Hi.

I'm looking to use the miniature cheap step up and step down types of converter that are so readily available nowadays. I have a fairly good grasp of how the basic circuit works and the calculations in design of these, (thanks to that TI pdf). One thing that I'm curious about is that I have seen problems getting these to kick in with no load applied, which doesn't surprise me. Instinctively I would have assumed there was a lowest duty cycle the system could cope with which would limit action below a certain load level, or have I got completely the wrong view on that aspect?

Knowing the component values, is there a way to estimate the minimum current draw required to ensure reliable regulation of the output?

 

Instinctively I would have assumed there was a lowest duty cycle the system could cope with which would limit action below a certain load level, or have I got completely the wrong view on that aspect?

Knowing the component values, is there a way to estimate the minimum current draw required to ensure reliable regulation of the output?

Your instinct is good. The limit caused by the high frequency ability of the output switch and related parts. If the output circuit is slow and the switching frequency high then you can't get a small duty cycle.

Determining the minimum duty cycle can be hard if you don't have all of the details of the design. It is fairly easy with a single chip switcher that has the output switch built-in because it is typically specified in the data sheet. It is much harder with circuits having a controller with external switches and a transformer driven in flyback mode, for instance.

 

Thanks for the confirmation I'm on the right track Richard. I'm tempted to set this up with an additional permanent resistive load or maybe even something like a simple shunt regulator type block to keep the current draw up to a minimum level. It seems wasteful to do that when the circuity has been designed with efficiency in mind but if it gets it to work...... I can get the minimum required current empirically of course but I would have preferred to see how it could actually be derived or designed.

Thanks for your help.

 

Hi there,

Boost circuits are more fussy than bucks. The boost circuits can put out a high voltage until there is some min load applied, which can really damage things.

The buck circuits generally are not as bad though, but of course we always have to be aware of this.

The simplest way to find out is to test if you already have a unit that is working. Simply apply some input and apply a variable load resistance. As you turn the resistance higher and higher, the voltage will start to go out of wack. You'll see the point where it goes out, and this probably also depends on what the input voltage is too so you may have to vary the input voltage as well.
You have to be more careful again with the boost circuit though because some boost circuits can actually blow themselves out if there is a load that is too high in resistance. Turning the load resistance higher and higher could be damaging if you dont do it slowly and carefully, watching for just a small increase in output voltage.

There is usually a correlation with power handling capability and min load resistance. The higher power units will often have a lower load resistance requirement although there are new circuits coming out all the time.

Yes making a decision to use a fixed constant load resistor is a necessary evil. The efficiency is immediately brought down somewhat. You could calculate the difference and try to come to some reasonable.compromise.
With a boost circuit sometimes a zener and resistor will work, so if the load goes away the zener conducts and the series resistor and zener load the circuit. Some care is in order though to make sure the zener does not overheat and blow open.

 

Thanks for the additional info MrAl. It does sound like an empirical solution using a simple resistor or zener/resistor load is the most easily implemented way to go. The wasting of the small current required to ensure reliable operation is a small price to pay for reliability in a non-demanding application. It just sticks in my craw as an engineer to not have the ability to be able to easily predict it fairly closely by calculation, especially when the circuit action is so well defined and understood.

I'm a little surprised to hear you say that some boost circuits can actually be damaged without a load in place though I can see how that could occur. Seems like they should be fitted with some sort of safety circuitry as standard. I'm going to play with this over the next week or so and see where it leads.

Thanks again to both of you for your help, it is greatly appreciated.

 

Thanks for the additional info MrAl. It does sound like an empirical solution using a simple resistor or zener/resistor load is the most easily implemented way to go. The wasting of the small current required to ensure reliable operation is a small price to pay for reliability in a non-demanding application. It just sticks in my craw as an engineer to not have the ability to be able to easily predict it fairly closely by calculation, especially when the circuit action is so well defined and understood.

I'm a little surprised to hear you say that some boost circuits can actually be damaged without a load in place though I can see how that could occur. Seems like they should be fitted with some sort of safety circuitry as standard. I'm going to play with this over the next week or so and see where it leads.

Thanks again to both of you for your help, it is greatly appreciated.

Hi again,

Some of the single chip LED drivers have the ability to blow themselves up (switching transistor mostly) because they are meant to be connected to an LED permanently. What happens without a load connected is even a small duty cycle causes the output voltage to ramp up and up and eventually it blows out the switching transistor which can possibly be just a 10v or 20v device. Of course you can use a higher voltage transistor, but the lower voltage ones are cheaper and fast and have low Vsat so it kind of degrades the circuit to use a higher voltage transistor just in case someone disconnects an LED that was never intended to be disconnected while the circuit is running.

