Poor man's 15 Amp battery charger?

[Alan B]

It would be good to sense the voltage difference between the ground and neutral connections, and refuse to operate if they are more than a few volts apart. Then perhaps one processor would be adequate. If two processors were to be used I'd want the second one to be a redundant safety control, we really don't want this charger to not shut down. The micro with the user interface is more prone to problems, so the other one should be really simple and insure the charging cycle is properly terminated. Redundant means they sense and turn off the cycle in different independent ways. So one runs the solid state relays, and the other operates a master relay that is not solid state and has actual airgap between the contacts.

I don't think a GFCI outlet protects against a neutral-hot swap, though it could.

Individual fuses on each capacitor would be nice. They could be lower values that way and a capacitor short would be much less exciting, and the charger would even mostly still work.

I have a 1000V 30A bridge on order.

[bearing]

Would it be possible to put a zener diode between the battery+ and the neutral, to trigger the GFCI if battery voltage gets too high?

Does the battery get a little warm when it starts to get overcharged? In that case you could use a mechanical termostat mounted on the battery to disconnect the charger, as a last resort.
It won't help if the battery only starts to get warm when the thermal runaway has been triggered, though.

[oldswamm]

ga2500ev, you should check your PMs.

I drew all this a couple days ago, then got busy (and also got sick), so I didn't post till now.
And keep in mind that I'm trying to design for 'everyone', not just myself, so would like it to work with 12 to 150V. I know this is a long post, but I tried to be concise, so please be patient while reading it.

First, I mentioned using a crowbar on the output earlier, but some might not know what it is, or how it would be implemented. The idea is to absolutely disconnect the battery from the charger. (Think of throwing a metal bar across a set of power leads.)
I haven't worked with SCRs for decades, or done any research, so left the interface as a box with a ? mark for now. The user interface (main) processor would send a 'watchdog' signal every 100ms or whatever, and if the crowbar interface doesn't see it, or if it's not the expected format, it will drop the crowbar. The user interface could also send a max over voltage spec to the crowbar processor, so it could sense over voltage and drop, but if the main processor was in an unintentional loop, it could possibly sent the wrong voltage (not absolutely foolproof), but in that case it probably wouldn't send the 'watchdog' signal. The zener could be replaced with an LM431. The over voltage portion of circuit wouldn't do any good if the processor is programed for too high a voltage either.

crowbar sch.jpg (27.37 KiB) Viewed 515 times

Next is the circuit I'm considering for putting a serial number ID on each battery. (Not having to program the charger every time you plug in a battery is part of making the system fool proof. If the user doesn't want this feature, leave it off!) The charger would then be programed with the charging parameters for that ID #. If said parameters change, as with removing a bad cell, all you would have to do would be to change the parameters stored in the charger.
Starting from the upper left corner and going clockwise I will try and describe what I am trying to do, and why I chose the components.
The 900M resistor would supply current, and would be different for different voltage ranges. The processor would sleep around 98% of the time for low current consumption (<100uA). The circuit with the 100k resistor is used to detect charge pulses, and their timing so we can transmit at a low noise point (wake on rise). The next circuit would ideally convert the 150khz square wave output of the processor to a sine wave superimposed on the Batt+. The diodes would be to absorb spikes during connect and disconnect. The resistors with the T are thermistors for measuring batt temps. They are the reason for the processor choice. The touch sensor inputs make resistance measurements easy (they put an accurate current source out, then measure the voltage). I would hope it's obvious why we want battery temperature sensors.
For the signal I'm thinking of a pulsed 150khz tone. By defining a 0 as no pulse followed by it's mirror image, and a 1 as a pulse followed by it's mirror, I have a sync check no more than every 2 bits, as well as bit by bit redundancy The red dotted lines represent the sample points for the main processor.
I show 9 bits for the serial, but if it went past 256, we could use the first bit = 1 to increase the number of bits in the SN. 6 bits should be plenty for temp for this application, right?
<edit> This would be a blob of epoxy with a red and a black wire and 2 wires for thermistors from it, marked with a SN and voltage range.

