
BTW, I did a rework on my boost only design. It's good for 10A and 40W now. I'm doing the assembly on the first of the new batch of boards in between work on a new design. I'll post some pics when I get it assembled. I've never actually posted pics of my controller boards here, (mods yes, boards no), but with the myriad of high output controller boards out there now, I'm not concerned about disclosing the details.
-
Great to hear you still have your EE spirit. Craig I feel like this path makes the most sense with components that are available. My hat is off to you sir. Keep fighting the good fight.
-
Wow I understand how your getting virtual ground by watching eevblog on opamps and how to switch around ground and always being -1v less to whats coming at the input on the output and somehow ground goes virtual where it is impossible to be ground but it is. lol
So a buck feeding the 3061 and the 3061 powering the buck and in return being fed the min ~4.5v. of a lipo 3.7v cell.. Mindboggling
Good build wishes to ya hope these parts are the sauce. To bad cant look over shoulder and ask a slurry of em... :Thinking:
-
Great to hear you still have your EE spirit.
Yeah, I do. Sometimes I lose it when things don't work out, but it usually comes back at some point.
Wow I understand how your getting virtual ground by watching eevblog on opamps and how to switch around ground and always being -1v less to whats coming at the input on the output and somehow ground goes virtual where it is impossible to be ground but it is.
Virtual grounds are just strange and they do present problems. I have an issue to work out, but I figure it will come out in the wash. It makes PCB ground negative with respect to the ground on the atomizer connector.
-
Thanks Dave,
BTW, I did a rework on my boost only design. It's good for 10A and 40W now.
Wow Craig, that's mighty impressive for a boost design :thumbsup:
-
It's pretty good I suppose. Boosters are more tricky to get big currents out of like you can with a buck converter. You're sort of going uphill with a booster, but it's not just that. In control system speak, there's an extra right half plane pole in the transfer function due to an extra term in the duty cycle expression. That extra pole makes feedback harder to compensate since it digs into phase margin a good amount. At very high duty cycles it can be impossible to stabilize a booster. That sounds out there, but it's just the mathematical way of saying a boost converter is harder to tune for high outputs than a buck converter. Buck-boost has the same problem when boosting so it's limited in the same way.
I'm also limited by PCB space. I sure could use a a bigger inductor for more inductance and less DCR. I need a smaller display to get more room on top for bigger parts or more parts, but I'm dead set on using the FSTN type. That's the only type that is readable in direct sunlight. The display I'm using is the smallest one of its type I can find. I'm always on the lookout for something smaller, but I really like the particular display I'm using. It uses a pin header instead of an unruly FPC ribbon cable like most do.
-
Craig - Your EE knowledge is impressive :P Far beyond mine for sure.
Just thinking the other day, since I cannot find a control that exists to do exactly what I want. Couldn't it be possible to have a boost circuit that takes Vin range of 3 - 4.2 volts and boosts it to 12v, non adjustable. The boost would then feed a buck circuit the 12v, which in turn can control from 3.5 v - 10 v? Excuse me if this is what you mentioned a couple of posts ago.
Just thinking it might be easier if you could have a constant boost output to alleviate tuning issues and crossover points between buck/boost. I found a couple of boost controllers capable of doing this at up 20 amps, surface mount stuff, well beyond my fabrication skills scared: A dual sided PCB could allow you to keep the overall size of the part quite small I would think.
TI has a buck controller I found that will take something like 6-36 volt input and allows you to regulate from 3-24 volts or so, can't recall exactly but the buck portion is definitely available already.
I also found a boost controller on digikey that would work for this purpose, again way too small for my fat fingers to work with.
Seeing your incredible fabrication skills, I would think you could do this without much of a challenge, then add something like the low voltage detector you linked to mamu in the other thread for battery cut off. Throw in a couple of fuses and an 1s6s lcd and digital pot and you'd have 20 amp capable from 3-10 volts @ up to 20 amps with full safety and adjust ability.
Just an idea, please don't shoot a layman for brainstorming :)
-
Hehe, I wont' shoot at you for brainstorming (funny comment).
Anyway, what you're thinking is the approach PWM mods from China use. First they boost the voltage, then they PWM it down to the desired output.
Generally, using two regulators in series is something to avoid. The reason is the losses from each regulator multiply. So, if you were to use a regulator with 90% efficiency in series with another regulator with 90% efficiency, you would end up with 81% efficiency, which is pretty dismal.
Anything under 90% is something to avoid. As it is, I'm not too pleased about the 85% at maximal outputs I'm getting with my current design. Though the boost only design I did before does pretty good with efficiency hovering around 90% at maximal outputs.
BTW, when you see someone claim their design is getting like 95% efficiency they're full of it. My boost design can do 98% under optimal conditions, but when you load the crap out of stuff, efficiency takes a hit. Nobody gets around that.
You didn't quote any part numbers there, but I think you're talking about controllers, yeah? That's the base part you start with that determines everything else in a converter design. I've been scouring listings for the next best controller literally for years now. I'm using the best and latest stuff that I can find from the usual suspects.
For anyone looking to build their own converter's there's a lot of FM involved in converter design and I can tell you it can be strange sometimes. Sometimes it's more art than science. It's certainly possible for a hobbyist to design and build a converter, but be prepared for some mystery and frustration.
In terms of modules, I don't pay as much attention to those since I build my own converters, but if there's a boost module capable of 20 Amps out there, I'd sure like to see it and I'm sure a lot of people would want to get a hold of it.
-
I know I ran across one recently but couldn't use it. The 10 thumbs I have don't agree with very fine soldering work!
