Electronic DC Load

Project 31 : Electronic DC Load Part 1

This is the first of a series of articles on the design, construction and operation of an analog Electronic Load for DC (click on any picture for full size image)

Exhibit 1 – A Precision Electronic Load

At the heart of this project is an implementation of the video presented by Dave Jones under “EEVblog #102 - DIY Constant Current Dummy Load for Power Supply and Battery Testing “, back in the day when the EEVblog was really about electronics design.

Dave's excellent analysis, design and description results in a fully functional tool made essentially from junk-box parts. In this series we present a discussion about performance, trade-offs and options, then extend and complete the construction.

In this the first article, we will build the schematic on a breadboard, and bench test the performance. At the end of this article you have should more than enough information to put together one of these extremely precise and useful tools.

In subsequent articles we will extend the functionality, talk through a suitable printed circuit board, and share our approach to building this out to become a valuable bench test tool.
We hope you enjoy this series, please feel free to add comments and questions, and if you click any image in the article it will download to full size ( typically 1024x683)

In Exhibit 2 below, we have translated the whiteboard DaveCad version from Dave's video, and made a few updates to fit with our requirements and availability of parts. At the outset if you have not already watched Dave's video we'd commend you to do just that, as in this article our explanations will go more into the differences we have chosen and rather less about how the Electronic Load actually works.
EEVblog 102 is located here

That said, a rudimentary description of how the Electronic Load works is surely also required.

So here goes...

The Electronic Load measures a current through precision resistors, and buffers that via and OpAmp into a MOSFET to create a constant current load. The load is adjustable by applying a finely tuned reference voltage to the other side of the OpAmp. The voltage across the load resistor is fedback to the operators by a 0-199mV panel meter.

This load is continuously variable  from 0 mA to 2000 mA. This is done using a 10 turn wire wound potentiometer connected to P3 and P4 in our circuit (the wiper of the pot connected to pin 1 of P3. This potentiometer sets a reference voltage at the input the first OpAmp which is configured as a voltage follower. The voltage at pin 3 for the LM324 is also reflected at the output on pin 1. In turn this is applied to the input of the second stage, noting that the second voltage will do everything possible to make the inputs at pin 5 & 6 equal. Hence the second stage drives the output MOSFET which is connected across the supply under test to pass a current through itself, and a precision 1 ohm resistor (10 of 10 ohms in parallel).
The voltage developed across this resistor is reflected on pin 6 of the second stage OpAmp and hence will very accurately drive the current in the load. This voltage is also monitored by the digital voltmeter, which is used to set the current at the load point (connected at P5 and P6).

You may note in this circuit we are using two 2 terminal blocks instead of 3 or 4 terminal blocks, but that is what we have, so hence that's what we used.
There is a resistor R2, which takes the voltage from the load into a resistor divider so that the 0 to 2000mV developed across the precision load, can be sub divided to 0 to 200mV to match the full scale deviation on the panel meter.

Exhibit 2 – The modified EEVBLOG DC load Rev.A circuit

Design Changes

The Panel Meter : In Dave Jones's original design he utilised the CX-101 panel meter, we found them difficult to source, but substituted the GDD5135A which is cheap and readily available. With careful tuning this little meter performs admirably against our Fluke mutlimeters for the task. The wiring on theGDD5135A  meter is somewhat simpler than Dave's original design.

The MOSFET : Dave's design specified using an MTP 3055, which has more than adequate performance for this project. Unfortunately we did not have any to hand, so substituted for a STP50N06 (rated at 50V and 60Amp), also more than enough performance . You could use any similarly rated switching MOSFET.

The Supply Voltage : In Dave's original design, he chose 5V as the supply rail, and was able to comfortably yield 1.3 Amp of load. Dave talks about how by lifting the supply rail you could also lift the voltage swing across the precision resistor, and hence the current presented to the load. With some experimentation we have selected 12v as our supply voltage, which gives us 0-2,000 mA. However this selection is not a simple trade off, as you raise the current many factors quickly come into play which affect your ability to manage the design.

