Driving IR2110 with SG3525 for a synchronous buck converter

Keep in mind that the normal CMOS gates have about 120R of internal resistance at its outputs. And may not be simmetrical. For example, 200R source and 100R sink. N channel MOSFET has usually lower resistance. Note that it's an example only and may or not be your case. Read carefully the datasheet to stay sure.

Over it, take a minute to understand this: if you pay attention, the buck and the boost topologies are the same with interchanging between input and output. Thus if you have a large output capacitor and you switch off the power supply from primary side, momentarily it may actuate as boost from output driving voltage into the primary with no voltage control. Thus primary side may be overvoltaged during few instants. This effect is more notorious when syncronous rectification is used. The effect is in some way, similar to bus boost in half bridge class D amps.
 
Interesting facts, thanks!

In my case Logic gates internal resistance is not a problem, because the inputs of IR2110 have extremely high impedance.

Some mosfet drivers are also not simmetrical in source / sink, like IR2184.

IR2110 is simmetrical though, 2A / 2A.

I am planning to use another one of 4.5A / 4.5A for even higher efficency.

To calculate the gate resistances for the mosfets, of course, i'm also taking into account their gate resistance.
 
Yes, i know...

Everything has to be designed carefully and all details have to be taken into account.

I tested the power part and it works very very well. Now i just need to implement the modification with the XOR gates.

I will find a way to get my hands on an oscilloscope, because it's a must in electronic design.

Based on some calculations the MOSFETs take less than 50nS to turn on / off. Thats why the switching losses are so low. With an oscilloscope, i could have measured this time more precisely.

So, for now, everything is fine in my circuit except the output voltage regulation due to the max 50% duty cycle problem. But now i'm about to solve it with those XOR gates, hopefully.
 
Yep, that is actually what i do while i repair mains power supplies. There is a fuse plus a 100w Light bulb in series.

In this case its not a main power supply. Its just a synchronous buck converter with a maximum DC input voltage of 80V.

But of course i still need to be careful.
The circuit is fused, and the power supply i'm using to power the synchronous buck converter has an adjustable current limit.

Heck no i'm gonna connect prototype circuits directly to batteries!

Even if they are fused. If the current limiting feature comes into play, i wont burn the components.

Only after testing the circuit in every regard i will then be able to connect it directly to batteries. Of course ALWAYS still fused.

Never forget the fuse, or...
Short circuit will be massive.
 
So in the end will finally get the output complementary waveforms that will look like this (see image) (ignore SD) and with a duty cycle that can go up to ~100%, right?

With 2110 (or any bootstrapped IC), a 100% duty (permanently on upper device) is not possible, as this would drain the bootstrap capacitor providing the VBS power supply, as shown in the datasheets. It is not recommended to go beyond 90% in any case, especially with high gate-charge MOSFETs.
 
Yes, i knew 100% is impossibile. I will limit it to 95 - 90%.

I'm using a polypropylene capacitor for bootstrapping. As far as i know, they are the best for high frequency operation. They have an extraordinarily low dissipation factor. They are even better than ceramic, i think.

Of course, bypass capacitors are in every IC, literally as close as phisically possible. They are multiplayer ceramic capacitors (SMD).

The bulk electrolithyc capacitors are 820uF 100V x3 on both input and output (Good capacitors by nippon chemi-con)

All those ensure reliable operation.

The design seems very good. Now, i am about to make the modification with the XOR gates.

I also found that the CD4070 has symmetrical source / sink currents, just in case this makes the design even better. But i dont think it matters in this case, because the inputs of the mosfet driver are very high resistance / impedance.

Since i'm making the circuit on a prototype board, i'm using DIP package ICs.

In the new PCB design they are all going to be SMD, which will improve the circuit even more.

I also noticed that i didnt add diodes in parallel of the gate resistors, so i'll add them now. It should make switching even faster.

Exsternal shottky diodes are also in parallel to the MOSFETs internal body diodes to improve efficency.
 
Its not over, i realized that the CD4070 i had is partially broken. I actually pulled it off from a broken circuit borad (i dont remember what it was). And so the only thing i could do is to order it. And i'm waiting a lot for it to arrive. Around 1 week of time remaining.

