A PWM Controller For Each Cell?

[dirty_d]

try lowering the load to about 100 milliohms and see how it works. i don't know of any motors that are 1k ohms.

 

[Randomly]

You remember when I told you that trying to put PWMs in series would be equivalent to having a single PWM at the end of the battery stack? That there was no way you were going to be pull different amounts of current out of series connected anythings?
Here is your schematic, I've simplified it a bit. You will also note that three 0.001mF capacitors in series is equivalent to a single 0.00033mF capacitor.


Try changing your spice schematic to this and see if you don't get the same output. It's just another version of the same old PWM FET for a battery pack, it's not a per cell PWM at all.

 

[safe]

Don't Forget About The Current Sag...

It's true that gates held wide open for a long time will stabilize so that they will match the capacitance of the last cell... but that was never the idea here.

The idea was to "pulse" the energy into the capacitor array and to do it so that you exploit the flow rates and the sag. Using the "fluid pressure" analogy you have to factor in the idea that the cells will sag slightly when they are being drained unequally. At least that was the idea. (however misguided)

Take a look at how when all the PWM gates are opening at the same time (and width) that the LAST charge builds more slowly because the previous cells are drawing current away. The previous cells are preventing the last cell from fully charging because for any given cell the current is divided between going forward into the next battery in the series verses going "sideways" into the capacitor.

The question is:

"Can I exploit this current sag?"

...and I don't know if it's possible or easy.

But you can't think in terms of wide open gates... the gates are being opened for a very short period.

It looks like some sequential issues are cropping up that make it even more complex. :? Take a look at the charging behavior of charge one... it actually starts out above the corresponding cell value and then drops.

 

[safe]

try lowering the load to about 100 milliohms and see how it works. i don't know of any motors that are 1k ohms.

Hey I'm just trying to get something in the "ballpark" to be visible. I'm not even sure if I can capture this concept at all (the idea of multiple levels charging unequally) so the least of my worries is not modeling the load properly. :lol:

But I accept what you say as true... 1K is way too high for a motors resistance...

 

[Randomly]

Don't Forget About The Current Sag...

It's true that gates held wide open for a long time will stabilize so that they will match the capacitance of the last cell... but that was never the idea here.

The idea was to "pulse" the energy into the capacitor array and to do it so that you exploit the flow rates and the sag. Using the "fluid pressure" analogy you have to factor in the idea that the cells will sag slightly when they are being drained unequally. At least that was the idea. (however misguided)

Take a look at how when all the PWM gates are opening at the same time (and width) that the LAST charge builds more slowly because the previous cells are drawing current away. The previous cells are preventing the last cell from fully charging because for any given cell the current is divided between going forward into the next battery in the series verses going "sideways" into the capacitor.

The question is:

"Can I exploit this current sag?"

...and I don't know if it's possible or easy.

But you can't think in terms of wide open gates... the gates are being opened for a very short period.

It looks like some sequential issues are cropping up that make it even more complex. :? Take a look at the charging behavior of charge one... it actually starts out above the corresponding cell value and then drops.

Errors:
1) why do your voltages start at 3.8v, 4.8V, and 7.0V? Do you know? You have your initial conditions set wrong.
2) Do you know why all the pulses show up on your waveform? Because you are probably using a large power FET model, the gate capacitance is on the same order of magnitude as your 0.001mF caps so the gate drive are showing up on your output. Change your caps to 0.1mF and see what happens.
3) I won't even get into the conceptual errors. You are fooling yourself with numerous errors in your simulation and analysis of it. GO BUILD A REAL PHYSICAL CIRCUIT and see what happens.

 

[safe]

Simulations Are Very Addictive...

There's no harm in playing around with ideas...

The other way had a problem in that the capacitors are connected in series from the start. What I really want to do is get a small chunk of energy from cell one into a capacitor, then grab another small chunk of energy into a capacitor and on and on... until I've got a row of capacitors with just as much energy extracted as I want to drain from that particular cell. Then I want to disconnect from the batteries and recombine the capacitors in series and then release that as a pulse of energy. So this is another way of looking at it. The voltage comes out too low, so I'll be fiddling with it for a while. (could be like that 50% problem from before)

This stuff is soooo addictive... (enough playing around for now)

This circuit starts out with capacitors in parallel and then switches them to series, except that it's only doing half of what it probably needs to do.

