EEVblog #811 - How The Varta 15 Minute Battery Charger Works

[hamdi.tn]

I have a little different take on what is going on in the charger.
Only one cell is being charged at any given time.
They are multiplexed.

+1

[mariush]

Bad user interface, those two buttons were just confusing to me. Lack of configuration for charge current is another fail for me. Could have added a third button and allow configuring current for each battery.

Would have loved to see the insides of the AC adapter.. is it a proper quality one that can actually do 45w or it's just something that overheats if used a lot (and they're relying on duty cycle,  people only using it for 15 minute periods and then idling for long times),.. how about temperatures inside that ac adapter when it outputs that current...

[IanB]

Exactly that. The "-dV" detection method implies that anyway: you send a current for a fixed period of time down the cell, then you let go and measure the open circuit voltage.

No! Not that. To do -dV detection reliably you must measure the voltage while the charging current is still being applied. The -dV signal is actually measuring the drop in internal resistance of the cell that happens when the cell starts to warm up at full charge. Internal resistance is only successfully measured by examining the voltage change caused by an applied current. It cannot be measured open circuit.

[andiz]

When I look at the wide-spread availability of the modern low self discharge cells, those crazy fast chargers are nothing else than pure marketing bullsh*t.
You pay a premium for those super-duper fast-chargers and all they are doing is ruining your cells by toasting them. It is a way better practice (and cheaper) to have a spare number of Eneloops or equivalent good quality cells to exchange the discharged cells in whatever device, and use a low-speed charger to recharge those discharged cells time controlled at a rate of 1/10C for about 12-14h.

What NiMHs don't like at all ist heat and overcharging (combined it's even worse). They will build up higher inner resistance because of separator and electrolyte damage, which leads to even higher heat losses (very practical for high-speed charging, isn't it? ). There are cells on the market, which can't even cope with 1C charge rate for long until they are building up additional inner resistance (after about only 20-50 cylces at 0,5-1C charge rate). This higher inner resistance makes those -dV controlled chargers pretty useless, because they only work (more or less reliable) with new cells (but not for the first 1-3 cycles, where they are likely to get hopelessly overcharged, because the chargers won't find the cut-off point). Ironically most -dV chargers claim to not overcharge, but without overcharging they wouldn't even work.

Actively cooling the cells during charging (like Varta) does not make anything better. First, because the inner heat of the cells will be way higher than on the outside and second, the active cooling will distort the detection of temperature- and voltage-signals.

[dr.diesel]

Not to mention that 15 minutes isn't fast enough to sit/stand and wait for it either, so why boil them? 

[Zbig]

Exactly that. The "-dV" detection method implies that anyway: you send a current for a fixed period of time down the cell, then you let go and measure the open circuit voltage.

No! Not that. To do -dV detection reliably you must measure the voltage while the charging current is still being applied. The -dV signal is actually measuring the drop in internal resistance of the cell that happens when the cell starts to warm up at full charge. Internal resistance is only successfully measured by examining the voltage change caused by an applied current. It cannot be measured open circuit.

You're right, I was wrong there. Anyway, I'm still pretty sure some multiplexing action is happening there. I have different (but I doubt radically different in terms of internal architecture) charger and this switching has apparent audible effects.

[andiz]

No! Not that. To do -dV detection reliably you must measure the voltage while the charging current is still being applied. The -dV signal is actually measuring the drop in internal resistance of the cell that happens when the cell starts to warm up at full charge. Internal resistance is only successfully measured by examining the voltage change caused by an applied current. It cannot be measured open circuit.

Sorry to say that, but this is not correct.
Measuring the voltage under load is a really bad technique (unless the cells are being soldered into a circiut), because of varying contact resistance between the battery terminals that can lead to false voltage readings of more than 10 mV and this is more than the detection threshold of the -dV cut-off. This is the reason, why commercially available charging controllers are measuring the open circuit voltage for -dV detection.
NiMH or NiCd cells have a negative temperature coefficient (because of the dropping of differential resistance) driven by heat an pressure, that reduces the open circuit voltage of the cells as well.

