High Voltage Charging

These MAINS and HIGH VOLTAGE experiments can kill you!

In order to charge this car's battery it will need a high voltage charger.
Or we attempt to charge the 60 x 12v batteries in parallel.

Fortunately this is quite easy and cheap.

Mains power in the UK is 240vAC.
The packs in the Mass-EV are going to be arranged as 3 x 20 batteries.
Each battery is, of course 12vDC thus giving a pack voltage of 240vDC.
Total voltage of 3 x 240 = 720vDC.

This is, of course, the design to make charging from domestic mains easy and cheap.
When charging, the packs will be charged in parallel so we only need to charge a 240vDC pack.

If this gets sold in America I guess it will have to have 6 x 120vDC.

Rectified mains peaks at 240vAC x 1.414 (squareroot of 2) = 339.4vDC.
Or about 340vDC so has plenty of voltage difference to charge the battery.

We obviously need to control this as we don't what to overcharge the battery.
We can do this by controlling the AC input to the rectifier using a simple circuit
which normally is incorporated into a dimmer unit for domestic lights.

Original gschem schematic
This circuit was originally published on Electronics Project Design
I recreated the schematic from that image.

It works by removing part of the rising edge (or falling edge on negative sweep) of the phase using a triac circuit.
You can reduce the output by up to 50% linearly.
There's some information about it on epanorama.net.

You can further reduce the output by half wave rectification also, giving a pretty good range.
Certainly enough for the charging of the battery.

The dimmers are sold in DIY stores for under £10 so I purchased one and modified it with a more powerful heatsunk TRIAC.

Original datasheet
Pin Layout of the BTA26

The dimmer is from maplin:

Image of the dimmer from Maplin's website.

Some shots of the board removed from the dimmer:

Original Gimp Image
Better shot of the PCB with some detail.

Once the existing triac was removed it can be identified and the pinout discovered from the datasheet:

Original datasheet
Pin Layout of the BT136

Original Gimp Image

So now we can connect the new high power triac in place, via extending wires.
Of course, we could use this as a 6kW dimmer!

The heatsink and triac were £10 plus a bridge (£2) and a second bridge and switch to use one diode for half wave rectification.
Add a case and wiring and the whole charger comes in around £30.

This gives 13A at around 85v to 340v DC.
The 3 packs are 240vDC 40Ah each giving 240vDC 120Ah in total.
At a 13A charge this would be 120Ah / 13A = 9.23 hours.

So it's quite capable of charging 3 x 240vDC 40Ah lead acid pack to full charge overnight.
In reality you would charge at 12A for a 10 hour charge.

The unit components are actually rated up to 25A so this could charge much faster:
At a 25A charge this would be 120Ah / 25A = 4.8 hours.
For this you would need to have a special supply from your meter.
Also the batteries would probably need cooling, venting and monitoring since they are charging at a much higher rate.


So work finally began on this as it is now needed to charge the batteries we are going to use in the first prototype.

The circuit actually tested was:

Original gschem schematic
This seemed to work in that we could vary the average voltage of a 500w/250vAC lamp.
This was, of course, being fed with a variable DC voltage and not AC.

The capacitor was added to measure the RMS DC voltage and verify the meter's reading.
In the final ciruit it will not be needed as the battery does not need a smoothed supply.

There was an issue with the choke once the capacitor was added.
It was being heated.
I guess this was because the current was spiking as it recharged the capacitor during each phase.
This would cause high reactance in the choke.

As a test the choke was shorted to see if it worked OK, but this blew the 2A fuse quite dramatically.
This would again be because at switch on a high current would pass for the initial charge of the capacitor.

Original gimp image
Apart from the choke there was nothing even remotely warming up.
Not even the TRIAC, although it was only loaded to 4A.
It the real charger it will run around 12A which it still well below the 25A rating.

Even though the ciruit did OK, it would benefit from being designed more for this purpose.

Original gschem schematic


Since the lamp was loading the output the RMS voltage was read from the meter, but we really need to see the peak voltage.
This should be around 340v.

As the battery charges the load should decrease and the average voltage should rise.
This would be working as a rugged uniform current charger.

