Buck Converter, Problems Problems and Failures!

Thread Starter

Mustafatarhan49

Joined Aug 16, 2012
56
3 Ohms is a lot of resistance for a wire. For example, 25 SWG wire with a cross sectional area of 0.2 sq mm will have a resistance of 80mOhm for 1 metre length.

Since we are expecting 1V output, 3 Ohms load will draw ~333mA for it and will satisfy the minimum load requirement. If the current is more, then the resistance is even less!

I was wondering if you could simulate to check the output voltage.

I checked the output with the 10V input. It worked well! But the problem is current is too high! I will feed uC with that and that I afraid that uC will not handle that value (50mA is the Imax for my case). Please see the attachement

What should I do to decrease that current ?

is not that output depend on Duty Cycle (Vout = Vin x D). If I want to have 5V output and 50mA current in the output, I need to use 100ohm. If I use 100ohm in the output that gives 17Voutput! That looks weird to me...I will totally lost
 

Attachments

Last edited:

crutschow

Joined Mar 14, 2008
34,281
I think the problem is that, which such a small load, the inductor is going into the discontinuous mode (inductor current drops to zero between pulses). If you view the inductor current, you will see that.

You need to increase the pulse frequency and increase the inductance until you reach continuous mode operation.
 

RamaD

Joined Dec 4, 2009
328
What is the minimum load current that you can guarantee? I guess that the 50mA load you had mentioned is the max. load. If you cannot guarantee any minimum load (the uC can go to power down mode) then you need to put a resistor as minimum load, but this resistor will be wasting power all the time.
For some sample values,
Vin = 14V
Vo = 5V
Imin = 10mA
Freq = 150kHz
D = 5/14
You would need to use 1mH inductor to stay in continious conduction mode.

What is the source of your PWM, since you had mentioned as discrete? You need to soft start to ensure there is no overshoot of the output voltage which can potentially kill the uC.

You also have stated that the input voltage can go upto 50V. Then the gate drive, should ensure that your Vgs do not exceed 20V (normal MOSFET). You could replace the resistor with a 12V zener, and insert a resistor between the collector and gate.
 

Thread Starter

Mustafatarhan49

Joined Aug 16, 2012
56
What is the minimum load current that you can guarantee? I guess that the 50mA load you had mentioned is the max. load. If you cannot guarantee any minimum load (the uC can go to power down mode) then you need to put a resistor as minimum load, but this resistor will be wasting power all the time.
For some sample values,
Vin = 14V
Vo = 5V
Imin = 10mA
Freq = 150kHz
D = 5/14
You would need to use 1mH inductor to stay in continious conduction mode.
My minimum load current is 10mA. how did you calculate the Inductor value? (I =L * di/dt ?) I am really bad in these calculations and I don't know which formulation to use.

What is the source of your PWM, since you had mentioned as discrete? You need to soft start to ensure there is no overshoot of the output voltage which can potentially kill the uC.
-what is the soft start for PWM? I am planning to take PWM from my uC. can you advise if there is another way or method which is more suitable?


You also have stated that the input voltage can go upto 50V. Then the gate drive, should ensure that your Vgs do not exceed 20V (normal MOSFET). You could replace the resistor with a 12V zener, and insert a resistor between the collector and gate.
--Actually, I was planning to use voltage divider in between. I believe If I use zener, I will have drop max ~30V in between Source to gate. Does not that mean huge power loss. Maybe I am wrong :confused:
 
Last edited:

RamaD

Joined Dec 4, 2009
328
My minimum load current is 10mA. how did you calculate the Inductor value? (I =L * di/dt ?) I am really bad in these calculations and I don't know which formulation to use.
V = L * di/dt
where V is the difference between Vin & Vout.
di =2 * Imin.
dt is the Ontime.
You can also calculate with V as Vout and dt as Offtime.
-what is the soft start for PWM? I am planning to take PWM from my uC. can you advise if there is another way or method which is more suitable?
Soft start will be to raise the PWM width slowly while powering on to final value as demanded by required output voltage.
Is this the same uC you are planning to power from the output? This uC will not have the power to start with!
Simpler will be to use some standard PWM controller.
--Actually, I was planning to use voltage divider in between. I believe If I use zener, I will have drop max ~30V in between Source to gate. Does not that mean huge power loss. Maybe I am wrong :confused:
With a resistive divider alone, it will be difficult to keep the gate voltage within spec. for the entire Vin range of 14 to 50V. So you need to have a resistive divider with a zener to ensure Vgs within specs. Replacing the gate resistor with a zener as I mentioned before, could delay turning off the MOSFET and you will need a resistor in parallel. The value of the resistor is decided on how fast to turn off (depends on gate cap of MOSFET). The gate is capacitive and requires very little power, but the simplistic driver here consumes power! If it turns out to be prohibitive, then you need to consider some other driver circuit. I remember SgtWookie gave a gate driver circuit, which I could not locate immediately.
 