If you want to measure the inner resistance of a cell in order to calculate the charging termination point, you have to measure the difference between open circuit and load voltage. But also this technique calls for a proper connection eg. cells being soldered into a circuit.

[mikerj]

I have a little different take on what is going on in the charger.
Only one cell is being charged at any given time.
They are multiplexed.

That makes some sense I guess, the gate and cell sense voltages would always be referenced to ground then (albeit via a variable number of MOSFET RDSon drops).

However, that means the peak charging current would need to be 16 amps to get an 8 amp average with two cells (and a 4 amp average with 4 cells).  The switcher doesn't really look capable of delivering that kind of current, even at a low voltage.  What are the MOSFETs rated at?

[IanB]

Sorry to say that, but this is not correct.
Measuring the voltage under load is a really bad technique (unless the cells are being soldered into a circiut), because of varying contact resistance between the battery terminals that can lead to false voltage readings of more than 10 mV and this is more than the detection threshold of the -dV cut-off. This is the reason, why commercially available charging controllers are measuring the open circuit voltage for -dV detection.
NiMH or NiCd cells have a negative temperature coefficient (because of the dropping of differential resistance) driven by heat an pressure, that reduces the open circuit voltage of the cells as well.

If you want to measure the inner resistance of a cell in order to calculate the charging termination point, you have to measure the difference between open circuit and load voltage. But also this technique calls for a proper connection eg. cells being soldered into a circuit.

I agree the contact resistance at the cell terminals can be a source of problems, but since the charger is left unattended while operating whatever contact resistance exists is likely to remain the same unless anyone touches it. (I always spin the cells in the slots before charging to remove any oxide layer and create a good contact.)

The -dV under charging current is unfortunately how the electro-chemistry works. We are stuck with it if we want to detect -dV reliably.

As soon as the charging current is disconnected the cell terminal voltage decays in an asymptotic fashion towards some lower stable voltage, but it takes a few seconds to stabilize. So if you are doing pulsed charging and you try to measure the voltage in the gaps between pulses you are aiming at a moving target.

Maha had this problem in their design of the C9000 charger. The first version to market tried to do -dV detection in the gaps between charging pulses and it became heavily criticized for missed terminations and overcharging. In later revisions Maha modified the firmware with an additional termination condition upon reaching an open circuit voltage of approximately 1.47 V. This terminates modern cells like Eneloops reliably, but leaves them at slightly less than a full charge.

Of course, if you try to measure the voltage while the charge current is being applied, you not only have contact resistance to deal with, but also noise signals from the switching converter that supplies the charging current.

Good modern chargers such as the ones sold by Panasonic for Eneloops seem to have very sensitive and sophisticated end of charge detection algorithms. They appear highly reliable. I have seen any number of other brand chargers that are were desperately unreliable and could not be trusted at all.

[AF6LJ]

I have a little different take on what is going on in the charger.
Only one cell is being charged at any given time.
They are multiplexed.

That makes some sense I guess, the gate and cell sense voltages would always be referenced to ground then (albeit via a variable number of MOSFET RDSon drops).

Yes it would be variable, however I argue it would be small. +/- 20% is no big deal in this equation.

However, that means the peak charging current would need to be 16 amps to get an 8 amp average with two cells (and a 4 amp average with 4 cells).  The switcher doesn't really look capable of delivering that kind of current, even at a low voltage.  What are the MOSFETs rated at?

Not if we are charging one battery at a time with eight amp pulses.
And one other thing; nobody said the eight amps was an average, continuous or peak.
As I said I see the charger multiplexing the charging, only charging one battery at a time by doing some switching. There certainly is enough MOSFETs to do the switching.
I would love to see Dave get in there with a four trace scope and and see what sequence those MOSFETs are being switched in.

[drussell]

And one other thing; nobody said the eight amps was an average, continuous or peak.
As I said I see the charger multiplexing the charging, only charging one battery at a time by doing some switching. There certainly is enough MOSFETs to do the switching.
I would love to see Dave get in there with a four trace scope and and see what sequence those MOSFETs are being switched in.