Let the search commence.
Found what stuck in my brain as a high amp booster. A bit big, obsolete according to TI, too low of an ouput voltage, probably very expensive, and not sure it would work. Great help, I know:-) Maybe it gives you an idea.
http://www.digikey.com/product-detail/en/PT7746N/PT7746N-ND/323476
-
I was wondering, and this is probably caveman stupid lol. but can you use like an arduino or some other controller to switch from buck to boost and not have both running at the same time. this may be what your already doin without the arduino I have no clue. I just thought it might help with efficiency if not. even possibly sensing the batteries output and if you are selecting under that voltage it goes into buck mode, and if your selecting higher it chooses to boost it. also if your right in the sweet spot which is likely never then just letting it flow? I have no clue what kind of chip you got going of much on the matter at all but I have played with arduino a bit. its open source robotics stuff and could make a nice programmable controller, but definitely will not handle the high end output. maybe the lower end but at the same time im not sure how clean the signal would be. I do know it uses PWM, which might give that weird rattle like the older vamo's. pulse vaping lol. also arduino is open source so there is a lot of code already written for it, you might be able to find code that does exactly what your looking for. anything that goes to market that uses arduino they have to supply source code for. I didn't even know it at the time when I built my uav quadcopter but she is powered on arduino, so I can swap between a few different flight controller programs. anyway now I'm way off topic. just thought it could get your creative juices flowing again. even all the brushless ESC's are arduino I have purchased which again I did not know at the time. Might be able to run a Arduino for buck and ESC for boost with some firmware modding and extra circuitry wizardry who knows. get yourself a 40a 14.8v mod that will pop your coil faster than you can say titty twister, or turn a 1k lumen flashlight head into the bat signal. for a split second.
-
http://www.digikey.com/product-detail/en/PT7746N/PT7746N-ND/323476
"The PT7746 only operates in combination with a PT7770 series regulator and is not a stand-alone product. Please refer to the PT7771, PT7772, or PT7777 series..."
Not really something you could use in an e-cig mod. The part itself is rather monsterous in size and it's not a stand-alone product. I'm sure the companion part is just as ridiculously large.
Our limitations in power output mainly come from our size constraints and portability requirements. If it weren't for that, you could simply plug into the wall limited only by the 1500 Watts available from the outlet. When you're limited to a board that can fit in a hand-held box or a 22mm tube running off a battery, that really narrows the ability to provide power output.
-
I was wondering, and this is probably caveman stupid lol. but can you use like an arduino or some other controller to switch from buck to boost and not have both running at the same time. this may be what your already doin without the arduino I have no clue. I just thought it might help with efficiency if not. even possibly sensing the batteries output and if you are selecting under that voltage it goes into buck mode, and if your selecting higher it chooses to boost it. also if your right in the sweet spot which is likely never then just letting it flow? I have no clue what kind of chip you got going of much on the matter at all but I have played with arduino a bit. its open source robotics stuff and could make a nice programmable controller, but definitely will not handle the high end output. maybe the lower end but at the same time im not sure how clean the signal would be. I do know it uses PWM, which might give that weird rattle like the older vamo's. pulse vaping lol. also arduino is open source so there is a lot of code already written for it, you might be able to find code that does exactly what your looking for. anything that goes to market that uses arduino they have to supply source code for. I didn't even know it at the time when I built my uav quadcopter but she is powered on arduino, so I can swap between a few different flight controller programs. anyway now I'm way off topic. just thought it could get your creative juices flowing again. even all the brushless ESC's are arduino I have purchased which again I did not know at the time. Might be able to run a Arduino for buck and ESC for boost with some firmware modding and extra circuitry wizardry who knows. get yourself a 40a 14.8v mod that will pop your coil faster than you can say titty twister, or turn a 1k lumen flashlight head into the bat signal. for a split second.
Not to be nitpicky (but I am), it would help me a lot in reading your posts if you could separate your thoughts in paragraphs. It's rather intimidating to see a bit block of text like that. I would rather like to read your posts.
</nitpicking>
You make some good points so I would not say you're being too elementary here.
I had actually thought in terms of using two controllers. Put one on standby while the other is engaged. The MCU would be able to sense when each is required and switch them over seamlessly. However, I run into a space issue with that. Not only is there the additional controller, but also some additional parts to handle switching them in and out. I just don't have room for something like that on my PCB.
The above is actually the 4 switch buck-boost topology I mentioned before, but utilized with two controllers. Ideally, you want to use one controller to handle both sets of switches. It just saves a ton of space and extra components that way. The problem is that the 4 switch topology is something new in terms of product designs. There's a number of controllers available for low outputs, but not for high outputs. There's only one high output controller I know of right now. I already tried it and it was a fail. TI is supposed to be releasing a high output 4 switch controller this year.
Obviously YiHi got it worked out with the SX350, however, there are a lot of ICs that are domestic to China we don't see here in the states. Evidently a high output buck-boost controller is one of them.
In terms of MCUs, right now I'm using Microchip 16 bit parts (PIC24F), but Arduino is nice in that it's open source and USB based. However, Microchip does a pretty good job with their tools. They're all free except for the programmer/debugger which is not terribly expensive. I like their stuff a lot and I'll stick with it. The Arduino stuff does have some nice advantages, but the Microchip stuff does too. You really can't knock either one.
There are simply tons and tons of MCU products out there. They are one of the most plentiful products in terms of integrated circuits. There's a wide array of options for any one of a big number of MCU platforms so its' hard to say there's one that's best all around. They all have their merits. Some have more merits than others.