There is an issue with seeing the waveform.
My scope is 100v maximum.

I will need a potential divider to see the waveform using this scope.
I'm guessing 100k high and 10k + 1k low should give me around 10% scale so the voltage should not rise above 34v on the scope.

So we tested this ciruit and experimented with different values for R1, R2 and R3.
Using a value of 0 for R2 and 6.8k (high power resistors !) we didn't really get better than 75% phase.

The low power control was pretty good, but not being able to reach the full power was a limitation.

Inspecting the standard dimmer circuit I understand why it was in this configuration.
The gate was switched from the supply across the TRIAC itself so that when the TRIAC conducted it removed the gate input.
This means you can achieve near 100% phase if the gate is connected to MT2 directly.

Undertanding Triacs

TRIAC or Triode for Alternating Current is the equivalent of a thyristor for AC.

Original gschem schematic

Since AC effectively switches off every half cycle this means it does not lock on in the same way as a thyristor.
The TRIAC does lock on, but only until the end of the current half cycle.

It's like having PWM but without the complex timing circuit.
Since it's AC you get the timing bit for free.

So we tried out the redesigned circuit and it was indeed more stable, but the current in the gate circuit was a bit too high.
The gate circuit in the TRIAC as little resistance so will blow the TRIAC if connected directly to a supply.
This meant it would be better to connect it to the MT2 so the gate supply would be removed when the TRIAC conducted.

A revised circuit was tested:

Original gschem schematic

Charging Times

NiMH cell has an internal resistance of 0.17 ohms

We have 38 of 6 cells so 228 cells.

Total resistance is:
R = 228 * 0.17
R = 38.76 ohms

NiMH cell has a voltage of 1.2v per cell, so
Vbatt = 1.2 * 228
Vbatt = 273.6v

Terminal voltage was actually 285v so the cell voltage is actually 1.25v.
I believe this the no-load voltage.

The battery potential is raised from 285 to about 298 so
Vcharge = 298 - 285
Vcharge = 13v

I = V / R
I = 13 / 38.76
I = 0.335A
I = 335mA

This is about C/20 so sub-trickle charge,
but higher than a maintainence (C/100)

Capacity of a cell is 6.5Ah so charge time is
t = C / Icharge * 1.2
t = 6.5 / 0.335 * 1.2
t = 23.3 hours

Real Battery Charging

These MAINS and HIGH VOLTAGE experiments can kill you!

Having 3 ex-NHW11 Prius batteries, but no charger, I feel the need to make one.

The basic idea is to just use rectified mains directly to charge the 273vDC battery.

Since rectified mains (in the UK) is over 300vDC this should be OK to at least trickle charge them.

From: HyperPhysics: Charging a Capacitor

Original gschem schematic
I have some resistors which appear to be 2w (Resistors),
so using 5 parallel capacitors connected directly to one terminal for testing:

I filled in the values on the above circuit.

Testing on Real Batteries

Started with a relatively safe circuit which was extremely low current just to get a feel for the batteries reaction
This is a test which I expanded to charge 3 complete full batteries in parallel which were all very low.

Original gschem schematic

Anyone who tells you that you need an expensive charger for Prius batteries just show them this :)

As the battery voltage rose I swapped out R1, R2 and R3 for 680 ohms (5w) at 200v
then 10 ohms (60w) at 280v and the batteries eventually rose to 320v.

200v is an average of 0.88v per cell.
280v is an average of 1.23v per cell.
320v is an average of 1.4v per cell.

Reading CHARGE METHODS FOR NICKEL METAL HYDRIDE BATTERIES it seems this was the best method for resurrecting these batteries.

"To charge excessively discharged or deep-discharged batteries, first allow a trickle current to flow,
and then proceed with the rapid charge current once the battery voltage has risen.
Rapid charge start voltage: Approx. 0.8V/cell
Rapid charge transition voltage restoration current: 0.2 ~ 0.3 CmA"

In this document it refers to "CmA" which is assuming the capacity (C) is in mAh.
This is indicating we should use a charge of C/5 to C/3.33.