JMac3108

Joined Aug 16, 2010
348
You're in over your head trying to design a buck regulator with discrete parts with no prior experience with switching power supplies.

I'd suggest using a swicthing controller IC from one of the big suppliers, TI, Linear Tech, etc. They're inexpensive and you can get app notes and design help. This will get you up and running so you can learn more.

One possibility is the SimpleSwitcher family from National Semi (recently purchased by TI). They have an entire family of parts, and have design software you can use. Follow the link below if interested.

http://www.ti.com/ww/en/simple_swit...er&DCMP=A_SVA_General&CMP=KNC-GoogleTI&247SEM
 

Thread Starter

Mustafatarhan49

Joined Aug 16, 2012
56
V = L * di/dt
where V is the difference between Vin & Vout.
di =2 * Imin.
dt is the Ontime.
You can also calculate with V as Vout and dt as Offtime.

V_out-V_in=L×d_i/d_t => 50-5=L×(2×10×〖10〗^(-3))/(6.66×〖10〗^(-7) )
==> L ==>150uF this is what actually I had found before too.


Soft start will be to raise the PWM width slowly while powering on to final value as demanded by required output voltage.
Is this the same uC you are planning to power from the output? This uC will not have the power to start with!
Simpler will be to use some standard PWM controller.
I was actaully planning to use the same uC for PWM :confused: But As you said, I will have start up problem...how can I handle this situation ? Any advise? or I will try to use PWM controller :S


With a resistive divider alone, it will be difficult to keep the gate voltage within spec. for the entire Vin range of 14 to 50V. So you need to have a resistive divider with a zener to ensure Vgs within specs. Replacing the gate resistor with a zener as I mentioned before, could delay turning off the MOSFET and you will need a resistor in parallel. The value of the resistor is decided on how fast to turn off (depends on gate cap of MOSFET). The gate is capacitive and requires very little power, but the simplistic driver here consumes power! If it turns out to be prohibitive, then you need to consider some other driver circuit. I remember SgtWookie gave a gate driver circuit, which I could not locate immediately.
I cany use zener with a Voltage Divider and parallely a resistor divider.
But I would like to see SgtWookie's circuit. Can you give the topic name if you remember?

You're in over your head trying to design a buck regulator with discrete parts with no prior experience with switching power supplies.

I'd suggest using a swicthing controller IC from one of the big suppliers, TI, Linear Tech, etc. They're inexpensive and you can get app notes and design help. This will get you up and running so you can learn more.

One possibility is the SimpleSwitcher family from National Semi (recently purchased by TI). They have an entire family of parts, and have design software you can use. Follow the link below if interested.

http://www.ti.com/ww/en/simple_switc...oogleTI&247SEM
For the suggestions, I really thank a lot! I checked the TI website.I found the "LM5574" converter was really interesting and exactly what I need. But unfortunately, It is not possible for me to use it (Chip it self and external components will be out of my price range for sure).
Indeed, you are right to say It is hard to make discrete soluiton without having any experience, but I am in the project at the moment :) So I need to handle it. I believe I will make it :) But I need some advise and suggesiton from you guys:)
 

bountyhunter

Joined Sep 7, 2009
2,512
I found the "LM5574" converter was really interesting and exactly what I need. But unfortunately, It is not possible for me to use it (Chip it self and external components will be out of my price range for sure).
Indeed, you are right to say It is hard to make discrete soluiton without having any experience, but I am in the project at the moment :) So I need to handle it. I believe I will make it :) But I need some advise and suggesiton from you guys:)
Even if you get some kind of cheap buck converter thrown together, it won't be manufacturable. It won't have any current limit or thermal protection. It won't be reliable and would not be viable as a consumer product. Just my opinion.

You might ponder something:

if IC solutions were NOT more cost effective than discrete..... why would it be that IC solutions are used in every power supply device on earth?

Open one of the cheap $5 units that run an external hard drive and you will see an offline flyback based on an integrated converter of some kind.

I say this because I spent 30 years in the business and I suspect somebody you work for believes the totally wrong assumption that "there must be a cheaper way to do it" and has dumped that task on you.

It will be your choice as to how much time you want to waste on it, but it's a learning curve.

It will be your task to introduce your boss to reality.
 

Thread Starter

Mustafatarhan49

Joined Aug 16, 2012
56
Even if you get some kind of cheap buck converter thrown together, it won't be manufacturable. It won't have any current limit or thermal protection. It won't be reliable and would not be viable as a consumer product. Just my opinion.

You might ponder something:

if IC solutions were NOT more cost effective than discrete..... why would it be that IC solutions are used in every power supply device on earth?