You can do some simulations with LT Spice for example (the free circuit simulator by Linear Tech).

One of the reasons this happens is because (i think someone mentioned already) is that the controller chips have a minimum non zero duty cycle so there is always some minium energy getting to the output. If there is no energy dissipator in the output then the voltage just keeps building and building until the switch transistor voltage rating is exceeded or the output filter cap voltage rating is exceeded and then that device eats up the energy as well as makes the circuit non functional after that.
If you can find a circuit that can reach a zero duty cycle then it may be self protected, but that's not usually a target goal for developing the chips because it's never intended to transfer zero energy from input to output but rather handle some sort of load. It's mostly the lab bench supplies that have to be able to do that and then they use special circuitry or a permanent non zero output load.

As to reducing the efficiency, a fixed resistor will reduce efficiency because it will be in parallel with the real load for all the time the converter is running, but a zener with series resistor may overcome this problem because the zener will only kick in say 1 volt above the output and thus apply a small load (th4e resistor) to the circuit output, but if the circuit has the normal load then the zener will never conduct so the efficiency should not be affected.
So say we have a 10v output when we have the load connected normally. If we also have an 11v zener in series with a 1 ohm resistor then if the load is disconnected and the max output current is 1 amp, then the output will rise up to about 12v. That's still higher than we want, but at least the circuit wont blow out.
A better solution of course is to use a voltage reference diode and maybe a transistor or something where the voltage of the reference is set to conduct if the output goes over a certain amount.
There are probably other ways to do this too.

Converter chips are becoming more sophisticated too these days so we never know what we might find when we look around today. There is one chip i know of that normally runs at 180kHz but during an overload switches down to about 50kHz or something like that. Efficiency is super high too (buck circuit) in the 90's. Cant find any for sale yet though through normal distributors.

 

Thanks for that additional info MrAl. Once again it all makes sense to me. The only question I would ask now is, what is a ballpark figure for that minimum current which the dummy load must draw. I realise it cannot be the same for all cases but how would you go about estimating it once you know the device? As you are talking about a permanently connected LED, a good idea anyway simply to demonstrate the circuit is functioning rationally, I am assuming that only a few milliamps will suffice. I'm considering the smaller units which seem to run up to about 1-1.5A and my maximum current draw should only be about 50-200mA, probably at the lower end of that. The powered devices are pretty basic analogue circuits fairly tolerant of supply voltage so the fact that it goes up a volt or two before being forced into action would not really be a problem.

I'm an avid user of LTSpice so that route for circuit verification is one I would have moved on to soon. I wasn't sure how accurately the models would simulate that side of the device characteristics. As always, I would hope pretty closely but I have been burned in the past with models moving into very inaccurate territory under on the edge operating conditions.

The light show guy inside me kind of likes the idea of one LED to act as a permanent load to keep the regulator stable and as an indicator, and another zener/LED combination set up to light up as a warning in an overvoltage situation as you described. I think I'm in danger of ending up making the supporting circuitry more complex than the darned regulator!

 

Hi,

Yes that's very understandable, we dont want to make the circuit too complicated. I like the LED idea though so there is an indication that the circuit is running normally.

The minimum current requirement is based on the minimum duty cycle and the losses in the circuit. IF the max power was 10 watts and the min duty cycle was 10 percent, then we might be able to reason that the min power would be around 1 watt. If the min duty cycle was 1 percent then we might reason that 0.1 watt is the minimum power output. But this is what simulators can tell us with a little thought, or at least give us some idea what might happen.

The simulators are devices that can be used to test ideas like what happens with a load of 0.1 watts, what happens with a load of 1 watt, etc. It may not match up perfectly with real life performance, but it can help.
The only way to know for sure is to test the final circuit. Connect a real load of say 50 percent of the known full load and test the output voltage. Let's say the full load is 10 watts, then connect a 5 watt load. Then we can start to reduce the load little by little and see if we start to lose voltage regulation. Once the voltage starts to move up we know that we are loosing regulation and so the circuit voltage will probably just keep going up and up after that.

Some data sheets for the control chips will show the minimum duty cycle and this can help in the simulation. In the simulation we will most likely see a higher output voltage than in real life because of the losses that are present in real life components that are not always modeled in the simulation devices. That means if we find the min load in the simulation it most likely will work in the real life circuit, although this is something that should always be bench tested. In power supply designs of all kinds there's no way to rely solely on a simulation anyway because of the nature of power components and parasitic elements that dont go into the simulation. All power supply designs should be tested for no load, half load, and full load, right on the bench. Higher power designs must be life tested as well which usually means at least an overnight run at the full load. Even new designs for wall warts are tested overnight at full load just to make sure they dont burn anybody's house down.