Charging Li-Ion with ripple current

[Martin]

If I were to charge a Li-Ion cell with a charger that terminates at 4.2 V but has 200mVpp ripple, would that do something bad to my cell ?

I mean, towards the end of charge, there will be an average voltage of 4.2 V with a superposed sinusodial hum of 200mVpp, so that I get a voltage that changes b/w 4.1 and 4.3 V periodically.

Will the chemistry integrate this hum and see 4.2 V or will it notice the 4.3 V peaks ?

[wasBlinded]

I think if left on the charger it would almost as damaging as charging to 4.3v. Could you set it for a nominal 4.1v end of charge voltage? That would probably be easier on the cell.

[SilverFox]

Hello Martin,

Interesting question...

I would be tempted to try it and see what happens.

Tom

[MrAl]

Hi there,

In building my best Li-ion charger i wanted to be able to use it with ANY
DC wall wart over say 6 volts and under say 30 volts and still charge
the same and not generate ANY heat (well almost not any). This required
a switcher, and as we all know, switchers do not always put out a perfectly
smooth DC voltage. It ends up being a bit choppie unlike a linear circuit.
A close look at this choppie DC voltage shows it's constantly moving up and
down (roughtly 50kHz) and is never constant.
What i did was looked at the spec of the Li-ion and saw that it can take
a max voltage of 4.250 volts, so i used enough filtering to limit the spikes
to about 50 millivolts (0.050 volts) peak to peak. This means the voltage
can be set to 4.200 volts nominal and it will wander up to 4.225 volts and
down to 4.175 volts. This wont hurt the cell, but i still set it a little bit
lower to make up for any meter tolerance.

Point is, if you simply limit the peak to 4.250 volts you wont hurt the cell.
Since your ripple is 200mv peak to peak, that means if you set the nominal
output to 4.1 it will vary from 4.0 to 4.2 volts, which wont hurt the cell.
Better yet however is to add some more filtering. If you add some more
electrolytic caps to the output you can reduce the ripple to 100mv or even
50mv like i have.

BTW i have been using this charger for quite some time now, charging both
small and larger Li-ion cells.

[Martin]

I'm not asking this question to get help with a constellation that I actually have, this is more a theoretical thing:

How cheesy can a Li-Ion charger be ?
What will be inside tomorrow's $ 2.69 Li-Ion charger ?

Does it actually need a smoothing capacitor after the transformer-rectifier ? The rechargeable battery will do some smoothing, the transformer limits the current, like some real cheap NiMH chargers.

Because it is Li-Ion, no way to get around circuitry to switch off (or shunt) at 4.2 V, but if that can be 4.2 V avg and not peak, that smoothing capacitor can really go, making it cheaper, smaller..

I'm tempted to try it but would rather like to hear some more ideas as I yet have to get my wife's consent for this sort of explosive experiment.

[Bandgap]

Does it actually need a smoothing capacitor after the transformer-rectifier ? The rechargeable battery will do some smoothing, the transformer limits the current, like some real cheap NiMH chargers.

Be careful.
Liion cells are nothing like as robust and over-charge-tolerant as NiCd, or even NiMH.
Read Newbie's dire warnings on overcharging Liions - he is wise.

In my view, you always need something a lot more sophisticated than a transformer and rectifier, with or without capacitor.

The problem is, Liion cells need current limiting and an accurate charge voltage - generally 4.2V* within 1% or so.

Mains voltage can vary at least 10%, so unless you pick a transformer that keeps the charger output below 4.2V when the mains is at its highest, there is a risk of cell damage and maybe even fire or explosion because sometimes the output of the transformer-rectifier will get above 4.2V.
Most of the time, such a 'safe' transformer would under-charge the cell.

All Liion makers (that I know of) recommend a current-limited, regulated constant-voltage charge cycle.
- with automatic charge termination in almost all situations.

A few notes -
Relying on the protection circuit in protected Liion cells is not a good strategy.
This is a ONLY safety circuit and not a charger - it limits voltage to over 4.2V - reducing cell cycle life.

Under charging - to less than 4.2V - is safe and can result in longer cycle life. However, less run-time will be available from that cell on that cycle.