We will discuss these factors in detail later in this article, but in the end we settled for 0-2,000 mA because that was all our panel meter could display.

The Breadboard Layout

Building this project on the bench first as a breadboard project has proved  a very smart move. Whilst this is an inherently simple design/project there are numerous traps, and design trade-offs to be made along the way to getting to be usable test instrument.

Exhibit 3 breadboard wiring for the LM324

Exhibits 3 and 4 show the layout we used, note the use of screw connectors for the components that by necessity cannot be on the breadboard. Eg, the panel meter, and the 10 turn potentiometer and the supply under test

Exhibit 4 – using 8A screw connectors on the breadboard

Exhibit 5 shows the heat sink we chose to go with to distribute the excess heat in the MOSFET. In our case we took a salvaged PC CPU heat sink from a 1.2GHz CPU, cut it in half, and drilled and tapped some mounting holes. The choice of heat sink will become an important part of how this project comes together. Some form of heat sink is mandatory, but bigger is not always better as we came to understand during the testing phase.

Exhibit 5 – Power MOSFET STP50N06

The precision resistors, Exhibit 6 shows how we have them connected for the bench test phase of this project. We chose to go the way Dave originally proposed and use 10 resistors of 10 ohms each at 0.1%. we think these resistors are rated at 0.5 watts. It has proved a valuable point to make sure the resistors are mounted well off the board, at 2 Amps, a total of 4 watts is dissipated in the resistors, and they become quite hot to touch.

The final design will definitely take this into account, and this is one of the design decisions about the usable range. For example if your tried to configure this load the carry 10 Amps, the resistors would have to dissipate 100 watts a significant amount of heat, and well out of practical limits.

Exhibit 6 – Using 0.1% 10 precision resistors

The GDD5135A panel (exhibit 7) below is a 3.5 digit meter capable of 0-199mV. The meter occupies a small footprint, has large digits and is configured to operate in a common ground configuration. All these are important design criteria. It also operates from 3v to 16v which is well within our target range of 12v, and draws typically only 20 mA in operation.

Exhibit 7 – using the GDD5135A 200mV meter

The photo below (Exhibit 8) shows the whole set up in one frame, the ten turn potentiometer on the left , the 3.5 digit voltmeter, the precision resistors front right, and the MOSFET and heat sink back right. Of note also is the meter reading 100% correlation with the Fluke multimeter.

Exhibit 8 all components wired together

To complete the test set up we are using a Rigol DP832 power supply to (a) power the Electronic Load (channel 1), and (b) to provide the power supply under test (channel 2). the third channel is not used for these tests. Exhibit 9.

Exhibit 9 Power supply and test set up

Testing

Test 1 : In Exhibit 10 , our first shot under test you can see the following : Mutimeter on the left looking at the supply voltage, multimeter in the middle looking at the numbers of mV developed across the precision resistors, and the multimeter on the right, looking at the temperature of the MOSFET heat sink. So in this example at 4.0V, drawing a current of 1.99 Amps the temperature of the heat sink is 40.3 degreesC. Much of the data that follows uses this test setup.

Exhibit 10 – our test set up

Test 2 : 8V at 2 amps

In Test 2 our setup we increase the voltage to 8V, and keep the current the same, the heat sink temperature increases to 47.8 degreesC after a couple of minutes.

Exhibit 11 – now at 8 volts

Test 3 : 12 Volts and 2 Amps
In Test 3 we increase the voltage to 12.0V, and keep the current at 2 Amps. The load is now dissipating 24 watts if input power (Exhibit 12)

Exhibit 12 – testing at 12 Volts

At 24 watts the temperature of the heat sink is increasing rapidly now at 60.3 degreesC after a few minutes.