 

[safe]

Steps Needed

I'm starting to get an idea of how many steps are needed in order to achieve the "canal analogy" of filling each lock with the necessary energy and then totaling that up at the end to form a pulse. As I see it so far the steps need to go something like:

Step One: Fill the Capacitors. (with the Batteries in series)

Step Two: Disconnect the Capacitors from the Battery series.

Step Three: Hold the Energy in a Loop with each Capacitor. (or you lose it)

Step Four: Reconnect the Capacitor Loops into it's own Series Loop so that it can perform work.

This is a diagram that attempts to do that. It does seem to work for one cell, but I'll have to see if it scales up later. Hopefully the more experienced people will say to themselves "oh yeah, that's exactly the thought process I went through in 1976..." or something other than "what a rookie train of thought." :wink:

 

[safe]

http://en.wikipedia.org/wiki/Pyrrhic_victory

A Pyrrhic Victory?

It does seem possible to do what I had imagined. You can fill the capacitors individually and then switch them and you don't even seem to need a separate loop for each capacitor as I had thought. (however, in order to really get the balancing feature you might need to do it) You then need to switch the parallel capacitors (separate) so that they become series. You then release the pulse.

But was it easy to get the pulses to align correctly?

No... and building this (as I see it now) would not be easy because you would have many very, very precise timing elements to make it work. Perhaps with more study the idea could be improved enough to make it feasible, but right now I'm just satisfied to have done it in the simulation.

So this is the diagram of just two cells in series with their corresponding circuitry to make it all work.

An interesting thought experiment.

(hope a few folks took an interest in it)

P.S: I have not even checked to see if the balancing idea works. (which would involve even more complicated switch timing) I think now I understand much better WHY things aren't being done on a cell-by-cell basis... it gets too complicated. And that is important to know. But I will keep thinking about this one as well as others.

It still seems like a metaphor that "should" work somehow.

 

[safe]

A Return To The Charge Pump

A Charge Pump works by changing the voltage level from BOTH ends of a capacitor. (in a cycle) It seems to me that when you place battery cells in series it's doing the functional equivalent of progressively stepping up the voltage of a capacitor.

So it makes me think...

Maybe you want to skip series battery connections altogether and allow the charge pump to make the series connection for you.

Since the charge pump steps up it's voltage with each additional cycle your balancing technique is simply the act of omission... in other words maybe you load cell 1,2,3,4,5... but then skip cell 6.... and continue on with 7,8,9,10,11, etc. The voltage attained will be slightly lower than had you tried to add the "runt" cell but that's okay, you want that cell to rest anyway. This is also a good low voltage cutoff scheme.

Anyway... this is a basic info segment from a popular entry level Charge Pump document (I forget where I got it, but people here have recommended it) and a SPICE diagram and voltage readout. The efficiency isn't great because I'm not sure I've got all the right parts and timing right, but it's in the ballpark for now.

This might also be of interest to people that use hub motors and want DC-to-DC conversion. Think about being able to cycle through your cells TWICE. There's no reason not to build something like this to include voltage doubling since that's what most Charge Pumps are used for. (expect less than half the current for double the voltage) What I'm thinking about is a strange twist on the Charge Pump where rather than pulling from the same cell with each cycle you instead switch cells on each cycle. Feel free to recommend links of people who might have tried something similiar to this since I am trying to learn everything about what's been done before.

The Charge Pump (type) approach is easier to do and seems more practical... it's one of the first ideas I thought of with this stuff, so I'm going back to that. Series battery cells are mostly unworkable as far as I can tell. But the "canal analogy" might be doable with a Charge Pump approach. (poor efficiency is the problem I fear most with this technique)

 

[safe]

If At First You Don't Succeed... (try, try again)

Wow, this charge pump idea makes for a great simulation result if nothing else!

By taking just two 3.2 volt cells and linking them in a series circuit where each section was it's own Charge Pump (and using the standard pulse frequency for a typical PWM controller) I was able to get some incredible multiplications of the voltage. (almost 14 volts at peak) This might be the way to get that DC-to-DC overvolting idea to work without incurring the double loss of the converter and the controller... make them into one and you still lose some energy (I'm sure you do) but at least you get that voltage increase in a slightly better fashion.