Steve

*sometimes 4.1V, it depends on the form of carbon used inside.
A few cells are designed for other voltages.

[uk_caver]

I'd assume that having ripple around wouldn't exactly make it easy to measure voltages accurately - workarounds to try and deal with ripple with precision might be more complicated than just having a smoother supply.

In the end, I suppose much depends on what someone bothers putting into mass-produced silicon. I suppose someone could make a chip that took a rough rectified input and charged a cell with the output, cutting off charge at the precise correct point, but would that (along with a regular transformer), work out cheaper than a direct swiching design with a consequently smaller transformer and possibly fewer heat issues.

[Martin]

.. but would that (along with a regular transformer), work out cheaper than a direct swiching design with a consequently smaller transformer and possibly fewer heat issues.

You are right, this is what we're seeing these days. The switchmode solutions have become so small and so cheap, they have left behind the traditional transformer.

Still I wonder: Is the chemistry having an integrating effect or will repeated 4.3 V peaks make it go boom ?
I think I talk to Newbie as Bandgap suggests.

[MrAl]

Hi again,

There is nothing wrong with smoothing the DC a bit more to aid in charging a
Li-ion cell. The trick, as has been mentioned, is to keep the peak lower than
the max voltage for an Li-ion cell (4.25 volts). The drawback is that for part
of the time the charge current isnt as high as it could be with a perfect DC
supply, and so charge time is increased. The higher the ripple (and assuming
the safe level of peak voltage is used as suggested) the longer the charge time,
because a higher ripple means the charge current will be lower for a longer
time. Smoothing the ripple (and keeping the peak voltage correct) means
the charging supply doesnt go as low during the wave dips, and this means
the cell charges up faster. How much difference in time? Well, with a supply
that varies from 4.000 to 4.200 it could take a lot longer to charge because
with simple rectifier type power supplies the peak only lasts for a fraction of
the total 120Hz cycle time. This means that smoothing the ripple out to
a very small peak to peak value (0.050mv or better yet 0.010mv) could have
a profound effect on the charge time. 200mv peak to peak probably isnt very
good at all, while 10mv peak to peak is very good.
It's not too hard to calculate the required capacitance to get the ripple down
to an acceptable level.
Also as noted previously, i have used a value of 50mv peak to peak and got
good results.

[Bandgap]

If you were not thinking too much about power efficiency, a low-cost possibility would be to use an internally current-limited wall wart type transformer and a shunt regulator.
The regulator could be three cheap components, but would dissipate some power (get hot).
Its voltage could be set to clip ripple peaks to 4.20 or whatever and forget the capacitor.

But really, IMHO there is no sensible point in making a Liion charger without electronic regulation.
And if you are going to have any silicon in it at all, you might as well have a proper regulator.

Take a look at the amazing chips from Power Integrations for example.
The LinkSwitch-TN at www.powerint.com can do the whole mains to battery thing without a transformer (providing you can't touch the battery during charging, this is OK. I wouldn't let a kid use it).
You would have to mess with the basic circuit a little to get the voltage precision up to the level required for Liion charging, but I suggest this is the way cheap chargers are likely to go - after all, the 12V car phone adapters you buy are not just a big resistor, which they could be, they are switching designs - and that must have proved to be the most cost-effective solution.
For example, the High-Side Buck – Optocoupler Feedback design with a precision 4.20V voltage sensor (The Nat Semi LM3620 would do - and it has 4.1 and 4.2V select pins).

An isolated design (where you can touch the battery during charging) would only be slightly more complicated.

I apologise if my concentration on regulation is hijacking the spirit of this posting, but I worry about explosions and fire in Liion charging and I am not sure there is an unregulated transformer-and-rectifier design that will provide a reasonable charge time combined with safe charging under all circumstances.
I will cease to hijack now!

Steve

[Calina]

On a theoretical point of view this is an interesting question but I am wondering if this is not just splitting hairs.

How much difference in capacity could it be between a cell charge at 4.1 V and 4.2V ? Is it worth the trouble or he risk?

[wasBlinded]

On a theoretical point of view this is an interesting question but I am wondering if this is not just splitting hairs.

How much difference in capacity could it be between a cell charge at 4.1 V and 4.2V ? Is it worth the trouble or he risk?

