The advantages of using a power supply incorporating digital control to power non-linear loads

I was recently tasked with doing a presentation on the advantages of power supplies incorporating Digital Control to power non-linear loads, so I thought I would share the content with you.

A non-linear load is one that does not behave like an ideal resistor, in that the current drawn from the power supply is not proportional to voltage and/or the initial currents are often much higher than the rating of the power supply.

These loads can cause problems for power supplies, but are actually present in many applications:

Large switched capacitor banks

Point of Load DC-DC converters

Thermal printers

DC motors

The main issue from a power supply’s point of view is that the load can activate the internal over-current protection.  Over-current protection (OCP) is an essential feature for a power supply, but the power supply is usually expected to recover automatically with no manual intervention.

So to start with, let’s look at the types of OCP. There are several basic methods used:

Constant Current

Fold Back

Fold Forward

Hiccup

Constant Current

When an overload condition occurs, the output voltage falls but the output current remains at a fixed level.  This type of protection is not well suited for delivering peak loads as it can lead to the power supply latching.

 

Fold back

When the current drawn reaches the OCP limit, the voltage falls, but this time the current decreases as the overload gets heavier.  Again this type of protection is not well suited for delivering peak loads as it can result in the power supply latching.

 

 

Fold forward

When the current drawn reaches the OCP limit, the voltage falls.  This time the output current increases to a set maximum at short circuit.

Fold forward is well suited for powering up motors, but requires heavier system load cabling to handle the additional overload current.

 

Hiccup

At the OCP limit, the power supply turns off for a short interval and then automatically tries to restart.  Hiccup mode reduces the need for heavy cabling or pcb traces, and this type of protection can be modified to deliver a peak load. 

With traditional Analog Control though, the OCP points and recovery timing are fixed.

With Digital Control we can use software settings to adjust the limits and timing; for example we can set:

10s for an initial overload condition

60ms for heavy overloads

5ms for a short circuit condition with recovery times or 1 to 2 seconds

Let’s take an application example of a discharged capacitor bank being switched onto an operating power supply with Analog control.

The lower (blue) trace is the power supply current; restarting twice with the power supply current limit set at around 60A.

The top (gold) trace shows the output eventually recovering, but tolerances with the hiccup mode timing could have prevented a full recovery. 

This time the same discharged capacitor bank is switched into a TDK-Lambda CFE400M supply incorporating digital control.

The blue trace is the current, gold trace is the output voltage

Using Digital Control, we can set the thresholds and timing accurately.

50A for 1.5ms (The short circuit condition)

30A for 50ms (The over current condition)

 

Notice that there are no multiple attempts to recover after the capacitor bank is applied to the power supply.

To summarize:

Digital control can allow for precise and repeatable current limiting using load dependant timing.  We are not restricted to the value of a timing capacitor which can change due to:

a) Batch tolerances

b) Aging of the capacitor

c) Capacitor values changing with temperature

Digital control allows for easy tailoring for different applications with no physical modification of the power supply is needed.  All changes are handled with software programming!

Power Guy

 

Reducing Switching Power Supply Radiated & Conducted EMI

One application issue that comes across my desk on a regular basis is where a customer has gone to an outside lab to certify their equipment for EMC, and they have failed conducted or radiated noise.

Usually the power supply in question is an open frame type, which does not have the shielding that a metal enclosed power supply has.  There are two areas that are worth checking; grounding points and wire harnesses.

1. Grounding Points

Power supplies utilize decoupling capacitors; two are typically connected from input to earth ground (see below). Likewise, two are connected from the output to earth ground. This keeps the noise currents circulating close to the power supply, rather than allowing them to radiate around the end user’s system.

 

In an enclosed power supply these capacitors will be grounded through the metal case, but with an open frame type, it is up to the user to connect these points to ground.  With the power density of products today, there often is more than one point on the power supply printed circuit board that needs to be connected.  A common mistake is to only connect one, which can cause excessive radiated and conducted noise.

The installation manual will show which mounting holes / points need to be grounded.  In the product below, three mounts should be connected (A, B & C).

 

 

The photo below shows the same power supply undergoing EMC testing, and it can be noted that the unit is connected to a metal plate with metal standoffs.

 

 

A quick glance of the underside of the printed circuit board will show which mounting holes have traces that need to be grounded. This smaller model has only one grounding point at the bottom right hand side of the board.

2. Wiring harnesses

In the test photo, it can be seen that the cable harnesses are neatly dressed and are kept away from the power supply.  Wiring that is positioned above or below the unit will pick up radiated noise, thus defeating the purpose of having the EMI filter components.

If I am assisting customers on site, I always pack some tie-wraps in my tool kit to re-route any offending harnesses.

An earlier blog covers the standards http://power-topics.blogspot.com/2007/11/guide-to-emc-standards-for-power.html

 

“Brute Force” Parallel of Power Supplies

You will see in many of our instruction manuals a warning about not connecting power supplies in parallel that do not have current share capabilities.

At first it would seem a nice easy way to get extra current.  Take two like power supplies, connect them together and they will deliver twice the current?

 

Unfortunately there is a good chance that the two power supplies will not current share due to their output voltage set points.  The power supply with the highest output voltage setting will deliver as much current as it can until it reaches its current limit threshold and then the output voltage starts to drop.  The second power supply will then take over and provide the balance.  The output voltage might glitch during the transition, affecting system operation.