Anyway... you can look at the circuit and after a while it should become clear that you could reroute around a cell and it would not inhibit the overall system.

So this does two things:

It increases the voltage because it's a series of Charge Pumps.

It adds the cells together in series and yet it makes it pretty easy to circumnavigate around a "runt" cell. (you just skip it) Not exactly what you would call PWM, but it does the same thing. When a cell went low you could just skip it.

...it's kind of cool when you take an idea and make it actually run on the simulator. 8)

 

[safe]

Three Modes... Three Results

It wasn't long before it dawned on me that I really could trim back the whole "Charge Pump" angle and limit the increase in voltage with each pump cycle. What you do is take the cells out of the circuit and bring them all back down to ground. The effect is to make the pumping action add only about as much energy as the cell itself and this pretty closely approximates the idea of a series connected battery.


So this is mode two... simple series, no big voltage doubling action...


The third mode is when you want to do "balancing by omission". What I do here is simply take the cell out of the circuit. (it would involve another MOSFET, but at 3.2 volts that should be cheap enough)


...so it would not be all that hard to toggle between a standard series connection and a hardcore Charge Pump connected circuit for those high voltage situations.

How It Would Be Built

I'm figuring that you would place all these circuits into one circuitboard that is compressed and close enough together so that timing related issues are minimized. The tighter the timing you can get the more time is spent adding energy into the Charge Pump and less time is spent doing nothing. Since this is a simple "two stroke" system the timing is not all that hard. There are only two sets of MOSFETS that open and close at opposite times. (this is typical of all Charge Pumps and they seem to have some fancy cycling circuits that would do a really good job)

In Summary

It appears that by using the Charge Pump design philosophy that you can achieve a series circuit while never actually connecting the batteries in series. On top of that you can "balance by omission" and even do some voltage doubling if you feel like it. Overall I'm pleased in that the initial claims of a "canal analogy" seem to have been met.

 

[safe]

Buffering?

Seems to me that one way to build this idea would be to have the "extraction" phase be doing the Charge Pump type behavior and that would be moving the energy into a buffer. Then you pulse that last stage as your way to produce PWM.

One would have to investigate the efficiency tradeoff between buffering verses the alteration of the Charge Pump cycle rate.

...after all Pulse Width Modulation involves looooong pulse widths when you are operating at near 100% duty cycles. I just don't see how a Charge Pump that would operate at it's peak efficiency at 50% duty cycle is going to be able to go any higher. (it has to prefer 50% because it's a two stroke pump) By using a buffer you "prestage" the energy you are going to use. To my knowledge most controllers buffer a little anyway simply because it works better with the batteries, so it's not all that far fetched an idea.

What I'm really going to need to do also is to model the DC motor in a SPICE program. I've seen a few, but nothing that will drop right in without some modifications for my needs. Using a static load as the way to test a circuit is really inadequate because the motor and it's inductance have a huge effect on what the controller can achieve. Much of the so called "current multiplication" of PWM does not become visible until you model the inductance of the motor.

What happens?

What happens when you use a Charge Pump that has it's nice 50% duty cycle and simply accept that for what it is? I guess I need to model that to see what happens, but given the way that PWM already behaves it might not be all that bad.

One might change from: PWM (Pulse Width Modulation)

To something like: PFM (Pulse Frequency Modulation)

...and while the duty cycle in the conventional sense is always 50% the pulse frequency increasing would mean more energy being transferred. (? :? maybe) At high frequencies the inductance behavior of the motor would diminish, but it already does that with PWM anyway.

High Voltage Usually Better

In most cases it seems that higher voltage is better than lower because you can use less current to achieve your goals. (less current means less heat) So I'm guessing that the full blown Charge Pump that steps up the voltage to a very high level being fed into a buffer and then metered out in a PWM manner is going to be a superior way to go. Maybe at peak you might have 200 volts entering the buffer, but when you actually release it the voltage would calm down a lot. This seems more efficient in principle. You might want to use a Supercapacitor as this last buffer because you can afford to hold a lot there.