Its about 10%. And about the same between 4.2 and 4.3 volts. Charging to 4.3 volts wouldn't mean the cell is going to explode, it would simply last many fewer cycles of cell life.

[Bandgap]

OK, I can't keep quiet...

Adding to what wasBlinded said....

According to a paper written by Khosrow Khy Vijeh of National Semiconductor in 2003.....

"Initial capacity increases by about 5% for each 1% increase in charge termination voltage."

He has a graph showing
170Wh/l at 4.0V
220 at 4.1
260 at 4.2
300 at 4.3
This graph s pretty close to a straight line at the voltages of interest.

However, the life reduction is incredible. The graph is a real ski jump.

2500 cycles at 4.1V
700 at 4.2
200 at 4.3
under 10* at 4.4

so 50mV in charge voltage has a big effect on life.

4.1 and 4.2 are either side of the lifetime curve 'knee'.
Interestingly, at 4.0V, his graphs show virtually limitless life.

The paper used to be available from planetanalog, but I can no longer find it.

Steve

*hard to tell, the curve has hit the bottom axis of the graph.

[SilverFox]

Hello Steve,

I think you may be reading the chart a little optimistically. Unfortunately, there are no grid lines on the chart, so it is open for interpretation.

My view is that charging to 4.2 volts gives you about 500 cycles, and charging to 4.1 volts gives you somewhere between 1500 - 2000 cycles.

From experience, charging to 4.4 volts gives you 4 cycles...

Here is the article.

Tom

[Bandgap]

I think you may be reading the chart a little optimistically. Unfortunately, there are no grid lines on the chart, so it is open for interpretation.

I am happy to go with that Tom.
A few years ago I discussed Liions with battery makers and at the time they seem to think laptops got about 600 cycles from their batteries.

However you read that graph, life suffers dreadfully with increasing charge voltage.

Thanks for the link. It is a fine article, I feel. Did you find a way of clicking and getting part 2?

Steve

[SilverFox]

Hello Steve,

No, but here is the direct link to Part 2.

Tom

[Martin]

Some excellent material that you guys have linked.

It's about time I come out with the truth, my real intention:
The idea is to charge a Li-Ion cell from a bicycle dynamo. The current is limited to around 500mA by the nature of the generator. It's output frequency is 0 to 100 Hz, whereas below 8 Hz the voltage will not reach the critical 4.2 V.
Because the frequency can be so low, it is difficult to smooth the supply with reasonably-sized capacitors. Chopping the voltage up electronically doesn't make a difference, as there's simply nothing b/w two pole steps of the generator. So there is ripple. What can I do ?
1) I can terminate the charge when the positive peaks of the voltage across the cell start to touch 4.2 V.
2) Or I can terminate the charge when the average voltage reaches 4.2 V
3) Finally I can do as Steve mentioned, clip whatever exceeds 4.2 V and eventually shunt the full current with the regulator.
What I like to do is (2), the best charge performance, the least losses. I suspect this approach is safe but because this is not the typical way people charge Li-Ions, I'm not 100% sure.
As NewBie suggested, I have meanwhile contacted a number of battery makers (no answer yet).

[SilverFox]

Hello Martin,

I would suggest you follow Al's advice and limit the charge to 4.1 volts. If the ripple is significant, you may end up with cells charged slightly above that, but that is OK.

Tom

[uk_caver]

Presumably, once charging has been terminated, you'd want to hold off recommencing it until the cell voltage drops to an appropriate level?
What would an appropriate recommence-charging voltage be, when charging and discharging are unpredictable, as they could be on a bicycle light?

With Lithium cells, is there some ideal depth of discharge?
If charging was only done to ~4.1V, would multiple shallow discharge/charge cycles be better than fewer deeper ones?
If charging was to ~4.2V, would the picture change, given that charging while the cell voltage is high seems to do the most damage?

[SilverFox]

Hello Uk caver,

4.05 volts is used as a trigger to start charging again when the cell has been charged to 4.200 volts. I am not sure what the best value would be when you charge to 4.1 volts.

The self discharge rate is so low that I don't recommend leaving cells on the charger. Charge them up and remove them from the charger in between use. If you are going to store the cells for an extended period of time (2 weeks or longer), you should store them at 40 - 60% charged.