Exhibit 13 – 12 volts – 24 watts – 60degreesC

Test 4 :18 volts and 2 Amps, (36 Watts of input power), everthing is going along fine (exhibits 14 and 15). We had noticed though that the temperature was rising sharply (86.5 DegreesC)

Exhibit 14 - 2 Amps at 18 volts – 36 watts

We started to observe some interesting characteristics of this load configuration.

Exhibit 15 – 36 Watts – 86.5 DegreesC

Long story short, the temperature was rising very fast, the heat sink was at 125.7 degreesC (Exhibit 16). Actually there was nothing to suggest that the maxim temperature rating of the MOSFET would NOT likely be exceeded in the next few moments. Especially as we had typically measured +10-15 degreesC higher than the heat sink at the case of the MOSFET, AND there there is a further thermal resistance from the case to the junction. The junction of the MOSFET could easily have been at 155 degreesC.

Exhibit 16 – 36 watts temperature 125 degreesC

And interestingly the anti static mat under the heat sink started to let out a particularly smelly smoke (Exhibit 17) Note the brown scorch marks matching the heat sink. At 125 degreesC if you touch the heat sink with a wet finger, you will get a pleasing sizzling sensation as your finger starts to cook.

Exhibit 17 - Anti-static Pad starts to smoke

Design Observations

When the voltage is low and the current is high (E.g 4V at 2 Amps), 50% of the power is being dissipated in the precision resistors. This is important, because precision resistors are expensive, and at 4 watts across 10 resistors, we are approaching the rated maximum power in the resistor of 0.5 watts. They are getting very hot !. The implication here is this setup is not going to be suitable for currents above 2 Amps. Which starts to look like a design constraint.

Secondly if the voltage is high (say 24V) well within the rated 50 V for the MOSFET, and the current is high (2 Amps), well within the rated current of the MOSFET (60 Amps), the power into the load is 48 watts, of which 4 watts is dissipated in the precision resistors (as above), BUT 44 watts is dissipated in the MOSFET well about the rated 40 watts of the device.

In our case the MOSFET let go the magic smoke, and shorted between Gate and Source … all very exciting, but not very useful.

These two parameters are important considerations in this design, as is the size, and suitability of the heat sink, which we will examine in some detail below

Exhibit 18 – dialed off the scale

Just a note before we get into the temperature analysis, if you dial a current of 2 Amps or more the Digital Panel meter will overflow, per Exhibit 18, and per Exhibit 19 below you can get a very high correlation between the Digital Panel Meter and The Fluke (Exhibit 20)

Exhibit 19 – A precision tool

Temperature Analysis

From our tests we could see that for this load to be effective it will need to be operated within some predefined parameters. We have already seen, two scenarios where we can smoke the precision resistors through too much current or smoke the MOSFET through too much power.
It is also a bit more complicated,  during what you might call normal operation. The question came to us about what would be an acceptable operating temperature for the DC load.

We could tell from experience, that at 36 watts input (18 Volts and 2 Amps), that the heat sink becomes very hot (125.7 degreesC), if left to operate unchecked.

Given the temperature was still rising at this point we could see how we might even exceed the rated junction temperature for the MOSFET (175 degreesC), especially in our tests the case of the MOSFET was typically 10-15 degrees hotter than the heat sink.

To better understand what was happening we chose a few static loads, then measured the temperature performance of our heat sink and test setup. The first static load was 10 watts input into the load, and you can see from Exhibit 20 that the temperature was still rising (albeit slowly) after 30 minutes (82 degreesC) . It was interesting to note, though that the rate of rise was slowing. We found this intriguing, especially as the power into the heat sink was constant

Exhibit 20 – Testing at 10 watts

We then chose another test case of 36 Watts (18V and 2 Amps). As can be seen from Exhibit 21 the rate of temperature rise is much faster , from ambient, (approx 20 degreesC) to over 94 degreesC in 5 minutes.

