LM2678-ADJ converted into current source, Looking for Help

We are looking at it right now. We will get back to you soon.

Yang

Hello,

Thank you for including the details of your design and the schematics. Your approach is quite close and we can suggest some simplifications that may help.

In your approach, the current sensing is in the high side lead leaving the supply and entering the battery. This is always a bit of a challenge due to possible limitations in the allowed input common mode range of the amplifier stage connected to the sense resistor. We propose the following approach: (See attached schematic which includes additional discussed in the following text.)

In this approach, it's simplest to connect the power supply rail of the op amp directly across Vin, so selecting an op amp that has a 36V (+/-18V) abs max rating is the first step. Many of such devices are available.

The next step is to create a stage that converts the sensed voltage into a proportional current that is headed toward ground. This requirement stems from the fact that the LM2678-ADJ internal reference is tied to ground. This new current is proportional to load current into the battery and is easily scaled with resistor ratios.

Another problem arises with the choice of the sense resistor value. In your schematics you have chosen 10 milliohms, which on the surface seems okay but at the lowest load current of 50 mA it will only generate 0.5mV of sense voltage. This may be a challenge to amplify accurately with a low cost 36V op amp. Not impossible to do, and 10 millohms will save power dissipation in the resistor.

We suggest a different strategy: Let's first increase the physical size of the sense resistor to 2512, they are slightly more expensive and obviously larger, but carry a 1 watt power dissipation rating. Dissipating the full 1 Watt rating at 2 Amps charging current isn't advisable for long term reliability of the resistor, so let's back off and say that at 2A we dissipate 0.6W. For an I squared R loss of 0.6W at 2A you get 0.15 ohms. This gives you 15 times the sense voltage compared with  the original 10 milliohm selection. The slight added cost of the larger sense resistor will be quickly earned back in relaxed specifications on the op amp; assuming that the increased power dissipation is permitted. In any case, you can re-work this exercise to balance the trade offs.

Now that the sense resistor value has been selected we choose the shown amplifier topology to create a current mirror with output headed toward ground. Resistor values in the mirror are all related by ratio, so there is a lot of flexibility in scaling. Some suggested starting values are shown that should meet your requirements. At this point the stage implements a Constant Current regulator, as you requested. But we need to point out some limitations. The current  sense amp, as shown does not have compliance down to zero volts. For output voltages below around 2 volts, the base emitter junction of the mirror transistor will be incorrectly biased and the stage will malfunction. If you need CC compliance all the way down to 0V output then a different stage will be needed. 0V compliance would be needed if there is ever a chance of a totally discharged (or shorted) battery installed. Some approaches for compliance clear down to 0V utilize op amps a bridge of 4 resistors. Such stages are viable but the tolerance of the resistors must be tightly spec'ed along with the precision of the op amp input offset voltage. We hope that 2V minimum is acceptable for this application.

Since this is a battery charging application there are several follow-on items to address. 

1) What happens when input voltage is disconnected and the battery remains attached? For a conventional buck stage, the battery voltage will back feed through the regulator and be present at Vin. This is generally undesirable and many charge circuits add an additional diode following the power output point to prevent this from occurring. It is unfortunate that the new diode dissipates power and increases the minimum operating voltage for the regulator input, but it is advisable when you consider the number of problems it solves.

2) Most battery chargers should be have not only constant current, but also have a constant voltage limit as well. These are often called CVCC stages. The voltage and current limit modes usually don't occur at the same time. The battery first is charged at constant current, and once it reaches the target stack voltage, it transitions into voltage regulating mode as the battery reaches full capacity. Continuing to charge at full current can damage many cell chemistries if they are forced above rated maximum stack voltage. The attached schematic adds a voltage limit loop along with the diode suggested in the preceding paragraph. Two small signal diodes appear which implement an analog OR into the feedback path; this is a simple way to transition between CV and CC modes. Note that the voltage feedback path is always connected across the battery. Be aware that if input charger voltage is disconnected, there is always a low current discharge down the divider string. 

3) The final concern involves the fact that this is an application that is in the category of being a "fast charger". Any fast charger needs to be reliably shut-down once the battery has reached rated capacity. Continued charging will at the least decrease the battery life by reducing the number of remaining charge and re-charge cycles. It is also possible to have a hazardous situation depending on the behavior of the battery chemistry. There are a number of appropriate charger controllers that can be applied as a control mechanism to this CVCC power stage. From TI most of those ICs use the BQ prefix, and applications questions can be directed through the forum devoted to those products.