In the mean time i conducted some other tests with a higher input voltage (40+V) and i noticed that the inductor gets very hot even with a low output current (The output voltage is still around half of Vin.) ONLY the inductor. Maybe i have to use a higher frequency? The inductor is a Iron powder core type the switching frequency is 50khz. Will raise the frequency improve this? I noticed that the CORE loss is high. The core itself gets very hot. Vasta majority of the heat is coming from the core loss.

Should i raise the frequency? (100 - 200khz)
 
The filter LC depends upon the frequency of the corner of the filter. Also on kind of filter. But as voltage is raised, closed magnetic circuits become saturated easily. So you will need a gapped ferrite core. In my last project I used drum cores recycled from PC CRT monitors (about 50 to 70KHz) and didn't run hot. But a relative large turns are needed to lower the Bmax on the core.

Read some articles from Lloyd Dixon at Texas Instruments about choke designings.
 
I'm going to search for those gapped cores and see what i will be able to find. For now i will just raise the frequency. As far as i know, with a higher frequency magnetic saturation can be lowered / removed. I tried with bigger inductance (100uH) but it still gets hot (less than lower inductances though). I will raise the frequency to 150kHz. Based on calculations, the IR2110 wont have any problems driving 17nF of gate capacitance at 150 - 200kHz.

I'll make this modification and see if the inductor heats up less.

Everything else in the power part of the circuit is highly optimized, there are also zener diodes on the gate of the MOSFETs to protect them from voltage spikes.

The gate drive IC is positioned INCREDIBLY close to the MOSFETs.

Added a PNP transistor + 1 ohm resistors on the gates as well to discharge them even faster than the driver IC could.

The switching efficency got so high that i can draw 10A from the output of the converter and leave the transistors without heatsink! I never expected this but hey, here i am.

It always seems impossible until its done. Then it still seems impossible but people are really impressed with you.
 
Last edited:
with a higher input voltage (40+V) and i noticed that the inductor gets very hot even with a low output current (The output voltage is still around half of Vin.) ONLY the inductor. Maybe i have to use a higher frequency? The inductor is a Iron powder core type the switching frequency is 50khz.
An iron powder core is the correct type for a 50kHz switching frequency. However, there're different grades for different applications and frequencies. You'd be alright with the yellow / white core from a scrap ATX PSU.

There should be little heat even under full load. If not, raise the switching frequency to cut the peak current in half.
But as voltage is raised, closed magnetic circuits become saturated easily. So you will need a gapped ferrite core.
Iron powder cores are internally gapped (within the material mix) and are therefore not closed magnetic circuits. Maybe the OP has the wrong grade .

In my last project I used drum cores recycled from PC CRT monitors (about 50 to 70KHz) and didn't run hot. But a relative large turns are needed to lower the Bmax on the core.

Drum cores are indeed open but they'd make a lot of EMI, especially at heavy loads.
 
Well, i actually used a white yellow core. Its a 100uH 10A rated inductor and i pulled it off from an induction heater board. It still gets hot! At 5A the temperature reaches like 60 °C and at 15A man it goes to like 150 °C... Maybe shootthrough? I will raise the frequency and also slightly raise the deadtime and see if it gets better. The SG3525 is actually set to a frequency of 100kHz but as far as i know, since i'm using a half bridge, the total frequency on the inductor is half, so 50kHz. I also tried a completely black core (its not ferrite) inductor (10uH) and it was getting super hot even at 3A. Basically only the core was getting super hot. Much better with the yellow white one but still it seems to dissipate too much.

Also, the output voltage being half of the input voltage... Is this the worst case scenario for the inductor? I May be wrong here.

And i also learned that DC-DC converters inductor values have to be calculate based in input voltage, output voltage and output current. And by running a lot of calculations i noticed that when the output voltage is half of the input voltage, the calculated inductor results in the highest value. If i deviate on MORE than half of LESS than half, the calculated inductor becomes equally lower. You will get something like a bell shaped curve.

And if i'm not wrong this means that adjustable output DC-DC converter have the disadvantage to use a bigger inductor for lower or higher output voltages in respect to the input voltage, when in those cases the inductance should be lower. Is this an actual drawback or it does not affect the DC-DC converter much?

The biggest inductance is needed for when the output voltage is half of input. But just for this scenario.

Basically if i set the output voltage of the converter much higher or much lower than half of Vin, the converter runs with a bigger inductor than it should be at that output voltage. It can go as far as triple / quintuple the value.