Interestingly though the rate of change was also slowing in this example. We concluded that even though the rate was slowing the set up was unlikely to sustain a power input to the load of 36 watts.
We were however intrigued by this rate of change that we observed in every test. Our theory was that as the gap to ambient became larger and larger, the heat loss the the air became faster and faster.

Exhibit 21 testing at 35 watts

Testing Heat Loss

To test our hypothesis, we heated the heat sink with 36 watts to a temperature of 132 degreesC. (refer to Exhibit 22). Then we removed the power, and monitored the temperature each minute for the next 20 minutes.

Exhibit 22 - Watching the heat sink cool down

Interestingly the temperature dropped as predicted, with a slowing rate as the temperature of the heat sink became closer to ambient (Exhibit 23)
To put that empirically, at 132 degreesC, the heat sink is losing heat more than 7 times faster than it is at 25 degreesC.
So at first glance it might seem better to run the DC load at a higher temperature as the faster heat loss is helping us to keep the size and cost down. However we are also shortening the life of the semiconductors and limiting the physical configurations we can use. If for example we ran the heat sink at 125 degreesC we would need at least 100mm clearance in all directions (somewhat impractical).

Exhibit 23 – cooling action

At this point in the process we have learned a great deal, we have found some viable scenarios where this design could be come a useful test instrument. We have also learned there are a few inherent limitations

The design appears an excellent solution for low power circuits, is highly precise, simple and very cost effective.

However there are a set of parameters which will determine the use and effectiveness of this design for a Electronic DC Load if we try to scale the solution, they relate to maximum current in the precision resistors, the maximum power in the MOSFET, and the controllable maximum temperature in the heat sink.

Exhibit 24 tries to capture some key elements of this discussion, notably the 2 Amp limit in the precision resistors, and the maximum power of 40 watts in the MOSFET. It is important that we avoid stressing the MOSFET unduly, as reliability and lifespan are massively shortened the closer we operate semiconductors to maximum tolerances.For these reasons, we are arbitrarily setting some design parameters :

• Maximum current through the load : 2Amps


• Maximum power in the MOSFET : 30 watts ( data sheet absolute max : 40 watts)


• Maximum heat sink operating temperature of 100 degreesC steady state

Exhibit 24 – Power rating table

Exhibit 24 shows some important factors to consider. Chart displays the power (voltage multiplied by the current) in the DC load.

• Black text on a white background show possible settings to get 10 watts or less in the load.


• White text on the blue background show settings to get from 10 watts to 20 watts in the load


• White text on the Orange background show settings to get 20 to 30 watts in the load


• White text on a Red background are greater than 30 watts.

Our tests so far suggest 10 watts is a viable configuration as presented (Exhibit 20), after 30 minutes at a temperature of 82 degreesC, it meets all of our criteria above. In addition we get a full range of voltage and current
In our 20 watt test ( not shown in these pages), the heat sink temperature reached 96.5 degreesC suggesting it is at the upper limit of our criteria. From the table above 10 volts is our practical maximum.
At 35 watts (Exhibit 21), the heat sink reached 94.5 degreesC in under 5 minutes, suggesting it is well outside our parameters, and from the table above 16 volts is a practical maximum, but this option fails the temperature criteria
So the opportunity exists to see if we can extend the range of the load from 10 watts and built as far out towards 30 watts as is possible. This is the objective of our next article !

Prologue

In the second article in this series we will examine options to extend the 10 watt range by changing the heat sink and cooling arrangement , and make some further suggestions on how to improve the viability of this DC Load as an effective test instrument.
As a precursor though, we will not be radically altering the design, the use of precision opamps, and resistors is such an attractive design, rather we will be enhancing the performance and reliability
We will also build a circuit board and propose a suitable housing. In addition we will make some comments on the power supply options you might use for this device

The Rev A circuit diagram is attached below. If you have already built one of these DC loads we'd be keen to include your experiences in the discussion, please leave us a note, or some feedback.

All the best ...Peter

Schematic for the Rev A circuit is located here