7245.LM2678-CVCC.pdf

Keep us posted on your progress.

Alan Martin

Hello,

Thank you for looking at my help request. I look forward to seeing how this works when I receive new LM2678-ADJ within the next 4 days.

To answer your first question. I guess I should of added them in the schematic, but yes you are correct that the output of the charger will need diode protection. For when the charger is unplugged before the battery and/or battery reversed plug in polarity. I will be adding protection diodes on the output to prevent the battery from feeding back into the charger circuitry.

For your second question. I never really looking into CC to CV transition for a Ni-MH battery. I thought this procedure was only used for Lithium batteries. Looks like I'll do some research on this topic.

For your last question. I thought about using a MCU (PIC based) to adjust the output of the charger. The MCU will also be the brain of the whole charging process. Am I current that if those adjustable resistor controlling CC and CV resistor junctions were replaced with digital pots that it would still work?

I will keep you posted on my progress.

Thank you

Lee

Hello Lee,

When used as a divider, digital pots are fine components because the division ratios are well defined by the nature of how they are designed. However, the way I draw them in the schematic I last sent, they are technically connected as rheostats. This doesn't violate anything but the absolute value of the digital pot may be a poorly controlled parameter which will impact the accuracy as you adjust decrease the current or voltage adjust settings. Probably not a problem, just something to be aware of. It's more important that max current and max voltage are well defined and determined by other precision resistor values, not the absolute value of the pots themselves. One aspect of digital pots is they have nonvolatile memory. So when you reconnect input power, the system starts up in its previous analog state. If there is other EEPROM available in your system you can use low cost DACs or even roll-your DACs with discrete resistors and implement the equivalent. It just depends on the accuracy and resolution needed for current and voltage trimming. 

My knowledge of battery chemistry and charge methods is out of date I'm afraid. Your battery vendor is the best point of contact for recommended methods for the specific model selected. Or look at the web resources or other TI forums. 

The voltage limiting loop just seems like good practice even if not absolutely necessary. On the other hand the ever present battery drain when the line source is disconnected is a direct result of this added circuitry.

Correct me if I'm wrong, even with the extra diode included;if you attach a battery backwards on the output, won't that diode still be forward biased? If I'm correct on this, the output of the regulator will be pulled negative, certainly damaging the IC and both diodes and possibly reversing the polarity of Cout.. It won't be pretty. A high current reversed biased diode across Vout plus a series fuse is an unattractive solution to this problem, I dislike fuses that blow, given such a simple error condition. This needs a clever approach, (probably using a MOSFET)  that works regardless of whether there is input power applied or not. It's late in the day and my clever brain cells went home early I'll have to ponder this for a bit.

Alan

Hello Alan,

You have a good point about the digital pots. Would the following solution around the volatile memory work? Using the ON/OFF pin of the LM2678. Pull it high and set the pot values and then low after setup. Would a brownout make this a problem (I believe it would)? You are correct using DACs with discrete resistors will implement the equivalent. The DAC idea will most likely make it simpler and more reliable.

The voltage limiting loop does sound like a good practice. There will also be two sensors a temp sense and voltage sense that will be connected to the MCU. The MCU will monitor the battery voltage and temperature to control the charge cycle. Would the voltage limiting loop and voltage sensor be acting like one an the same?

I must of misted a few details when explaining the diode protection idea I brought up last night. In the attached .Zip file contains Schematic-3 which included everything we talked about. The blue squares on the output represent the two protection circuit brought up thus far. The blue square on the right is the idea I meant to bring up and the blue square on the left is the idea I believe you brought up.

Is the square on the left with the PMOS FET setup the idea you were referring too? (I'm new to using MOS FETs which led me to find the following PDF - http://www.ti.com/lit/an/slva139/slva139.pdf)

Regards,

Lee

3377.Schematic-3_CVCC_TI.zip

Hi,

I have finally built the design that was proposed in a earlier post and it didn't seem to due anything but kill the LM2678 without heating up or showing any signs. This is what i did. I tested the LM2678 in the voltage source configuration to confirm the IC was good. After I added the feedback loop and put a ~8 ohm load to the circuit then turned it on. I looked at my Amp Meter to see if there was a output but there was nothing. After I checked to see if the IC still worked by reverting it back to a voltage source and it didn't function the same. Instead of being adjustable from 1.21V-30V (input voltage is 31.5V). Now its only adjustable form 1.21 - 5V. This same result would happen every time I try different feedback circuits to create a current source. I don't understand why this happens. Any theory's on this problem?

Regards