Open one of the cheap $5 units that run an external hard drive and you will see an offline flyback based on an integrated converter of some kind.

I say this because I spent 30 years in the business and I suspect somebody you work for believes the totally wrong assumption that "there must be a cheaper way to do it" and has dumped that task on you.

It will be your choice as to how much time you want to waste on it, but it's a learning curve.

It will be your task to introduce your boss to reality.

From the fırst sentence until the end of your line, I tottally agree with you. The thing what I really want is to show them my design works properly however this solution will cost more than IC solution. I really dont care about its manufactuability but I just want from this design to work in prototype. I will prepare an assesment and inside of that paper, I will explain the disadvantages of this discrete solution.
 

JMac3108

Joined Aug 16, 2010
348
I've been designing power supplies for a long time and the ONLY time I've ever done a discrete design done was the following...

Needed a circuit to boost 12V up to 60V, at very low current (a few mA), and a micro-controller was available on the board. In this case I designed a simple type of boost converter called a "gated oscillator" out of discrete parts, and generated the clock with the micro. BUT, even in this case, I only did it because my boss wanted it. I would have much preferred an IC controller solution because of all the protections built in.

In a discrete design you better test thoroughly under every possible transient condition to make sure nothing goes wrong. Switching power supplies look simple, but theres a lot happening and things can quickly go wrong. Make sure to look at startup, shutdown, and load transients.
 

Thread Starter

Mustafatarhan49

Joined Aug 16, 2012
56
I've been designing power supplies for a long time and the ONLY time I've ever done a discrete design done was the following...

Needed a circuit to boost 12V up to 60V, at very low current (a few mA), and a micro-controller was available on the board. In this case I designed a simple type of boost converter called a "gated oscillator" out of discrete parts, and generated the clock with the micro. BUT, even in this case, I only did it because my boss wanted it. I would have much preferred an IC controller solution because of all the protections built in.

In a discrete design you better test thoroughly under every possible transient condition to make sure nothing goes wrong. Switching power supplies look simple, but theres a lot happening and things can quickly go wrong. Make sure to look at startup, shutdown, and load transients.
Thanks for encouraging me Jmac:D I also believe that will be one and only discrete design of me too.
By the way, I need a way to calculate these capacitor values. Can you please give me which formula that you used to design your converters? or I will appriciate if you can give the name of some article or book to read.
Now, I affraid that ı will not over come from this work.
 
Last edited:

bountyhunter

Joined Sep 7, 2009
2,512
If you are set on this, I can warn you that the primary failure modes will be from not having a fast enough short circuit current limiting circuit to save the switch transistor and also not having good enough "soft start" to make sure the magnetics don't saturate at turn on.
 

RamaD

Joined Dec 4, 2009
328
You have too many problems remaining unanswered.
bountyhunter & JMac3108 have explained in detail about the additional protection and features offered by a controller, and you need those protections! If you implement soft start, switch current limit, OVP etc., apart from the (?)PWM, then the discrete solution would turn out to be more expensive and also less reliable, beating the original design goal of cost.
You would have seen that Vout = Vin * D, is not, repeat, is not true under all conditions!
 

Thread Starter

Mustafatarhan49

Joined Aug 16, 2012
56
If you are set on this, I can warn you that the primary failure modes will be from not having a fast enough short circuit current limiting circuit to save the switch transistor and also not having good enough "soft start" to make sure the magnetics don't saturate at turn on.
We decided to use PWM controller, so do we still need to think about soft start issue?
Can you please correct if my circuit formation and below comments are true?
-I connected output to an error amplifier than I checked the difference over there, by looking my desired output voltage, I defined my reference from the voltage divider. Instead of driver circuit I will use an IC over there.
 

Attachments

bountyhunter

Joined Sep 7, 2009
2,512
The Q5 pull up circuit for the FET's gate is very weak. The FET turn off will not be sharp and fast. If Q5's base drive was biased to a higher potential it's base drive would be stronger and it would saturate better.
 

Thread Starter

Mustafatarhan49

Joined Aug 16, 2012
56
The Q5 pull up circuit for the FET's gate is very weak. The FET turn off will not be sharp and fast. If Q5's base drive was biased to a higher potential it's base drive would be stronger and it would saturate better.

Yes You are right! I increase the value! but It still have some problems. The output is too small:S Is it because of voltage divider?
 

crutschow

Joined Mar 14, 2008
34,281
We decided to use PWM controller, so do we still need to think about soft start issue?
.....................................
Most controllers use PWM which doesn't eliminate the need for soft start. You need soft start unless you don't mind a large voltage overshoot at startup. That occurs with the first pulse which applies full voltage across the inductor. The generates a large pulse of current which will then overshoot the desired voltage due to the resonance between the inductor and the output capacitor.
 
Top