Batt SID sch.jpg (32.33 KiB) Viewed 515 times

This is what I've come up with for the main board. After much effort, ga2500ev convinced me to use 2 isolated circuits. For myself it would be unnecessary (IMHO), but since I'm designing for everyman.... This way it can be assembled with whatever button switches you have laying around.
Again clockwise, the transformer could be 1.1VA for power for the board only, or 2.4 to run a small fan, or 6VA to run most any fan, or even 2. It's drawn as 120v, but if I designed a 'production' board it would default to 220, and would require cutting a trace and soldering in jumpers for 110 (better to connect 110 to a 220 circuit than 220 to 110). It would supply power for the isolated processor through the 5v regulator.
The resistor limited power supply for the power side (unisolated) processor is fairly straight forward although not efficient (and not good for less than 16s as drawn). Again, the zener could be replaced with an LM431. Next a voltage divider to measure BattV. The first op amp and associated filter would be to convert the 150khz tone pulses to relatively noise free pulse inputs to the processor. (Input from those more experienced with this type of circuit would be appreciated. This is all off the top of my head with no research.) The shunt amp is straight forward. It could use 1, 2, 3, or 6 shunts for 12 to 75A at 1/2 the rated resistor dissipation. If you're willing to dissipate 1w from 1w shunts it would be good for 100A. I think it should be sampled every 100uS or so and averaged from 0 to 0 (I would appreciate it if someone with one of these operational would post an o-scope pic of the battery current). The opto is used to isolate the 2 processors, and there would be serial data with voltage, current, and battery ID data every 17ms(?) or so. Since accuracy of voltage measurement is critical, and the internal references are NOT accurate, I opted to use an external reference (the LM4040). We don't need accuracy at this point, as I would provide for calibration, but we do need repeatability. The only reason to use the 12F510 here is to minimize the number of different processors I need to buy.
The 18F26k80 processor is the user interface. The reason I chose it, is I wanted more than 256 bytes of EEPROM, for various program and battery parameters, and didn't want to use an external IC. Note that I changed the isolated power labels to VSS' and VDD'. Since I now have lots of pins I redrew the LCD interface for 8bit. Does anybody need/want r/w? Since I have lots of pins I included a fan tach input (could be 2). The serial output to the crowbar should be obvious. The HVC input is intended for the optoisolated output of a high cell voltage detector such as methods is selling. PB are push button inputs, which could be capacitive sensors. I added open collector transistors to the SSR outputs so 12v SSRs or even mechanical relays could be used. The transistors could be left out if using low current 5V SSRs. The 10k resistor to the transformer is a simple 'zero crossing' detector to get a 50/60 hz input for the timer.

capacitive chgr control sch.jpg (51.6 KiB) Viewed 515 times

This circuit isn't really part of the above, but I mentioned it earlier, and probably nobody understood what I was getting at. We in the US tend to have rather wimpy wall power, especially if you are connecting to someone else's power where you can't be sure of the breaker capacity. This is a circuit so you can have 2 cords (with GFCI plugs) coming from your charger for double the power. The danger is that you could connect to outlets on different legs (220). If that happened with this circuit, the relay would open and the bulbs would glow. The resistor would be needed if the bulbs flow more current than the relay uses, and the diode wouldn't be needed with an AC relay. With a processor, you would disconnect the first 5 SSRs then connect SSR6 for 1/2 cycle every couple seconds. If current is present with SSR6 active, that cord is connected, but it could still be to the wrong leg and the relay just hasn't had time to open, so you would want the processor to wait a few hundred mS before connecting SSR6 and the other SSRs at the same time.

2 plug setup.jpg (17.17 KiB) Viewed 515 times

[317537]

I made a decent hvc cut off from an dpdt old relay for my solar system, I tuned it to shut down at 14.4 and every day it did just that with out failure. I used a zenner because this thing liked to switch at about 8v, a resistor to mimic a float, and I switched to and use the solar panels to keep the reed off after full charge , only until night the coil-reed would reactivate until morning. With 6.2v zener (I think) and the V drop over the relay coil and fine tune pot, it drew little current at rest voltages what 4v through a 120ma coild @12v lucky to draw 50ma. and the relay could handle 25 amps. I had to add a blocking diod somewhere as this thing would go zzzzzzzzzzzzzzzzzzzzzzt on and off so fast, I sort of wished I wrote this circuit down. I may recreate this device and see if I can make them shut off at 56v or something.