What is very interesting is that you get the most cycle life if you charge to 3.92 volts, and discharge to 3.0 volts. You loose some capacity, but greatly increase the cycle life to something like well over 50,000 cycles.

Tom

[uk_caver]

typoFor charge maintenance on a charger, I guess a different restart-charge voltage might be used than in an intermittent drain application.
Presumably in a charger, a voltage is picked that a cell is unlikely to quickly discharge to after being charged up?

One scenario on a bike could be a stop-start pattern that results in charging being repeatedly triggered, so the cell is almost permanently in a state of being charged, possibly hovering around the retriggering point, or maybe continually being just topped up to cutoff point, then cycled back down to the retrigger point. Presumably in a case like that, the lower the retriggering voltage, the better for the cell, though at the expense of possibly reduced stationary runtime.
However, could it be that somewhat deeper cycling and/or a lower maximum voltage would actually end up over time with a greater [average or minimum] stationary runtime, compared to a system based round a less-well-treated cell which was kept more topped up but [hence] lost capacity faster? Much might depend on the pattern of cycling, and the frequency of light use.

[Bandgap]

Hello Steve,

No, but here is the direct link to Part 2.

Tom

Thanks Tom - have beed away for a few days.

Steve

[Bandgap]

It's about time I come out with the truth, my real intention:
The idea is to charge a Li-Ion cell from a bicycle dynamo.

Ah ha, my speciallist interest.

OK, I have not done this yet, but I have been charging batteries off bike dynamos for years and I have put a lot of research into this for my Mk5 bike light. - delayed this year by too much gardening so far!

- of course, I might have got it all wrong, but here goes.

You do need regulation - absolutely - don't be daft.

You do not need a capacitor.

Series regulation is tricky with dynamos because the voltage climbs high as the load reduces - a Shmidt can give 170V . AVOID unless you are an EE

The same applies to a buck regulator. - AVOID unless you are an EE.

The cleanest way to do it is with a simple shunt regulator, using an accurate voltage reference chip like the one I refered to in an earlier e-mail.
The circuit willl be smaller than a matchbox - loads smaller than a suitable capacitor.

The way dynamos work, the gaps between the cycles closs right up, so you get charging for most of the cycle - much more than a clipped sine wave.

ANY use?

If this is, I have also got some useful ( I claim ! ) stuff to say about charge termination and charge voltage.

Steve

[Bandgap]

I have just posted some stuff further up [now copied below in this email] that I claim will be useful - I am a bike dynamo nut.

A dynamo nut who is failing to get used to this new hierarchical posting structure.

in short - I think I know what you have to do.


Now copied in here--------------------------

It's about time I come out with the truth, my real intention:
The idea is to charge a Li-Ion cell from a bicycle dynamo.

Ah ha, my specialist interest.

OK, I have not done this yet, but I have been charging batteries off bike dynamos for years and I have put a lot of research into this for my Mk5 bike light. - delayed this year by too much gardening so far!

- of course, I might have got it all wrong, but here goes.

You do need regulation - absolutely - don't be daft.

You do not need a capacitor.

Series regulation is tricky with dynamos because the voltage climbs high as the load reduces - a Shmidt can give 170V . AVOID unless you are an EE

The same applies to a buck regulator. - AVOID unless you are an EE.

The cleanest way to do it is with a simple shunt regulator, using an accurate voltage reference chip like the one I referred to in an earlier e-mail.
The circuit will be smaller than a matchbox - loads smaller than a suitable capacitor.

The way dynamos work, the gaps between the cycles close right up, so you get charging for most of the cycle - much more than a clipped sine wave.

ANY use?

If this is, I have also got some useful ( I claim ! ) stuff to say about charge termination and charge voltage.

--------------------------------------------

For - charge termination - take a look at

http://www.electronicsweekly.com/Art...+ion+cells.htm

Something to note here is that these are all professional-grade cells.
I have no idea how this reads across to consumer cells, or cells from any other manufacturer.

DON'T FLOAT-CHARGE LIION CELLS UNLESS YOU KNOW WHAT YOU ARE DOING OR ARE GUIDED BY SOMEONE WHO DOES. YOU CANNOT ASSUME ALL LIION CELLS CAN BE FLOAT CHARGED.
- end of safety announcement.

