Paralleling Diodes with Resistors

Summary

This blog demonstrates how diodes can be connected in parallel with a suitable series resistance per diode. These current sharing resistors are required for each paralleled diode primarily to handle the negative temperature coefficient (NTC) effect on diode forward voltage.

Using resistors with multiple diodes may be required for a number of reasons such as design costs, package size, availability of required diode(s) or something as novel as limited space on a printed circuit board (PCB).

The method using current sharing between diodes is particularly useful for designs where space is a premium and adding a thermal mass between diodes is not possible.

Literature
There is documentation relating to paralleling diodes although it can be difficult to locate online. ST Micro published an Application Note AN599: Parallel Operation of Power Rectifiers in 1993 and Application Note AN4381: Current Sharing in Parallel Diodes.

NXP also have a similar Application Note AN11358: Parallel Schottkys as secondary rectifiers in flyback adapter.

Diode Matching
Diodes bearing the same part number are made with similar characteristics however devices are never identical. Slight variations in diode characteristics such as forward voltage 'Vf', during parallel diode operation, can result in one diode sharing more current than others.

Diode matching, similar to LED binning, can be performed through a process of measurements targeting diode forward voltage drop for various currents (VI curves). Attention should also be made to the diode temperature. The solution of matching diodes may be suitable for low volume specific designs or product prototyping however this is not suited for volume (mass) production.

Diode Testing - No resistance
To illustrate how one diode, in a two diode parallel setup, will eventually end up at a higher temperature than the other a bench test was conducted.

Dual Schottky diodes in parallel

Two Schottky diodes (ON Semiconductor parts) were soldered in parallel. A 1.8R 5W resistor was used to limit the power supply current. 

Temperature of one Schottky Diodes when operated in parallel

Temperature measurements were performed at regular intervals using a FLIR C2 thermal camera and the power supply used for testing was a Rigol DP832.

The power supply voltage was under 2.5VDC. The total power supply power was limited to 1.3W.

Temperature Measurements: Schottky Diodes no resistor 

As shown in the table above, one of the diodes began to heat more rapidly in under thirty seconds. After four minutes the test was not continued because the diode junction temperature Tj was approaching the devices specified rating of 125C.

Diode Testing - 0.01R resistance

For the second test the two diode setup was modified to include a 0.01R resistor on the cathode of each of the diodes. This resistor is the standard type used for current sensing.

These resistors were added to test if a small resistance, such as a PCB traces or the resistance in leaded components, would balance the current between the diodes.

Dual Schottky diodes in parallel with serial 0.01R resistors

The power supply power was again limited to 1.3W and temperature measurements conducted.

Temperature Measurements: Schottky Diodes 0.01R resistor

The results above show that the diodes began to heat at a more consistent rate compared to no series resistance until five minutes. After six minutes one of the diodes had begun to heat more than the other and was drawing most of the current.

This test indicated that such as low resistance was insignificant and unable to balance the drop in diode forward voltage caused by the heating of the diode.

Diode Testing - 1.3R resistance

The two diode setup with resistors was modified to use a 1.3R resistor (0805) on the cathode of each of the diodes. These resistors values were guided by the formula shown below.

Measuring each of the diodes forward voltage Vf with a digital multimeter, diode 1 read 0.64V and diode 2 read 0.65V. With approximately 250mA being equally passed by each diode the series resistance calculates to 2.6R, this value was halved to reduce the resistor power dissipation and tested.

Dual Schottky diodes in parallel with serial 1R3 resistors

Resistors with the case size 0805 were fitted as shown in the image above.

Temperature Measurements: Schottky Diodes 1R3 resistor

The test results showed the diode temperature stabilized after two minutes.

The above results however were a brief snapshot of a lengthy soak test which did show that some variation in diode temperature did eventuate. The difference in the temperature of the two diodes was stable and no runaway was seen during testing.

Temperature of one Schottky diodes when operated in parallel with series resistance

When viewed thermally the balance between the diodes can easily been seen when compared to the test where no series resistance was fitted.

Current Sharing - 1.3R resistance

Using a multimeter the forward voltage across the resistors was measured after 10 minutes. One resistor showed a voltage drop of 0.374V and the send 0.363V. 

Using the voltage across the resistors to determine currents showed that 0.289A and 0.279A flowed through each resistor diode legs.

It should be noted that the current through each resistor diode leg is dynamic and should never be considered exact. Additionally the current through each leg should not be thought of as balanced. When the third test was allowed to run for a considerable time a temperature delta of 10 degrees was often seen between the two diodes.

Summary
For the tests conducted in this blog, with a pair of ON Semiconductor Schottky diodes, adding a suitable value series resistor to each of the diodes facilitated current sharing, without the need for thermal bonding.