The relay came from an old GM controller and they have a few poles and what nots on them. They were used for LVC in those controllers. I remember hearing the click when the voltage got too low on my SLA's.

[oldswamm]

Is there a way to distinguish between the term for a high cell voltage monitor digital output, and the voltage across the entire battery.
In this thread HVC seams to stand for the voltage across the entire battery, whereby I thought it stood for 'high voltage-cell'. Apparently this is another acronym that has multiple meanings, which is confusing.
Do we need to come up with a new acronym for the voltages across the high and low cells, which are our real limits with lithium batteries?
Maybe we should start using HCV even though that makes it sound like an analog value rather than digital, and of course then all the BMS and HVC/LVC monitor manufactures would have to change their terms, which they aren't going to do since their outputs mean one cell is high or low.....
I guess I'm going to have to use the more verbose term 'High Voltage-Cell'. Hard to fit on a schematic though....
Bob

[ga2500ev]

Is there a way to distinguish between the term for a high cell voltage monitor digital output, and the voltage across the entire battery.
In this thread HVC seams to stand for the voltage across the entire battery, whereby I thought it stood for 'high voltage-cell'. Apparently this is another acronym that has multiple meanings, which is confusing.

Welcome to the world of battery management. Both techniques are used. Many in the EV world have moved to thinking that monitoring at the cell level is overkill, but that whole battery monitor is insufficient. The current discussion has turned to the half pack monitor:

http://www.diyelectriccar.com/forums/showthread.php/battery-balance-monitoring-system-54845.html

Which measures the voltage difference between the two halves of the battery. In both charge and discharge modes, if one cell gets out of whack, the total voltage for the half of the battery that out of whack cell is located will change radically as compared to the other half of the battery. Measuring that difference is sufficient to indicate that either charging or discharging should be discontinued.

Do we need to come up with a new acronym for the voltages across the high and low cells, which are our real limits with lithium batteries?

They are called cell monitors.

Maybe we should start using HCV even though that makes it sound like an analog value rather than digital, and of course then all the BMS and HVC/LVC monitor manufactures would have to change their terms, which they aren't going to do since their outputs mean one cell is high or low.....
I guess I'm going to have to use the more verbose term 'High Voltage-Cell'. Hard to fit on a schematic though....
Bob

The term applies to both cases. The rationale is that if the cells are all balanced, then the sum of N cells in series is equivalent to N * (HVC of a single cell). So if you have 20 cells, and the single cell cutoff is 3.7V, then the HVC for the battery is 74V.

The failure occurs when the cells are not balanced. With lithium the charge/discharge curve is virtually flat during the entire charge discharge cycle. So both the cell voltage and the battery voltage is stable until one cell completely charges/discharges while the others still have available capacity. In charging, that single cell's voltage shoots off the chart, while the others stay flat. Say the per-cell charging voltage is 3.5V. So in the above scenario, the battery voltage will be 70V flatlined during charging. One fills early because it out of balance and it's voltage rises to 3.7V. Since all the others are still at 3.5V, only 0.2V is added to the battery total giving a total of 70.2V, well below the 74V battery HVC. So we continue to charge at full power. That cell, if it's way out of whack with the others in the pack would literally need to get to 7.5V before it'll trigger the battery HVC. But of course by that time, the cell is destroyed.

With cell monitoring, a signal would be generated when the first cell reaches 3.7V. Typical BMS systems will either shunt or redistribute the excess energy around the cell(in a top balancing scenario) , or will direct the charger to turn off (in a bottom balancing) one. In the top balacing case, the expectation is that the other cells are not far behind, and so often the single cell trigger will direct the charger to cut back the amps to a trickle making it easier for the charging energy to be redirected.