So you may not need to terminate and re-start.
I am going to float-charge - at 4.0 or 4.1 - Liion or LiPo - but not inside a building, and I am going to give my lighting system plenty of facility and room to ventilate safely if the cell bursts.
I have an idea to put a solar charger and Liion cells on a pole in the garden and let it float at 4.1 or 4.2 for a few months to test the idea where it can do no harm.

As a last thought, I just browsed across these that look interesting.
www.flyelectric.ukgateway.net/lithium-a123.htm

Steve

[Martin]

Presumably, once charging has been terminated, you'd want to hold off recommencing it until the cell voltage drops to an appropriate level?
What would an appropriate recommence-charging voltage be, when charging and discharging are unpredictable, as they could be on a bicycle light?

Well, exactly this was my problem with NiMH. The ever-changing charge current from the dynamo would have made it difficult to detect the voltage dip that indicates EOC. And once the bike stops and then moves on with a fully-charged cell, the dip will not come again and the cell gets overcharged.
I figured that Li-Ion is easier, it's a simple CCCV charge. The dynamo takes care of the CC, I just have to limit the voltage. When the bike stops, the current ceases. When it restarts, charging continues until the cell voltage exceeds a certain level, then the charge current is interrupted (or it just ceases by the nature of a CV charge). I assumed these repeated charge attempts don't harm a fully-charged cell as long as a certain cell voltage is not exceeded.

..delayed this year by too much gardening so far!

Sadly the exact same thing with all my hobby projects.

You do not need a capacitor.
..
The way dynamos work, the gaps between the cycles close right up, so you get charging for most of the cycle - much more than a clipped sine wave.

My Shimano DH-3D71 + bridge rectifier delivers something very close to a rectified sine wave (ABS(sin(t)) into a resistive load.
The current into an 18650 Li-Ion follows the same function, from 0 to 830mA peak. It averages at 500mA. The voltage across the cell varies b/w its open circuit voltage and open circuit voltage + 140mV => an AC part of 140mVpp.
Adding a smoothing capacitor of 10mF, the charge current follows more or less a sinusodial curve with a lower peak of 280mA and an upper peak of 700mA. The voltage across the cell varies b/w its open circuit voltage +50mV and its open circuit voltage + 120mV => an AC part of 70mVpp.
Still I don't need a capacitor as regulation is essential and it will just clip any excess voltage.

Series regulation is tricky with dynamos because the voltage climbs high as the load reduces - a Shmidt can give 170V . AVOID unless you are an EE

Absolutely. With my Shimano hub, I was able to direct-drive a flourescent tube. When trying to disconnect the dynamo or do any kind of regulation that reduces the input current, a number of components will have to be rated for a rather high voltage and that will typically punish me with higher cost and higher losses (Rds_on, Vfw). So the plan is to shunt any extra current.

The cleanest way to do it is with a simple shunt regulator, using an accurate voltage reference chip like the one I referred to in an earlier e-mail.

Right. It will work more efficiently with a smoothing capacitor (as MrAI explained), while it will still work without.

For - charge termination - take a look at
http://www.electronicsweekly.com/Art...+ion+cells.htm
..
I am going to float-charge - at 4.0 or 4.1 - Liion or LiPo - but not inside a building, and I am going to give my lighting system plenty of facility and room to ventilate safely if the cell bursts.
I have an idea to put a solar charger and Liion cells on a pole in the garden and let it float at 4.1 or 4.2 for a few months to test the idea where it can do no harm.

I assumed that float-charging is perfectly safe as long as I do not exceed a certain voltage. My reasoning is that the current is zero if the battery voltage equals the charge voltage, so the battery would not even know it's being float charged. I also assumed there's virtually no temperature-dependency of the charge voltage. Lots of assumptions.. Don't miss to publish the outcome of your experiments on CPF !


Thanks to all your detailed explanations, I can see a bit clearer at this stage of the thread:
My initial worry that ripple current adding to the DC charge current of a Li-Ion cell could push the cell beyond its limits and cause catastrophic failure, is pointless. This is because a Li-Ion cell will not fail catastrophically even when charged to 4.4 V. However, its useful life will be reduced so much, that it's a good idea to charge to 4.0 or 4.1 V only. With the constraint being cell life, even a large voltage ripple will not be able to create a dangerous condition.