 

[safe]

Simplified Charge Pump Description

A Charge Pump works by first adding energy to the positive side of the capacitor (Filled From Top) and then goes "underneath" and "pumps" the capacitor upwards (Filled From Bottom) so that the final voltage is equal to double what the voltage used to pump the capacitor was operating at.


Hopefully this makes it easy to understand how you can use this philosophy to create a series like connection of cells without actually needing to place them in series. Since in the "canal analogy" the locks are closed part of the time and then switched you can choose to fill with any energy you want.

 

[Ypedal]

Quoting Randomly on page 7
"3) I won't even get into the conceptual errors. You are fooling yourself with numerous errors in your simulation and analysis of it. GO BUILD A REAL PHYSICAL CIRCUIT and see what happens. "

Safe.. this has to stop.


I give you an " A " for effort, but please look at the big picture here.. this is not a Blog.. this is a public forum to share knowledge and experiences, you are in way over your head with all this and causing confusion.

I do encourage you to keep learning, but posting picture after graph after spreadsheet of nothing but fictional data that is full of errors.. will result in some of it being found on " google " by a randome search, some poor bastard will end up right into page 8 without reading the rest of it and risk blowing himself up creating it in real life..

When all this was being explained in detail by EE's it was ok.. but you support seems to be getting pretty thin and i fear the worst.. not for you but for the rest of the world.

 

[EMF]

Damn! I wanted to see how many times he would post to himself before he calved this off into another useless thread!

 

[safe]

I had proposed an analogy:

"Make a battery system that used the idea of locks in a canal and that way you can step up or not step up the voltage depending on the power available to you in an individual cell."

At first that idea was difficult to figure out and it did seem hopeless, but once I latched onto the Charge Pump idea it suddenly started to make sense and actually looks to be a decent idea. (no one is going to suggest it isn't working as a Charge Pump because the design is pretty easy to figure out)

If the analogy was successful then the purpose of the thread was achieved. So it's a successful thought experiment.

Abstract idea -> Actual Circuit

...so I see no problems, the idea turned out okay after all.

(not all abstract ideas turn out to be correct)

 

[Randomly]

I had proposed an analogy:

"Make a battery system that used the idea of locks in a canal and that way you can step up or not step up the voltage depending on the power available to you in an individual cell."

At first that idea was difficult to figure out and it did seem hopeless, but once I latched onto the Charge Pump idea it suddenly started to make sense and actually looks to be a decent idea. (no one is going to suggest it isn't working as a Charge Pump because the design is pretty easy to figure out)

If the analogy was successful then the purpose of the thread was achieved. So it's a successful thought experiment.

Abstract idea -> Actual Circuit

...so I see no problems, the idea turned out okay after all.

(not all abstract ideas turn out to be correct)

Except your original Canal analogy was incorrect and was an erroneous representation of how electricity actually works.

After much flailing around you've ended up copying from a book the stock standard capacitive charge pump topology that's at least 30 years old. However you haven't noticed the real world implications of the design. Maybe the fact that no engineer uses switched cap systems for anything other than very low current systems is a clue.

For a usable system to drive an ebike motor it's totally unfeasible.

1) It requires a huge number of very high current switches. So it is bulky, and very expensive. For 16 cells you'd need more than 64 High power FETs. This also implies all the high current gate drives you would need to drive this huge mess. Also almost none of the FETs are ground referenced so they would all need to be Floating gate drives, requiring many different gate drive power supplies which adds considerable complexity, bulk, and expense. Not to mention the level shifting circuits required to control the FETs. With that much current flowing through that many devices losses would be high, and you would also have a great deal of heat to get rid of from your circuit.

2) It requires very large and very high current capacitors to store the required charge. These would be very bulky and expensive.

3) It would have terribly poor efficiency. in a mighty effort to recover a few percent of pack energy not recoverable from a normal simple cell stack due to slight capacity differences in cells you've come up with an extremely expensive solution that throws more than HALF the total pack energy away, making things vastly worse than when you started.

An interesting mental exercise for you it may be, but you have not come up with anything that improves any aspect of a simple pack of cells wired together.

Yes you have an idea, however it is in no way a useful idea.

 

[safe]

Yes you have an idea, however it is in no way a useful idea.