But now it's easy to see how half pack monitoring can accomplish the same task. When the first cells starts to shoot off to the stratosphere, its half-pack voltage will rise significantly relative to the other half-pack. If one sets a cutoff of say 0.4V difference, then a single cell would not be able to get above 3.9V before triggering the alarm. Of course you'd still need the whole-pack HVC so that if the pack is perfectly balanced and everyone rises to 3.7V/cell and 74V for the pack together, that the charger still will cut off.

So digitally, the inputs would need to be the whole-pack voltage and the half-pack voltage to the monitor controller. From those two, the difference between the measured half-pack voltage and the expected whole-pack voltage/2 can be computed.

Hope this helps...

ga2500ev

[oldswamm]

Any old timers want to weigh in as to what HVC stands for? I did a google search and the closest I found, out of over 50 definitions, was 'high voltage circuit' and 'high voltage capacitor'.

I know some people use the M in BMS to mean monitor, so you have to ask what they mean by BMS (whether it balances or not).

That link on DIYCAR discusses how to compare the 2 halves of a battery for 6 pages without anyone pointing out that it might not be a good idea, although they're using it for seeing an imbalance when driving, not charging. The worst you can do while driving is ruin a cell.

For charging I would consider it to be a bad idea. This goes much more for LiPo. If you have 2 high cells, one in each half, it's a matter of which catches fire first. If you have a low cell in the same half as the high cell you will be in the same danger. This does NOT qualify as foolproof.

For a bike with LiPo, the extra cost of monitoring each cell is high, but not as high as a fire.
For a car with LiFePo, the extra cost is trivial compared to the cost of ruined cells (again IMHO).

ga2500ev, I stick by my choices of processors for the already stated reasons. The 12F part you suggested (via PM) costs double what the 12F510 does, and is less widely available. The 16F1938 doesn't have 1k of EEPROM, and given the accuracy of the internal reference (+8/-6%) it probably doesn't have a decent temperature coefficient either (I couldn't find a spec). With a calibration routine, the accuracy of the LM4040 (.1%) doesn't matter, but we need repeatability to end up with accurate voltage measurement. The LM4040 probably wouldn't make it into production if I found a cheaper part with as good of repeatability, but if this is a one off project, the cost ($1.70) is irrelevant.

I would like some feedback.

Is anyone other than ga2500ev interested in what I'm doing here, or am I wasting my time?

Does anyone but me think automatic battery ID with battery temp monitoring is useful? Is wireless objectionable and people would really rather have a bunch of extra wires?

How about current monitoring with subsequent Watt-hour calculation?

Is multiple current selection something that nobody would use?

No comment on a crowbar as opposed to an extra mechanical relay (which could weld). It would ONLY activate in the case of a GROSS malfunction of the charger control. The only problem I can see would be if you were charging (for example) a 2Ahr sla with a charger fused for 100A, the battery and charger wouldn't be able to blow the fuses(<edit> Alan's idea of separate fuses on each cap would solve this on the chg side, if properly sized).
The crowbar processor could be configured to release the redundant relay(s) which would add to the cost and size (one advantage to that would be that the charger COULDN'T activate without a battery connected since this processor gets it's power from the battery).

Bob

[317537]

Any old timers want to weigh in as to what HVC stands for? I did a google search and the closest I found, out of over 50 definitions, was 'high voltage circuit' and 'high voltage capacitor'.

Bob

Did I say HVC for BMS? If we type the letters "LVC " I certainly do not get our google definition return LVC as low voltage cutoff, or otherwze we have been suggesting low voltage CMOS.

These are acronyms and they can mean anything you want them to mean. Like my bike is a POS. Or, WTH? STHU with this BS. We just use them to make ease of both language iteration and literation..

I ment the relay idea for the voltage cut off (VCO? or HVC?) for a poor mans charger at full charge.

The BBS, battery balance system does not require to be onboard the pack for the ride. LVC only requires a simple analogue device to measure all cells down the series. It just makes it easier to plug in a two wire charger instead of a 18 wire charger 16s. This concept sort of escapes the realm of the poor mans charger definition , or PMC.