A voltage-limitation is essential for charging Li-Ions. The regulator will not only deal with the gradual rise of the cell voltage as it's being charged, it will also clamp any voltage ripple that exceeds the set voltage. So if there's ripple, it's within safe limits. The charge performance can be improved by smoothing the supply, while this is optional.

The circuit could look like this, while I will probably use an off-the-shelf integrated voltage regulator for the op amp + voltage reference.

[Bandgap]

For what it is worth, a have a few more comments.

1st, glad I am not the only one wrestling with spring weeds.

next....

My Shimano DH-3D71 + bridge rectifier delivers something very close to a rectified sine wave (ABS(sin(t)) into a resistive load...

I was really talking of the voltage waveform. So few people have a clue about current and phase that I was steering clear of current.
In my measurements with the Schmidt, using an RMS current meter, the dynamo seems to deliver the same RMS current into leds or incandescent loads. - providing the output voltage demands was roughly similar - For me the voltage waveform varied a lot.

The url="http://i168.photobucket.com/albums/u200/Vankoff/bikecharge.jpg"]circuit could look like this[/url], while I will probably use an off-the-shelf integrated voltage regulator for the op amp + voltage reference.

That looks good to me.

Another National chip, the LM385-ADJ I think, lends itself to a simple but accurate shunt regulator. As far as I remember, you only have to add one NPN medium-power transistor and a resistor, plus the two potentiometer resistors - that can be high value.
Afaik it senses voltage between the feedback pin and its positive terminal. So you can dump its negative terminal current straight into the base of the NPN, and just use a base-emitter resistor to keep down the Vbe with quiescent current flowing.
I may (as usual, be wrong in all of this!) a diode is needed somewhere in the negative system to stop the tail current discharging the battery when there is no charge current.
I would post a circuit, but I don't know how.

Bob Pease does something similar in this circuit
http://www.electronicdesign.com/Arti...ArticleID=4251

Charging efficiency (electrons in to electrons stored) is almost 100% for Liion cells, whit is why a favour them for pedal-power use - plus the remaining charge is easy to estimate and they are light.

Just out of interest, what are you going to do with the cell, and what capacity cell do you feel safe to charge at 500mA?
I am agonising over the minimum cell that can reliably deliver 500mA. My intention is only to charge at 100mA max - stealing it from the headlight when needed.

Steve

[Martin]

Another National chip, the LM385-ADJ I think, lends itself to a simple but accurate shunt regulator. As far as I remember, you only have to add one NPN medium-power transistor and a resistor, plus the two potentiometer resistors - that can be high value.

LM385-ADJ ? Wasn't aware of the ADJ version, nice to know, Steve. So far I was looking at the TL431 plus PNP.

Just out of interest, what are you going to do with the cell, and what capacity cell do you feel safe to charge at 500mA?

At this time, I only want to be able to charge single cells for flashlights, GPS, cellular. During daytime, when not using the dynamo for light. I'm looking at single-cell applications of 1000 .. 2000 mAh. Nothing complicated.

In my dreams I thought abt backing up the dynamo light system with a battery, but then this is getting tricky: Charge during the day and use at night if the dynamo doesn't deliver enough energy ? Or should I charge the battery day and night, losing some light at night ? Instead of running 3 series-LEDs, I could run 2 series-LEDs plus a Li-Ion cell in series. And why not try to maximize light output by holding off the night-charging until the cell is below 30% ? I like that. Using two Li-Ion cells in series would probably be better than one. These 2 feed a step-up converter that maintains a constant current thru a 3-LED headlight, so if the dynamo drives a higher current thru the LEDs, the battery and converter won't engage at all. Well, just ideas that will not be built very fast.

[AndyTiedye]

Added to Bike Light Thread Links

[Martin]

Added to Bike Light Thread Links

Amazing your filter found this ! Tnx !

[Martin]

Received a call from Sanyo today. They said that it's not acceptable that the ripple voltage exceeds 4.2 V.
So it's confirmed, need to clip the peaks.