The whole point of a "thought experiment" is to take an idea (like the canal analogy) and simply see where it goes. The fact that it actually went somewhere was a good thing. Imagine if I had stopped short of the Charge Pump and just left it without an ending? (there would have been nothing educational in that)

Water being pressed by gravity produces pressure. Electricity can be thought of as being like water in that it produces a type of pressure. A canal is raised by the addition of water into locks, as long as the locks are properly timed they can be filled with water so that the pressure at the bottom of the lock rises. (deeper water has higher pressure) The Charge Pump works more or less the same way as a lock in a canal.

So the idea "worked"... whether it's practical or easy is another matter...

 

[safe]

...you've come up with an extremely expensive solution that throws more than HALF the total pack energy away, making things vastly worse than when you started.

I'm going to grill you on this one a little...

Charge Pumps are going to naturally run best when they cycle at 50% duty cycle. This is because you have to load the energy in one phase and then pump in the next. (and they naturally need to be equal) So in the abstract I "think" what you are saying is that compared to a standard PWM system you are limited in the peak amps you can pull.

But people aren't always pulling ALL their amps all the time. In fact I've been designing my bikes so that they have a big gap between the peak amps of the battery supply and the demanded amps that the controller uses. If someone were to be able to operate at a "C" rate that was 50% of the battery peak then this type of system would make a little more sense. (on my current projects the SubC's are rated at 10C, but I only plan to pull 4C from them)

This was what you "meant" right?

The Charge Pump cannot deliver the peak amps of the battery because the pumping action will place a cap equal to about half.

Or did you have another idea?

(was this a "throughput" statement or an "efficiency" statement?)

 

[Randomly]

No, I meant that you will lose half your battery power as heat in your circuit.

dirty_d already posted a link to a technical explanation as to why this happens. which you ignored or failed to understand.
It was pointed out several times after that but apparently you aren't interested in learning how things actually work.

If your understanding of the basic principles of electronics is wrong, nothing you derive or invent from that incorrect basis has any validity.

You may find it personally entertaining, but the results have no value to anyone and in fact they have a negative impact since they force people to spend time evaluating and editing them out as useless misleading information. The earlier analogy likening your technical ramblings as SPAM and all the implications that implies is remarkably accurate.

You don't seem to have any interest in learning how things actually work. Until you make an effort to actually learn what you don't understand it's unreasonable to expect me to invest any time in helping you.

Go get a book and read up.
http://www.amazon.com/Art-Electronics-Paul-Horowitz/dp/0521370957/ref=sr_1_4?ie=UTF8&s=books&qid=1215573210&sr=1-4

Or better yet take a class.

I actually have no expectations that you will do any of this, I'm fairly confident you will just ignore those things that don't fit into your world view. However somebody else reading this thread might gain some grain of useful information out of it.

 

[safe]

Efficiency?

From Wikipedia:

http://en.wikipedia.org/wiki/Charge_pump

A charge pump is an electronic circuit that uses capacitors as energy storage elements to create either a higher or lower voltage power source. Charge pump circuits are capable of high efficiencies, sometimes as high as 90-95% while being electrically simple circuits.

So obviously when you are talking about efficiencies in the 50% range I have to wonder where you got it. My "guess" is that this is because of the use of MOSFET's rather than the lower energy cost options that low voltage Charge Pumps deal with. (you had mentioned this) You might be right (Randomly) that a significant loss will take place with this idea, but the BASE idea of a Charge Pump doesn't seem to support such excessive losses.

They are claiming some can reach 90-95% which is comparable to a typical PWM controller.

I'm not saying you are wrong, but that there is conflicting information coming from you and Wikipedia.

Maybe Wikipedia is wrong?

(or Wikipedia's definition of a Charge Pump can't apply to this circuit)

 

[safe]

Who Might Want This?

While the efficiency question does have some real nagging concerns, if we were to say that (for instance) the heat losses produced a system that was in the 80-85% efficiency range (rather than 90-95% or 50%) and you were able to do all the things you wanted:

Balancing By Omission (LVC)
Charging in Parallel
Voltage Doubling

...then one could argue that for people that are using hub motors this would effectively be like having a DC-to-DC converter built right into the controller. At low motor rpms you could drive the system in simple series mode and then when at high motor rpms you could drive the system with the voltage doubled. The improvement in motor efficiency related to rpm very well could produce a better OVERALL result when you factor in things like hills and high speed straight aways.