I turned a crappy thowout 5 amp 15v(18v peak) power transformer into 48v charger with an AC voltage multplier, 18v*4=72v.

BTW the AC voltage multplier was the best SLA charger I ever used and it used to desulphanate very well. The multpier took 15v @ 5A to 48v @ 1.2 amp to 60v @ 700ma using a more pulse mode charge via the AC rectification from initial charge to over charge, by the multimeter, inefficiencies inclusive. The current to voltage knee current was perfect in its ability to pack energy into an SLA. , I could of used my relay VCO if I had invented it at the time but most of the time it was easy to catch the charge state at full charge because it would take so long to get from 95% cpacity to 100%. in most cases 95% was fine. I think I only overcharged twice in 8mths I used it and all this seem to do is knock sulphur off the plates.

[317537]

Everything can fails, even fails proof fails. A Wireless transmission, crow bar, digital, analogue, wire connections dry joints. You name it, it can fail and leave you in a puddle of doo.

If I trust my BMS I could end up in as much trouble or even more than if I trust my witts. Electronic reliablility vs Human error. In most case both is advisable. But the learning process requires one to go through the motion of being alert at all times and manual mangament is the best learners awarness course.

The best solution is often the least complicated and the essence of this thread can be, or better said, should be maintained with the ideal of safety being intact.

[oldswamm]

I think a lot of the 'disagreement' is because we each have a chemistry were thinking about.
When I think safety, I'm thinking primarily LiPo.

Leslie, you wouldn't recommend this charger, under manual control, into LiPo, especially if we're shooting for 2C+, as a 'learning process' would you?
Most people I know would learn that lesson real soon. (With me near the top of the list)

At any rate if we can make a charger safe for LiPo, just changing the charging parameters should make it safe with other lithium chemistries, and then all we have to do is add a drop detector and we can do the Nickle chemistries, every bit as safe as a commercial charger.
If the nickle based batteries had an ID chip (or if you only have one battery), the charger could 'learn' the normal knee parameters for even more dependable detection.
You surely wouldn't try and charge nickle based batteries manually, would you.

SLA are easy, although it occurs to me that (if it can ID it) the charger could be programed to do a regular maintenance charge.

Are you really recommending that I forget it, because people are better off with a manual or jury rigged setup?

To answer your points, if the ID fails, the battery will have to be programed in manually, just like a battery that doesn't have an ID chip, at least till you fix it.

The crowbar is a totally separate circuit (one of the important redundancies). It would require 2 different serial streams, generated by 2 different routines, in different parts of the program to reset it's timer, on top of which it will drop if the voltage goes above the set trip point. The crowbar's problem of a 100A charger connected to a low current battery can be solved by a low amperage fuse in the battery's charge line (not a bad idea anyway), and if we add fuses on individual caps, AND an extra relay on the input as Alan suggests, operated by the crowbar processor, we solve any possible charge side fuse problems. If you want to trust the redundant relay, the crowbar could be eliminated (it becomes redundant ). I guess I need to rename the crowbar circuit when I redraw it.

The idea is, if there's a failure, it will be that the charger 'doesn't work'. This is the safe failure mode. If it overcharges a battery, it will be because the end user miss set it, more than likely, if they have more than one battery, by charging one with the charger set for another, which is the problem I'm trying to address with the battery IDs. If you only have one battery, you probably won't miss set a charger.

Digital/analogue. Yes. What computers, including microcontrollers are good at, is analyzing feedback from the 'outside' (analogue) world.
I wouldn't consider doing what I'm doing with pure analogue electronics if that's what you meant. A foolproof seven (or more) chemistry, wide range, analogue circuit would be a nightmare. Computers are so much easier to change parameters on as well.

If done right connections can be extremely dependable. Almost everything but the fuses would be soldered in mine. I'm no fan of connectors. I even like screw mounted fuses.... Tie the wiring away from danger.

My ideal for safety is to design each aspect as safe and dependable as I can, with multiple redundancies, not simplicity for the sake of simplicity.

The design I'm working on offers numerous OPTIONS. You can use what you feel you need or want. The conditions that cause LiPo to burst into flame, desulfate lead acid.

