Synchronous Buck converter DCM/CCM efficiency comparison in light load condition

[yonglim]

Hi everyone,

I designed a synchronous buck converter with the zero current detector (ZCD) to have a option to select between CCM/DCM at light load condition. The control method is traditional voltage mode PWM and no pulse skipping/PFM in DCM mode.

At first, I thought the efficiency is always better in DCM mode during light load but surprisingly in my simulation the efficiency in CCM mode is actually better than DCM mode. The way I measure the efficiency is have a fix load at the output and let the converter run for a long time, after the converter reach steady state I take the average of the input current flow into the converter.

When I compare the input current waveform between CCM/DCM mode, there is a reverse current flow back to the input source in CCM mode when the high side FET turned on, since the inductor current is negative, while in DCM mode the input current always flow into the converter.

Now, if I take the average of the input current for both CCM/DCM mode, the average input current in CCM mode is actually less than the average input current in DCM mode, which results in the efficiency in CCM mode is better than DCM mode.

People always implement pulse skipping/PFM mode in DCM but I think this is not a fair comparison, I am interested in the efficiency comparison between two mode under the same voltage mode PWM control.

Someone said I should take rms current instead of average current, I think this is not correct because if you take rms current no way the converter can reach 90% efficiency number.

I hope anyone who experienced in switching converter design can join this thread to discuss or answer my doubt.

 

[mtwieg]

Now, if I take the average of the input current for both CCM/DCM mode, the average input current in CCM mode is actually less than the average input current in DCM mode, which results in the efficiency in CCM mode is better than DCM mode.

What about the voltage and current on the output? Are those the same? Are the two converters regulated with a feedback loop?

Someone said I should take rms current instead of average current, I think this is not correct because if you take rms current no way the converter can reach 90% efficiency number.

If your source is an ideal DC voltage, then use the average current.

 

[yonglim]

Hi mtwieg,

Yes, both mode are regulated with the feedback loop with same output voltage and load current.

 

[mtwieg]

Then DCM should be more efficient due to lower RMS current and lower switching losses. I would look at your simulation and measure losses in each component and then compare.

 

[yonglim]

Then DCM should be more efficient due to lower RMS current and lower switching losses. I would look at your simulation and measure losses in each component and then compare.

Hi, I agree DCM should be more efficient but let's assume the converter is a black box and what I can measure is only the input source of the converter, what is the correct way to measure the efficiency?rms current or average current? average current seems to be make more sense to me but as I said in the first post, due to the reverse current at light load in CCM mode, the average current is actually less than the DCM mode, results in higher efficiency than DCM.

So if I try to take the absolute value of the current waveform before I take the average, the result seems correct to me now, the efficiency number in heavy load looks the same but in light load there are significant different in CCM/DCM mode, DCM efficiency is higher than CCM in light load. But I am not sure I am doing the right thing or not.

 

[FvM]

So if I try to take the absolute value of the current waveform before I take the average, the result seems correct to me now, the efficiency number in heavy load looks the same but in light load there are significant different in CCM/DCM mode, DCM efficiency is higher than CCM in light load. But I am not sure I am doing the right thing or not.

No, not right.

Generally, real power transported through a line is the average of instantaneous power, P(t) = V(t)*I(t)

Similarly, if V is constant (or nearly constant), Pavg = V*average(I(t))

Taking an absolute value would drop the energy flow direction, which isn't correct.

Finally, efficiency is simply η = Pin,avg/Pout,avg

 

[D.A.(Tony)Stewart]

Since your thread is concerned only about efficiency, the only valid way to measure this is with true RMS power input.
It is a relatively new requirement for no load efficiency with IEC 16301 sect4.5 with 5mW max for no load and standby.

What are your goals and results?

 

[FvM]

Since your thread is concerned only about efficiency, the only valid way to measure this is with true RMS power input.

Sounds mistakable. We have real ("true") power, which is average of instantaneous power, as discussed above, and we have RMS current or voltage. They are related in case of a resistive load, but not generally. There is no thing like "RMS power".

 

[BradtheRad]

One of the desirable things about operating in CCM, is that current peaks are not as great compared to DCM. Therefore CCM has lower I x I x R losses, which improves its efficiency to that degree.

 

[treez]

so you are measuring efficiency but only in light load case?

With CCM the current goes back into the input cap in light load and so the average input current looks less, because of the backflowing current. However, that current will flow through stuff and cause dissipation.
Is this a simulation or a prototype?

If its a simulation then just put a diode upstream of your input cap (for both the dcm and the ccm versions) and measure the current through that.

Then efficiency, as you know, is

(VOUT * IOUT) / (VIN * IIN)

In light load, most efficienct controllers that are good for light load efficiency always go into skip mode...they do not switch continuously in light load...so you may look for a better controller if light load efficiency is your panacea

 

[mtwieg]

One of the desirable things about operating in CCM, is that current peaks are not as great compared to DCM. Therefore CCM has lower I x I x R losses, which improves its efficiency to that degree.

Lower peak current does not necessarily mean lower RMS current, even when the average current is the same.

 

[treez]

regarding light load, any hard switched smps, when the load goes light enough, will eventually go into dcm.....unless it goes into pulse skipping first.

as said by others, dcm avoids reverse recovery of rectifier

 

[yonglim]

so you are measuring efficiency but only in light load case?

With CCM the current goes back into the input cap in light load and so the average input current looks less, because of the backflowing current. However, that current will flow through stuff and cause dissipation.
Is this a simulation or a prototype?

If its a simulation then just put a diode upstream of your input cap (for both the dcm and the ccm versions) and measure the current through that.

Then efficiency, as you know, is

(VOUT * IOUT) / (VIN * IIN)

In light load, most efficienct controllers that are good for light load efficiency always go into skip mode...they do not switch continuously in light load...so you may look for a better controller if light load efficiency is your panacea

Hi treez, this is a simulation, just want to clarify, you suggest to put a diode in series with the input cap to block the reverse current? If this is the case then where the reverse current will flow to? I understand during light load people always do pulse skipping/PFM to improve the efficiency, but what I am interested in now is the comparison of the two identical synchronous buck converter at light load efficiency, with the DCM mode enabled and disabled.