Fitting the two diodes to a common thermal mass or thermally coupling the diodes may also be another solution to balancing the current sharing although not investigated in the scope of this blog.

Additionally the tests employed a benchtop DC power supply as the source. Other supplies such as AC or pulsing supplies (PWM) with shaped waveforms and parallel diodes may require more specific testing to determine operational characteristics for a parallel diode design.

Design Suggestions
Listed below are some design suggestions when using diodes in parallel.

1. Choose diodes with the lowest forward voltage (Vf) that can be afforded for the design. This aids in reducing heat dissipated which is especially a concern for enclosed spaces.

A single bigger diode is not always a better solution if it is dissipating significantly more heat.


2. For the current and voltage rating of the diode, add an appropriate safety margin suitable for the design.


3. A design with parallel diodes operating with continuous DC current may experience less stress than parallel diodes operating as freewheeling (flyback) diodes for inductive loads and therefore should be rated accordingly.


4. For the current sharing resistor select an appropriately rated device for power dissipation. When designs use PWM at high currents the resistors may also benefit from being rated for pulse withstanding.


5. Test, test and test. Testing does not refer to the product operational or consumer specifications but the working limits of the design. To understand your product it should be tested to failure!

Noctua 120mm fan testing (NF-F12 PWM) 7 x PWM using Teensy 2.0

Summary

This blog covers the replacement of a number of Cooler Master fans with Noctua PWM fans. Both fans were 120mm in size, to fit standard PC cases. Some measurements relating to PWM and current were taken as a comparison. PWM control was performed using a Teensy 2.0 programmed using Arduino.

Cooler Master Fan

Fan Replacement
The Cooler Master fans (A12025-12CB-3EN-F1) used in a tower of RAID drives, mentioned in this blog, were lacking the airflow to keep the temperature of the drives below 40 degrees Celsius. Noctua fans (NF-F12 PWM) were trialled with PWM control. 

It should be noted that the Cooler Master fans are rated at 2000RPM compared to the Noctua 1500RPM.

Noctua NF-F12 PWM

Fan PWM Control
Since PWM fan control from the existing motherboard was not possible, a Teensy 2.0 board was utilised. This board has seven PWM outputs and libraries for Arduino were already written. Image courtesy of PJRC.

Teensy 2.0 Pinouts

When using the PWM outputs on the Teensy there was a difference in the frequency between some of the PWM output frequencies due to the timers. This turned out to be of no consequence as the Noctua fans worked across a large range of PWM frequencies. Even with the PWM frequency set in the low Kilohertz no clicking or other audible noises were heard from the Noctua fan.

Teensy 2.0 with Noctua NF-F12

For the fan connections, pinouts were referenced from the site All Pinouts. Power was applied to the required two relevant power pins. A common 0VDC connected was required between the fan and the GND pin on the Teensy. 

Teensy 2.0 Close Up

The image above has only PWM1 connected. A series 2.2k resistor for current limiting was connected between the PWM output of the Teensy to the PWM input of the fan.

Arduino Program
To control the duty for each PWM output, a simple program was written using Arduino. The code allowed each output to be individually changed. Only basic error checking was performed in the software, something to be aware of.

From a terminal program the number of the PWM output (1 to 7) followed by the duty cycle in percent (PWM ON time) was entered with a new line (enter key) to process the change. For example:

     Set PWM 1 to 75%: 175

     Set PWM 6 to 10%: 610

     Set PWM 7 to 100%: 7100

Example of working code below.

int PWM1 = 4;    /* PWM Output 1 */
int PWM2 = 5;    /* PWM Output 2 */
int PWM3 = 9;    /* PWM Output 3 */
int PWM4 = 10;   /* PWM Output 4 */
int PWM5 = 12;   /* PWM Output 5 */
int PWM6 = 14;   /* PWM Output 6 */
int PWM7 = 15;   /* PWM Output 7 */
int PWM = 0;
int PWM_Serial = 0;
int percent = 0;
char buffer[] = {' ',' ',' ',' '};

/* Start fans at full power then run down to approx 10% */
void setup() {
      Serial.begin(115200);
      pinMode(PWM1, OUTPUT);
      pinMode(PWM2, OUTPUT);
      pinMode(PWM3, OUTPUT);
      pinMode(PWM4, OUTPUT);
      pinMode(PWM5, OUTPUT);
      pinMode(PWM6, OUTPUT);
      pinMode(PWM7, OUTPUT);
      analogWrite(PWM1, 255);
      analogWrite(PWM2, 255);
      analogWrite(PWM3, 255);
      analogWrite(PWM4, 255);
      analogWrite(PWM5, 255);
      analogWrite(PWM6, 255);
      analogWrite(PWM7, 255);
      delay(2000);
      analogWrite(PWM1, 26);
      analogWrite(PWM2, 26);
      analogWrite(PWM3, 26);
      analogWrite(PWM4, 26);
      analogWrite(PWM5, 26);
      analogWrite(PWM6, 26);
      analogWrite(PWM7, 26);
}