Bob

[fechter]

To me,
LVC= Low Voltage Cutoff
HVC=High Voltage Cutoff
BMS= Battery Management System

HVC and LVC are typically referring to individual cell measurements.

These terms were really just made up by guys on this forum and do not necessarily comply with industry standard terminology.

Industry standards are more likely
OVP= Over Voltage Protection
UVP= Under Voltage Protection

These terms could apply to either individual cells or the entire pack.

Any battery charger, regardless of whether it's an AC capacitor limited unisolated thing this thread is about, or the most sophisticated power factor correcting switching mode charger will still have the same issue of individual cell voltages vs. pack voltages. Individual cell voltage monitoring is going to be safer, especially if the 'cells' are made from multiple parallel cells.

As many people have found, if you never get too close to fully charged or fully discharged, cell level monitoring is less important, but I personally recommend it for any battery.

I think the circuit I posted way back will be adequate for safe charging, but not nearly as fancy as a microprocessor based one. There's something to be said for keeping things as simple as possible, but you still need to consider likely failure modes and the results of them to maintain some degree of safety. Nothing will be really foolproof.

[IBScootn]

Farfle & Skippic,

How are your chargers working? Has anyone hooked up one of these "bad boy" chargers to the J1772 changing stations yet (using one of the available J1772 to 240Vac adapters)? Or any other standard 240Vac charging station? The only use I would have for one of these chargers would be for fast charging off a J1772 station.

My concern is whether the smart public charging stations will except a charger with such a poor power factor. To get around the poor PF issue, I've been wondering if I shouldn't just build a charger that uses a power resistive load (heating element w/fan or element boiling away water) to limit the current instead of a motor-run cap or SSR. Most motorcycles have a muffler for noise/heat output. Since I have an e-motorcycle I don't have a muffler, but that doesn't mean I couldn't make a similar device to exhaust my current-limiting heat. Not sure how to calculate how much heat I would have to waste? I would want to charge my lifepo4 pack from min of 77Vdc to 83Vdc max at 20A. So I take the 240Vac input, rectify to 340Vdc, then take 340V-77V and would have 263V to drop across the heating element?? At 20A, that would require ~ 5.2KW rated heating device or equiv to 3 hair dryers?? Is that correct?

[IBScootn]

P.S. I wouldn't care about wasting energy as heat as these stations are currently free to use or charged by time. So, I'm more concerned with building a very cheap, high current, small charger than how efficient it is.

[parabellum]

I've been wondering if I shouldn't just build a charger that uses a power resistive load (heating element w/fan or element boiling away water) to limit the current instead of a motor-run cap or SSR

http://endless-sphere.com/forums/viewto ... 45#p507804
http://www.youtube.com/watch?v=3tN7vkgC ... e=youtu.be

[Skippic]

Mine is holding up great. I've been using a 70uF motor run capacitor with 3ohm resistors and something like 3000mF electrolytic caps to stabilize the current. For control I used a voltage divider, an Arduino and the TextStar display. The Arduino measurement is far from accurate, so I've been averaging 100k measurements to display the battery voltage. To turn charging on and off I used an SSR:
http://www.ebay.com/itm/SSR-25A-Solid-S ... 4ab6626024

Now I'm about to try using a 400uF cap as suggested by Farfle (big thanks):
http://www.amazon.com/dp/B005FUXGZG/ref ... _3p_dp_1_B

In the second charger I'm eliminating the resistors (maybe even the stabilizing caps).
Instead of the Arduino I'll use a much simpler and cheaper solution. My idea is to use a SSR, voltage dividers, voltage detectors and a cheap digital display.

Should be lighter, smaller with less connections and components to fail.

I'm attaching my Arduino code (dirty). I tried attaching the file, but ES didn't like it, so I tried renaming it to .txt, but it didn't help.

