Pulse-skipping – Minimum on-time: reasons, effects and relevance for automotive applications – Part 2
Switch-mode power supplies (SMPS) mostly operate at a fixed frequency, using pulse-width-modulation to regulate the output. There is a physical limit regarding the minimum on-time, hence how long a FET has to be kept on before it can be turned off again. In particular at high switching frequencies and low duty-cycle, this may lead to pulse-skipping modes of operation.
This article discusses what implications this involves and if it is of concern in normal operation. Furthermore, solutions to avoid pulse-skipping mode are introduced.
1. Aside from minimum on-time, Low-Power-Mode or ‘Drop-out’ mode (if input voltage is close to output voltage, hence duty-cycle approaching 100%) could be reasons for pulse-skipping, but those are not discussed in more detail herein.
There are various ways to implement pulse-skipping into an IC:
Hysteretic-mode pulse-skipping leaves out pulses that are not required, and will initiate a single pulse once the output voltage has fallen below a threshold. This allows for a relatively low output ripple.
Another implementation is Burst-mode, which, as the name says, sends a burst of pulses if the output-voltage drops below a threshold. This tends to introduce a higher output ripple.
Further options include constant on-time or constant off-time, both coming with the disadvantage that even during normal operation the switching frequency varies with load and is difficult to compensate and filter. Those variants will not be discussed in more detail in this report.
Today, switch-mode-supplies usually come with a minimum on-time in the tens to hundreds of nanoseconds. This accounts for gate-charging, blanking time, etc. See also “Understanding output voltage limitations of DC/DC buck converters”. In a typical automotive environment, the battery-voltage is nominal 12..14 V, however, during load-dump may easily double. Output-voltages are defined by the applications and their demands, mostly ranging from 1.2 V to 5 V.
Figure1 shows the output voltage (AC-coupled) of a switch-mode-supply, TPS43340-Q1, (Channel1, yellow trace), and the corresponding switch-node voltage (CH2, red trace). The part is supplied with 40 V and is set to switch at its maximum switching frequency, which is 600 kHz, driving a load of 500 mA. The measurements were taken on the standard-TPS43340EVM (Evaluation module), however the output voltage has been reduced to 1.2 V: With a switching frequency of 600 kHz and a minimum on-time of about 100 ns, the part would not enter pulse-skipping at higher output-voltages.
Figure 1: 600 kHz, VIN=40 V, Vout=1.2 V, Iout=500 mA, CH1=Vout (AC), CH2=PhaseNode, CH4=Inductor-current
The phase-node shows mostly two or three consecutive pulses, before one is skipped. The output voltage shows almost no variance, thanks to the rather frequent pulses, however, at different load-conditions it may increase.
40 V is not a normal use-case, however, using other parts supporting higher switching frequencies or with lower output voltages, it may be of concern under nominal conditions.
An example for this could be TPS65320-Q1-buck-regulator, supporting up to 2.5 MHz switching frequency. If we were to use it in order to generate 1.2 V from a 14 V input voltage, with a typical minimum on-time of 100 ns, the maximum frequency to avoid pulse-skipping would have to be reduced to ~2.4 MHz, with some margin for higher minimum on-time or higher input voltages (for example during load-dump), it has to be even lower.
The maximum allowable switching frequency to avoid burst-mode under normal conditions is calculated as follows:
For an actual design, considering losses, the situation is more relaxed, but still needs a reduction of switching frequency (in addition, account for margins and potentially higher input-voltages):
(with VIN = 14 V, VOUT = 1.2V, VF = 500 mV, ILoad = 500 mA, RDSon = 130 mΩ, DRC = 50 mΩ, tON min =100 ns)
However, potentially, a 3.3 V-rail does not need to be driven directly from battery, but have e.g. a 5 V-rail that can be used as supply, changing the duty-cycle to 66% and such more than doubling the on-time at otherwise identical conditions. Some multi-rail products, like the previously shown TPS43340-Q1, allow for cascading of rails, e.g. using Buck1 to generate 5 V and use these 5 V to drive a 3.3 V or 1.2 V output of Buck2. Here one rail is powered from 40, generating 5 V. With the ratio, even at 600 kHz, no pulse-skipping will occur. The 5 V output is then used to supply a second step-down-regulator, generating 1.2 V, as shown in the figure below.
Figure 2: Cascading of Rails
Since we can use 5 V as VIN for the 1.2 V-rail (VIN2), we will avoid pulse-skipping:
Figure 3: 600 kHz, VIN=40 V, VIN2=5 V, Vout=1.2 V, Iout=500 mA, CH1=Vout, CH2=PhaseNode, CH4=Inductor-current
Below is a summary of parameters that effect entering pulse-skipping-mode and their potential to help avoiding it:
- Output voltage: defined by to-be-supplied systems, can usually not be changed
- Input voltage: defined by the battery-voltage, but potentially another, lower rail can be used as supply for low-voltage outputs.
- Switching frequency: potentially, a reduction of the switching frequency, e.g. from 2.5 MHz to 1.8 MHz, avoids minimum-on-time-violations and still avoids the AM-band.
- Output load: defined by to-be-supplied systems: the voltage-ripple must be in a tolerable range and the inductor must be able to handle the excessive current.
- Device selection: Current-mode converters tend to have a longer minimum-on-time, but are easier to compensate. With careful design, a voltage-mode buck might be a suitable alternative.
Bottom line:
The foremost effects of pulse-skipping are output ripple and the shift in frequency.
If a frequency-shift causes the design to fail emission-tests or might couple as audible noise into AM-band during normal operation, one will have to avoid it by applying one or a combination of above given options.
If pulse-skipping only occurs during ‘extreme’ conditions, e.g. during load-dump, it is unlikely to be audible. Emission-tests are performed at nominal voltage, so the shifted frequency and its side-effects will not be recognized.
If the voltage-ripple on the output is off-limits, the system may demand e.g. for an intermediate voltage-rail in case of very low output-voltages, instead of supplying those rails directly from battery.
About the author
Frank Dehmelt is Application Engineer, Mixed Signal Automotive at Texas Instruments.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