There is also a current limiting feature implemented with a EMF / EMI shielded hall effect current sensor.

So that even if i short the output nothing bad happens, but the voltage will drop SUPER low. Like the input is 100V and the output is less than 1V. The inductance should be much much lower in this scenario.

But adjustable high power current limited buck - buck boost or boost converters are all over the place and they seem to work fine on every scenario... I guess there is a limit in the ranges of voltages that a DC-DC converter can be used with.

Is this right or i messed up?
 
Also, the output voltage being half of the input voltage... Is this the worst case scenario for the inductor?

Yes, peak current = Ton * (Vin - Vout) / L . The max Ton value is at 50% since Ton +Toff = constant. With higher L, you'd get lower peak current, but if this current is still saturation capable, then it will get hot.

Well, i actually used a white yellow core. Its a 100uH 10A rated inductor and i pulled it off from an induction heater board. It still gets hot! At 5A the temperature reaches like 60 °C and at 15A man it goes to like 150 °C... Maybe shootthrough?

There are yellow-white cores of different sizes, maybe you've the wrong size for your application. The material is of the soft saturating type (pg.25). The yellow white material is called -26 & is made by many companies. See attachment for details.

Shoot-through current heats up only the devices as it does not go through the inductor.
 

Attachments

  • PC_L.pdf
    6.6 MB · Views: 23
Nice PDF, it contains a lot of information about cores. Thanks!! I rose the frequency to 175kHz and the core loss went down. Now i can run it at 10A continuously. It gets hot but it does not reach 100°C. Above 10A it gets way too hot but thats because i'm above the rating of the inductor. I will pick a bigger inductor. I know where to buy them, they are wound on a dark blue core. Completely dark blue. It should be very good. Rated 30A continuos peak 38A. They are also made of multiple thinner wires in parallel (around 20). This reduces skin effect losses and other losses a lot!

I also have two grey/red cores. I could use them to make inductors. They should work for high frequency.

For the power part the only odd thing i noticed is that when i draw a lot of current (30+A) the high side mosfet gets much hotter than the low side one. Maybe its normal? I will still increase the bootstrap capacitor from 1uF to 2uF to add some margin. Two 1uF in parallel (polypropilene)
 
Last edited:
So high side switching losses are always higher... Probably because of the way the bootstrap system works. Overall its still very efficent.

I will add a big inductor and the XOR gates to allow higher output voltages and... That should be it!

The voltage and current feedbacks are going to be ORed or summed with an OpAmp.

As far as i know the current feedback is dominant on the voltage one by 'default'.

Basically, if i short the output the voltage will drop a lot because of the low resistance of the wires. Like a lab bench power supply. It does not break the switching converter. The feedback should be quite fast though. Hopefully the sg3525 is fast enough. The current sensor has a bandwith of 150kHz. I dont know how to interpret this, but by logic it seems like it should be fast enough.

A lab bench power supply is very fast at limiting current. If the output shorts, nothing happens because the current gets limited. I'm aiming for high relaibility, it should work.
 
Hi. You are designing a dc/dc converter, but looks like you don't know one thing. High side and low side mosfets operation is not a just inverted control signal to driver. This simplification would work only for some constant load, but will not work for wide range of output current. Why? Because when load is decreased, inductor current must be sensed, and bottom mosfet turned only when current is going from inductor to load and output capacitor. If current is not sensed, effeciency is lost because at low duty cycle of upper mosfet, allmost all time bottom mosfet would be just open and current would go in reverse direction - from output to ground through inductor. Reverse current flow must be prevented, if you want to keep the effeciency at varying load / input voltage.
As example, there are many good dc dc synchronous buck converters ic, LT LTC series,, you just have to comnect external mosfets and adjust feedback for voltage you need, and they have pins for a shunt resistor connection ,for sensing inductor current, to prevent current backflow. There may be other solutions too, as example with small current transformer, or you may use ic like ucc24630 for controlling bottom mosfet.
 
I decreased the temp of toroids using quasi-litz wire twisting several wires whose diameter is slightly less than the penetration factor for copper. The number of wires is taken to be able to conduct full DC current. This is a quite simple computation.

Remember that inductor value is chosen from minimum output current and you must chech temperature rise at maximum load.

Remember that in PC PSU's there are fan cooling the inside of them I ignore if you are using fans.