/*
AnalogReadSerial
Reads an analog input on pin 0, prints the result to the serial monitor

This example code is in the public domain.
*/
#include <SoftwareSerial.h>

/** Pins *******************************************************/
//#define rxPin 4 // define the RX pin
//#define txPin 5 // define the TX pin

#define maxV 855
#define minV 770
#define coef 8.77



/*void lcd_WriteLine(int lineNum);
void lcd_ShowLine(int lineNum);
void lcd_init();
*/
//SoftwareSerial buttonSerial = SoftwareSerial(rxPin, txPin);


//int Line1 = 0;
//float Line2 = 0;
float finish = 0;



void setup() {

//pinMode(rxPin, INPUT);
//pinMode(txPin, OUTPUT);



Serial.begin(9600);// sets the serial BAUD rate, must match that of the display

lcd_init();


delay(2000); // 2 second delay to allow screen to turn on if the power has only just been turned on.

Serial.write(254);// tells display we are sending a command
Serial.write(76);// tells the display to go to a line
Serial.write(1);// clears the line 1 aready for re-filling
Serial.println(""); // prints instructions
Serial.println(""); // prints instructions

lcd_WriteLine(2);
Serial.println("");
lcd_ShowLine(2);
}




void loop() {
String Line1 = "";
String Line2 = "";
int voltage = 0;
// float sum_voltage = 0;
int aux = 0;
int high = 0;




float sum_voltage = 0;

for(float i=0; i<100000; i++)
{
voltage = analogRead(A0);
if(voltage > (maxV+30))
{
analogWrite(13, 0);
high = 1;
}

if(voltage < (minV-100))
{
analogWrite(13, 0);
high = 1;
}

sum_voltage += voltage;
}

if(high == 1) finish++;
if((sum_voltage/100000) < (maxV))
{
if(finish < 10) analogWrite(13, 255);
}
else
{
analogWrite(13, 0);
finish++;
}
// Line1 = maxV;
// Line2 = sum_voltage/100000;

lcd_WriteLine(1);
Line1 = String("M");
Line1 += String(int(maxV/coef));
Line1 += String('.');
Line1 += String(int(((maxV/coef) - int(maxV/coef))*100));
Line1 += String("[");
Line1 += String(maxV);
Line1 += String("]");
Line1 += String(int((sum_voltage/100000-770)/(maxV-minV)*100));
Line1 += String("%");
Serial.println(Line1);

lcd_WriteLine(2);
Line1 = "A";
Line1 += int(sum_voltage/100000/coef);
Line1 += '.';
Line1 += int(((sum_voltage/100000/coef) - int(sum_voltage/100000/coef)) * 100);
Line1 += "V[";
Line1 += int(sum_voltage/100000);
Line1 += "]";
Line1 += int(finish);
Serial.println(Line1);




/* lcd_WriteLine(2);
Serial.print((Line2-770)/(maxV-minV)*100);
Serial.print(" ");
Serial.println(finish);
*/
lcd_ShowLine(1);

// Serial.println(sensorValue, DEC);
}





/* Sends LCD command sequence to initialise the LCD
* Sets cursor style to 'no cursor'
*/
void lcd_init() {
delay(2500); // let the screen initialise if it has only just been powered up

/* Clear the display and take us to line 1 cell 1
* Notice that this instruction to the LCD does not
* require a command-bit to be send first
*/
//Serial.println(12,BYTE);

/* Setup the Cursor Style */
Serial.write(254); // we are sending a command
Serial.write(67); // we wish to set the cursor style
//Serial.println(0,BYTE); // we wish to have no cursor
}

/* Sends LCD command sequence to write to LCD line n
* @param int lineNum Line Number to write to
*/
void lcd_WriteLine(int lineNum) {
Serial.write(254); // we are sending a command
Serial.write(76); // goto line n command
Serial.write(lineNum); // clears the line the line number (1-16)
}

/* Sends LCD command sequence to display to LCD line n
* @param int lineNum Line Number to display (top of screen)
* will show n and n+1
*/
void lcd_ShowLine(int lineNum) {
Serial.write(254); // we are sending a command
Serial.write(71); // display the following line number
Serial.write(lineNum); // the line number (1-16)
}

[Skippic]

Forgot to answer the question

No I haven't connected it to 220/240V yet.