Pretty much all these cells were well below the 0.8v, in fact alot would be 0v.

MCU Controlled PWM

Since, in a different project here MCU control of mains is actually quite easy and cheap,
This can be extended using AC-DC then using an ADC to measure and control voltage.

So a first basic AC circuit driving a lamp was constructed:

Original gschem schematic
This was adapted to DC with smoothing and symmetric cross detect:

Original gschem schematic
Peak voltage across the LP1 was +320vDC on full and went to 0v.


Also added a low power 5vDC supply for the Arduino.

Original gschem schematic

This is OK, but R1 gets very hot.
A better way would be to use a PWM switch mode for the 5vDC too, using the resistor circuit to bootstrap.

Moving away from pure transformerless as this can be unstable.
Just using a small transformer for a simple PSU to power the Arduino.

Also updated to include sensor information and lower level software:

Original gschem schematic

415vAC 3-phase

These 3-phase MAINS, HIGH VOLTAGE experiments can kill you!

The battery packs in the Mass-EV v0.1 are 3 x ~400vDC, so this will not charge from 240vAC even rectified to approx 330vDC.
In order to charge these a 3-phase charger using the 415vAC would be needed, unless we use a step-up transformer.

Original gschem schematic
So 3-phase 415vAC UK mains rectifies to 587vDC.

To start with a primitive charge using a resistor dropdown:

Original gschem schematic

This would take in the region of 1.5 months to charge from flat to full.
Really targeting something in the region of a week to be useful at the moment.

Which would be this:

Original gschem schematic
This is a very basic charger, not really that cheap since it uses a £16 resistor.
But it's reliable would work at C/100.

Top left is the pack voltage (418vDC) bottom left is the input voltage (583vDC) and you can see the current is 820mA.
The resistor is dissipating about 130W, well within the rated 200W.
This was charging 32 x 12v 100Ah batteries (384vDC total), which of course was no problem.

Looking at testing an IGBT controlled PWM.

Original gschem schematic
This is an untested triac design, which would be cheaper than IGBT.

Original gschem schematic
This should provide up to 580vDC which means it would charge a 400vDC battery.

G-Wiz Charging

The battery in a G-Wiz is a simple 48v Lead Acid wet cell, so this doesn't need a sophisticated charger.
This is just a 500w transformer, bridge and smoother and using a 10R/60W resistor for current limit.

Charge rate is about 1A into 200Ah battery so C/200.
The open circuit voltage is about 80v which is 3.33v per cell.

This can't be used for trickle charge as the gassing voltage is less than 2.5v per cell.
It would need to be limited to 60v for the pack for that which can be done with a LM317.

48v EV Motorbike charger

Using a purpose build charger from a 400va 48v transformer we get better results:

The secondary is 2 x 24vAC so in series is 48vAC.

Original gschem schematic
T1 secondary:

Yellow:open circuit, Blue:charging

The open circuit voltage is -78 to +77, so 77vAC amplitude which is 54v RMS
Charging this is reduced to -70 to +70, so 70vAC amplitude which is 50v RMS

Bridge rectifier got a bit hot so added a heatsink, and that cured it.

Charge rate is 5-10amps which is good enough for an overnight charger.

Parallel Charging

Parallel charging means we can charge at a low voltage.

How is this done with a series connected battery?
One at a time, with an isolated charger.

Knightbridge optocouplers are 10p each for 4 on a chip.
This is possibly cheaper than the extra wires to connect back to a central processor board.

This is a simple one to get the theory:

Original gschem schematic

Original inkscape drawing | Drilling mask
So a module can be made to use binary (gray code) selection and a single power cable pair to index the cell for (dis)charge

This circuit is for 4 cells, but the schema would work up to 16 as we have 4-bits:

Original gschem schematic

Original inkscape drawing | Drilling mask

This will need diodes in series with the opto LED to stop conduction across the 5v rails via the opto LEDs.
Probably need 4 (@ 0.7v each = 2.8v) plus the opto LED (@ 1.5v + diodes = 4.4v) this would be 8.8v for both sides so would not self-conduct