It is quite possibly charging up to two cells at a time, hence the reason it switches down to 4A max average with more than two cells inserted.  It must be averaging right around that 8A per cell to be able to get anywhere close to full charge on a 2Ah cell in 1/4 hour, though, regardless of whether it is a higher peak current being pulsed or a continuous 8A.

[AF6LJ]

And one other thing; nobody said the eight amps was an average, continuous or peak.
As I said I see the charger multiplexing the charging, only charging one battery at a time by doing some switching. There certainly is enough MOSFETs to do the switching.
I would love to see Dave get in there with a four trace scope and and see what sequence those MOSFETs are being switched in.

It is quite possibly charging up to two cells at a time, hence the reason it switches down to 4A max average with more than two cells inserted.  It must be averaging right around that 8A per cell to be able to get anywhere close to full charge on a 2Ah cell in 1/4 hour, though, regardless of whether it is a higher peak current being pulsed or a continuous 8A.

It could very well be doing two at a time, that would explain why the current is halved when the third and fourth cells are installed.
 
If I had that on my bench I would be so all over it with my scope to see what is actually going on.

[EEVblog]

No! Not that. To do -dV detection reliably you must measure the voltage while the charging current is still being applied. The -dV signal is actually measuring the drop in internal resistance of the cell that happens when the cell starts to warm up at full charge. Internal resistance is only successfully measured by examining the voltage change caused by an applied current. It cannot be measured open circuit.

Sorry to say that, but this is not correct.

IanB is in fact correct.
http://na.industrial.panasonic.com/sites/default/pidsa/files/panasonic_nimh_chargemethods.pdf
The voltage drops 5-10mV per cell when full charge is reached.

[mikerj]

However, that means the peak charging current would need to be 16 amps to get an 8 amp average with two cells (and a 4 amp average with 4 cells).  The switcher doesn't really look capable of delivering that kind of current, even at a low voltage.  What are the MOSFETs rated at?

Not if we are charging one battery at a time with eight amp pulses.
And one other thing; nobody said the eight amps was an average, continuous or peak.

8 amps multiplexed evenly between two cells is a maximum average of 4 amps each.  4 amps * 15 minutes = 1 ampere/hour, so you couldn't possibly charge a 2100mAh cell to 75% in 15 minutes.

[mikerj]

IanB is in fact correct.
http://na.industrial.panasonic.com/sites/default/pidsa/files/panasonic_nimh_chargemethods.pdf
The voltage drops 5-10mV per cell when full charge is reached.

I always understood the dv/dt was performed under load, but I noticed that my cheap Imax B6 hobby charger periodically stops charging on the NiMh/NiCd profile, presumably to sample the open circuit battery voltage.

[uwezi]

They are multiplexed.

That makes some sense I guess, the gate and cell sense voltages would always be referenced to ground then (albeit via a variable number of MOSFET RDSon drops).

However, that means the peak charging current would need to be 16 amps to get an 8 amp average with two cells (and a 4 amp average with 4 cells).  The switcher doesn't really look capable of delivering that kind of current, even at a low voltage.  What are the MOSFETs rated at?

If you look at the coil you will see that it has two parallel windings (at 17:32 into the video) and also given the quite big core I would assume that it could possibly be rated for 16+ amps. And the MOSFETs would be up to the job with up to 18A continuous and 50A pulsed drain current.

It would also possibly explain the weird use of the upper n-MOSFETs: instead of switching off these (which is not really possible, because the body diode would conduct) you turn on the lower n-MOSFET and bypass the battery in that slot. After all a constant current source would not like to be disconnected anyway...

And the discharge goes the same way, pulsed through the shunt resistor at the bottom of the chain, with all the measurement possibilities you need.

However, this also means that the batteries are exposed to pulsed currents of 8C instead of dc at 4C. But didn't I recently read something about a quick-charge method which used pulsed currents in order to optimize charge speed and battery health?

[Groucho2005]

Maha had this problem in their design of the C9000 charger. The first version to market tried to do -dV detection in the gaps between charging pulses and it became heavily criticized for missed terminations and overcharging. In later revisions Maha modified the firmware with an additional termination condition upon reaching an open circuit voltage of approximately 1.47 V. This terminates modern cells like Eneloops reliably, but leaves them at slightly less than a full charge.