[Farfle]

Mine is holding up great. I've been using a 70uF motor run capacitor with 3ohm resistors and something like 3000mF electrolytic caps to stabilize the current. For control I used a voltage divider, an Arduino and the TextStar display. The Arduino measurement is far from accurate, so I've been averaging 100k measurements to display the battery voltage. To turn charging on and off I used an SSR:
http://www.ebay.com/itm/SSR-25A-Solid-S ... 4ab6626024

Now I'm about to try using a 400uF cap as suggested by Farfle (big thanks):
http://www.amazon.com/dp/B005FUXGZG/ref ... _3p_dp_1_B

In the second charger I'm eliminating the resistors (maybe even the stabilizing caps).
Instead of the Arduino I'll use a much simpler and cheaper solution. My idea is to use a SSR, voltage dividers, voltage detectors and a cheap digital display.

Should be lighter, smaller with less connections and components to fail.

I'm attaching my Arduino code (dirty). I tried attaching the file, but ES didn't like it, so I tried renaming it to .txt, but it didn't help.

Don't use that cap! they are falsely advertised as motor run caps, they are only start caps and are not rated for continuous duty. Motor run caps only go up to about 100uf, always have metal cans and are physically larger for the same capacity.

[Skippic]

Don't use that cap!

Do you mean, the 400uF one you suggested?

[Farfle]

yup, the ones i got failed within 15 minutes of use, not an interesting failure, some hot oil leaked out and the current dropped to zero. But they did fail very quickly
*edit*, thought the cap was labeled as a motor run cap, its not falsely advertised, but I did find a similar one that was mis-advertised *edit*

[Skippic]

So then:
http://stores.directhvacsupply.com/-str ... gories.bok

I'm getting some 10 to make 1 portable charger and a fast one for home.

[Skippic]

What do you guys recommend to use as a rectifying bridge? I'm looking for something, that will create as little heat as possible. Say 50A max.

[Skippic]

From my simulations I see a great benefit from having higher mains voltage.

For a 85uF capacitor and 82V battery:
@ 110V 1.2Aaverage 5Amax
@ 220V 3.9Aaverage 9.3Amax
@ 340V 6.9Aaverage 15Amax

So for the same capacity you get much faster charging. You could in theory get caps for smaller voltage, but it looks like 370V and 440V is standard.

Furthermore even if you get more capacity, the spikes at lower voltage are much higher compared to the average current:
@ 110V 700uF 10Aaverage 37.5Amax
@ 220V 215uF 10Aaverage 27Amax
@ 340V 123uF 10Aaverage 21.5Amax

The capacity with these cheap caps is not such a big issue, the max current, if you want a fast charge is.

[Skippic]

I just received the 10 85uF 440V capacitors. While in the US I will connect all 10 to get 12A average with 45A max. My batteries are rated 5C (charging) so with 10Ah up to 50A should be OK.

I will of course put a selection switch for 85uF/850uF (to slow charge when I have time).

My current charger is pretty complex with 70uF cap, SSR, rectifying bridge, electrolytic caps, resistors, Arduino, 5V power supply...

In the next one I'm eliminating the electrolytic caps, resistors, Arduino and the 5V power supply.

The end of charge will be determined by the pack voltage. I'll use a resistive voltage divider to bring the voltage down and compare it to a reference with a LM311. Since the current needed is really small I'll use a resistor and a zener as the power supply. No stabilization as discussed previously - an oscilloscope is on it's way from Hong Kong so I can see what is really happening when one charges Lipo with impulses

[Skippic]

In Europe 5 of those caps should be enough for 19.7A average and 47A max, so I'll be able to make two chargers.

I'll publish the blueprints for the charger when it's ready (still waiting for some components).

[IBScootn]

Thanks for sharing your progress and it will be interesting to see the data from your oscilloscope. My GBS 40H cells have a recommended charge current of up to 32A and a max of <120A. I have no data on how the charging current spikes would affect cycle life; but if I wanted to charge from 240Vac at 30A, interpretating from your data, it looks like my current spikes would be ~64A (yikes). Wish I knew whether that is safe.