/* Main to parse serial commands, no error handling */
void loop() {
  if (Serial.available() >0) {
    PWM_Serial = Serial.read() - '0';
    switch (PWM_Serial) {
      case 1: PWM = PWM1; break;
      case 2: PWM = PWM2; break;
      case 3: PWM = PWM3; break;
      case 4: PWM = PWM4; break;
      case 5: PWM = PWM5; break;
      case 6: PWM = PWM6; break;
      case 7: PWM = PWM7; break;
      default: break;
    }
    while (!Serial.available());
    Serial.readBytesUntil('n', buffer, 4);
    percent = ((atoi(buffer)) * 255)/100;
    analogWrite(PWM, percent);
  }
}

The terminal application TeraTerm, was used to send the commands to the Teensy. As there was no feedback provided by the Arduino code, the actual PWM signal was monitored using an oscilloscope. The Tacho output was also connected to the oscilloscope.

Power supply current and oscilloscope frequency measurements were taken for the duty cycle range of 10% to 100%.

Measurements
To determine what current (mA) the Noctua fans would consume, the duty cycle of the PWM was varied with a single fan connected and the current measured using a multimeter.

Noctua Current vs Duty Cycle

Undoubtedly the Noctua fans Tacho output against duty cycle, as seen below, has been recorded before however for completeness this was also measured.

Noctua Tacho Output vs Duty Cycle

Measurements were performed with the PWM duty starting at 10%, as the fan stall speed occurred at a duty between 7-10%.

Cable Assemblies
As a side note, Noctua provide a number of adapter cables with the fan. These range from an extension cable, Y adapter and a Low Noise adapter. Regarding the low noise adapter the Noctua cable features what appears to be a single 82R resistor, meaning that low noise is simply a by-product of limiting the total current to the fan. The same result could be implemented by adjusting the total power to the fan.

Noctua Low Noise Adaptor

Adding the low noise adapter in series with the power supply to the Noctua fan, reduces the fan speed. The current drawn by the fan is reduced to around 40mA.

Noctua Low Noise Adapter Unsheathed

Operational Temperature

In relation to the heat generated by the fan itself, a thermal camera was used to measure the Noctua fan motor which was running at full speed in free air. After several minutes, with an ambient temperature of 24C, the temperature of the fan motor was approximately 28C.

Noctua Fan Temperature

Testing

For testing the effectiveness of the Noctua fans, the RAID system inside a server was kept running using Caffeine on Linux Mint Cinnamon. The system (RAID) did not spin the drives down.

Caffeine Package

The server was left powered and running the application Caffeine for 60min to stabilise the internal temperature without fans. The temperature of the drives was taken from the Smart data available under the Disks app. The highest drive temperature was 41C.

No Cooling RAID Temperature

A single the Cooler Master fan was tested first. After 60min the temperature of the warmest drive was 37C.

Single Cooler Master Fan - RAID Temperature

For the second test the original fan was removed and the RAID was allowed to reach 41C again. A single Noctua fan was then tested. After 60min the temperature of the warmest drive was 38C. 

Single Noctua Fan - RAID Temperature

For the next test the Noctua fan was stopped and RAID was allowed to reach 41C again. Dual Noctua fans were tested. After 60min the temperature of the warmest drive was 32C.

Twin Noctua Fans - RAID Temperature

Three fans were tried although no difference in RAID temperature was noted when compared to two Noctua fans.

Final Thoughts

There is little doubt the Noctua NF-F12PWM fan performs extremely well, almost as well as a fan running 15% faster, for the limited tests conducted above. It shall be noted however that at the time of writing this blog, the Noctua fan was four times more expensive than a Cooler Master fan or a nearest equivalent non PWM fan. This cost may be prohibitive to some looking to move to PWM. 

Additionally if large airflow is all that is required then a fast, reliable, non-PWM fan may be the more cost effective solution. Mass air flow fans manufactured by Delta or Orion may be more suitable for some installations. Orions models such as the OD1238-12HBXJ10 have an airflow of 250CFM compared to the Noctua's 56CFM however the Orion fan has an ear drumming 69dBA, compared to the Noctua almost inaudible acoustics at 23dBA.

For alternate installations such as a media centre, bedroom computer, home cinema or similar installation were low noise is paramount, then a PWM fan would be an ideal solution.