I have the Maha (Powerex) C9000 and use exclusively Eneloops. It's the best charger I have used. The voltage termination and following trickle charge do charge the cells to almost 100% of their capacity. Besides, -dV algorithms don't work reliably at low charge currents and/or older cells.
Check out this excellent review of the C9000:
http://lygte-info.dk/review/Review%20Charger%20Powerex%20MH-C9000%20UK.html

[IanB]

I always understood the dv/dt was performed under load, but I noticed that my cheap Imax B6 hobby charger periodically stops charging on the NiMh/NiCd profile, presumably to sample the open circuit battery voltage.

If you measure the open circuit voltage immediately after stopping the charging current you get a residual "echo" of the charging voltage until it decays away. Although not as strong an indicator as the actual voltage under load it can still be used as a means to detect the -dV signal. Some chargers evidently are designed this way.

[nuno]

1/2 cells @ 8A or 3/4 cells @ 4A may have something to do with power, the power the DC-DC can provide. Although, if it were the case I guess they could do better, but there can be a limit in some other place like PCB current capacity and so on.

[BravoV]

I have the Maha (Powerex) C9000 and use exclusively Eneloops. It's the best charger I have used. The voltage termination and following trickle charge do charge the cells to almost 100% of their capacity. Besides, -dV algorithms don't work reliably at low charge currents and/or older cells.
Check out this excellent review of the C9000:
http://lygte-info.dk/review/Review%20Charger%20Powerex%20MH-C9000%20UK.html

Same here for C9000, have been using it for years, and my 1st gen eneloops are still performing very well, I'm guessing it has to do with they're not fully charged all this time. No complain here with C9000 + Eneloop combination. 

Just curious if there is better charger out there especially for Eneloop cells ? C9000 is quite old now.

[Skimask]

Could the DAVE CAD circuit show in the video be modified to charge a bank of 4 batteries, albeit one at a time, using an input source which would normally only be able to charge one battery, and at the same time, be able to draw a load from the entire pack?
For instance, 4 NiMH cells, Vout about 4.8v on average, an input V of only 3.3v, battery 1 gets a bit of a charge, switch, battery 2 gets a bit of a charge, switch, battery 3 gets a bit of a charge...and so on.  Of course there would have to be a boost circuit in there somewhere to get Vgs high enough to kick the FET gates.

[ez24]

Check out this excellent review of the C9000:
http://lygte-info.dk/review/Review%20Charger%20Powerex%20MH-C9000%20UK.html

lygte-info is a member here - but I forgot his forum name.

Lygte  are you in on this topic?

[HKJ]

I am reading here.

All the chargers I have tested looks to be doing -dv/dt detection with current off or use voltage termination. Besides connection resistance another reason to do it with current off is noise from the switching circuit, the filtering requirements are much less when measuring with current off..

My guess is that the charger is fixed 8A and it will timeshare the charging current with 3 or 4 batteries in the charger.

[apis]

Whenever reading about quick charging NiMHs they recommend dT/dt termination and using max-voltage, max-temperature and max-duration for backup since detecting dV/dt is going to be difficult.

Charging them in series shouldn't be a problem as long as you can detect when individual cells are fully charged and disconnect them, but I suppose you could pulse charge them one by one as well.

For detecting discharge current, could they somehow be grounding the positive battery terminals one by one and thereby measure current via the shunt resistor?

Don't really understand how it detects if it's an AA or AAA battery though, there must be more to it and all the 1 ohm resistors must have some purpose, but the only reason I can think of is current sensing?

This thing needs more probing.

[drussell]

I would love to see Dave get in there with a four trace scope and and see what sequence those MOSFETs are being switched in.

Oh, I forgot to mention....

I'd love to see a four trace scope of the four + battery terminals during a charge cycle.  Even just looking at the voltages there relative to each other would tell us pretty much exactly what it's doing with the charge profiles.  Current to each cell and investigating how the mosfets are being switched would be interesting also but even just seeing those voltage curves would be very revealing.