The voltage doubling effect that is natural to Charge Pumps is an often sought after feature by the hub motor crowd.

 

[Ypedal]

Safe.. the whole pump thing is just one component in a " system " requried, 10 % loss on this.. + X % loss in the fets + " X % " loss on the next set of components and by the time it's all said and done it's a big o'le wasteful system. Added weight, added cost, added ........

Not to mention the entire concept of draining " ANY " battery pack down to 100 % is a bad idea to begin with.

If i open this thread one more time and read a paragraph that makes me go " wha... ? " . i'm locking it and moving it to spam...

I tried to be nice about the whole thing .. but that don't seem to work on you.. you really won't like my next move, so please don't go there.

 

[Randomly]

Does this look familiar?
http://www.olino.org/us/articles/2006/11/22/charge-efficiency-capacitor
it was posted in this thread earlier by dirtyd.
Several times it was pointed out to you.
Did you read it? no

Charge pumps are capable of high efficiency, but only under very light loads. Your application is not a light load. The efficiency will be terrible.

Why does the efficiency suffer?

http://www.olino.org/us/articles/2006/11/22/charge-efficiency-capacitor

try reading and understanding that.

There are a lot of really smart engineers in the world, why do you think nobody bothers to build high power charge pump circuits?
Why are inductive based circuits used EVERYWHERE? cost, efficiency, power density, simplicity....

1) So how much does balancing by omission buy you? do you even know? You're willing to spend endless amounts of money and effort to achieve something yet you don't even know what it will really accomplish. The cure is worse than the disease.

2) charging in parallel... what is the advantage here besides increasing the currents in the system by 16 times with the resultant heat and large wiring requirements... balancing? Then what is the point of #1? you've come up with the most expensive and inefficient balancing scheme on the planet.

3) Voltage doubling requires a whole new set of FET switches in a different topology on top of the original huge set. But of course you totally ignore costs since you're working with the budget of NASA. It also decreases efficiency even further.

This topology is

Extremely expensive
Extremely inefficient
Extremely complex
Extremely bulky

What are the sellings points again? You're going to get that extra 1% out of that pack from cell imbalances?
What does this system do that can't be accomplished much cheaper and easier by other approaches? nothing.

Will you address my points? no
will you read that link with the technical info? no
will you remain supremely confident in your viewpoint despite all evidence to the contrary? yes
will you brush aside and ignore any evidence or argument which doesn't support your world view? yes

The functioning of minds like yours is an amazing phenomenon, and I must remind myself that as irritating as your eternally confident and stubborn tunnel vision approach is that it's not your fault. That's just the way your mind works and you have little control over it. We get what we get and do the best we can with it. Unfortunately for all that obvious mental activity you are not going down paths of real world discovery but wandering around in a fantasy world divorced from reality. You would get so much farther if you just went and learned the real science.

 

[safe]

Does this look familiar?
http://www.olino.org/us/articles/2006/11/22/charge-efficiency-capacitor
it was posted in this thread earlier by dirtyd.

In fact I quoted from it:

"So with a charge time of 20 RC, and that is 4 times longer than with the constant voltage source, we reduced the amount of lost energy over the ESR to 10 %!
Of course, when we allow for a longer charge time, we can reduce the energy loss over the ESR evern more!"

...they recommended adding an inductor to slow the rate of current:

This is a "technical forum" and so part of it is to review history that might be stored deep in some textbook in some library somewhere. Maybe some people have seen the "actual textbook" and that's great... more power to you... but these days most of us are picking up our information through things like Wikipedia or other sources and through our own processes evaluating what course to take.

It would be nearly impossible for me to imagine anyone reading this thread and then suddenly saying to themselves:

"I'm going to buy the parts and build a Charge Pump based controller WITHOUT actually building it in a simulation with enough detail to prove to myself that I've covered all the bases."

...I mean do people like that even exist?

Maybe you (Randomly) assume people are dumb enough to jump to such conclusions so easily, but I give people the benefit of the doubt that they will not. People reading will release this is a "philosophical thread" and not an "immediate action / I've got ADD and must act now" kind of thread.

Of course you may be right... :wink:

The art of philosophical discussion seems to be a lost art these days with ADD on the rise. Blame tv I guess.