For example, take two 24V 10A power supplies with an over current set point of 120% powering a 15A load:

Power supply A might deliver 12A (now at its current limit point)

Power supply B would then deliver 3A.

One could argue that the power supply is being protected by the current limit.  There are two issues with this though:

1.  A power supply is not designed to operate in current limit indefinitely. Internal temperatures will rise, reducing the life of the product

2.   The safety certifications for UL, CSA are based on 100% load, not 120%

My recommendation is to use a power supply with a higher current rating, or choose one with a current share feature.

What Makes Switching Power Supplies Hiss?

I attended a technical briefing two years ago given by one of our Product Managers on TDK-Lambda’s 400W CFE series, and was shown the various ways the product could potentially reduce internal energy losses - particularly at very light loads – using features available from the power supply’s digital control circuits.  He did say that there were a couple of small side effects if all the features were enabled.

Recently one of our salespeople emailed me stating that in an off-load condition, his customer could hear a 10 kHz hissing sound coming from one of our 60W output power supplies and was curious if the unit was faulty.  Normally I would have asked about capacitive loading conditions or if the product was being grounded correctly.  Remembering the briefing I checked the product specification for no-load power draw.  Seeing that it was a low 200mW, I talked to one of our Design Engineers.

Before Energy Star legislation came into force, power supplies would often draw 5, 10 or even 20 Watts when in a no-load condition.  Apparently this was accounting for about four percent of the total electricity used in the US.  Now most small power supplies will draw less than 300mW.  Although the legislation is aimed at consumer products like phone chargers, many industrial customers are now requesting similar energy saving products.

I was informed that many of the new power supply control chips employ advanced energy saving features, and what my customer was hearing was probably the converter going into “burst-mode”.

AC-DC power supplies above 70W will often have two converters in them; one is the boost PFC (power factor correction) converter which ensures that the input current is not drawn in big “gulps”, but more sinusoidal like the applied AC voltage.  An interleaved PFC converter shown below is used for this purpose.

 

The other is the main switching converter which chops up the high voltage DC from the PFC section and provides isolation, voltage step down, regulation and filtering.  A commonly used “half bridge” circuit is shown below.

 

Back to the subject of hissing!  Without energy saving modes, both of these converters run continuously, regardless of the applied load.  Each time the FETs switch, energy is used by the gate drive circuits and lost in parasitic inductance and capacitance.

To avoid this, designers of control ICs have implemented features where the power supply engineer can have the PFC converter, the main converter, or both converters turn off for short periods of time, aka “burst-mode” or pulse skipping, during light loading.

This means that instead of the FETs being turned on every 10us for a power supply operating at 100 kHz, the FETs will be switched on for a few cycles say every 100us.  This is enough time to replenish the internal capacitors with energy and keep the output voltage in regulation.

 

The upside is much less power is dissipated in the power supply, but the downside is that instead of the product operating at 100 kHz (well above the human hearing range) a faint hissing might be audible as the burst mode operation operates at 10 kHz.

If the energy saving feature of the control IC is fully maximized, whereby a much longer interval between bursts is enabled, an increase in output ripple may be also observed.  Often this is acceptable to the end user provided that the output voltage remains regulated.

My suggestion to the salesperson was to offer our customer an alternative power supply that had a slightly higher off load power draw, but no burst-mode operation.  That suggestion was accepted.

Power Guy

Power Supply Filter Capacitor Values Can Change with the Applied Voltage

I read an article in one of the publications we advertise in recently where an Engineer had designed a timing circuit, but when he tested it, the frequency was too high.  He rechecked his calculations and found out from the capacitor datasheet that the value of the particular multilayer ceramic capacitor he had chosen, changes with the applied voltage.

I did not pay much attention to it, being more involved with power supplies, until I was talking with one of our TDK-Lambda Engineers regarding non isolated POL (Point of Load) converters, and he gave me the same warning about the value selection of the filter capacitors.  This is covered in the ceramic capacitor datasheets under DC Bias characteristics.

POL converters rely on quite large ceramic capacitors on the input and output to reduce the effect of the fast transient currents drawn by FPGAs, (which can cause the output voltage to deviate), with values sometimes approaching 2,000uF.

The concern our Engineer had was that our customers might not know about this.  Intrigued I decided to investigate.  Below is the DC-Bias Characteristic for a 22uF 16V multilayer ceramic capacitor.

If a capacitor value of 22uF was recommended by the application note for filtering the 12V input voltage on a non-isolated converter and the user picked this particular part, in actuality, the real capacitance would be closer to 12uF.

Upon testing the filtering, the user might complain that the application note was incorrect, whereas in fact the capacitor datasheet had not been interpreted correctly.

Something to bear in mind!

 

Power Guy

What Size Heatsink Do I Need?

Today there are a large array of high power modules in the range of 300 to 1,000-watts, both DC-DC converters as well as AC-DC power modules, which are commonly referred to as “bricks.” Even though these devices feature high conversion efficiencies in the area of 85 to 90% (or higher), some power is lost in the form of heat that must be dealt with in order to maximize the lifespan of the end product. For example, a 500-watt power module with 90% conversion efficiency would generate over 55-watts of wasted heat within the module that must be removed to maximize its reliability.


The concentrated power density (watts per cubic inch) within these power modules make them a challenge to cool in real world applications. Most high power bricks are packaged in thermally conductive plastic or epoxy cases with integral metal baseplates. The high power components within the bricks (i.e., semiconductors, inductors, transformers, etc.) are thermally coupled to these baseplates, which in turn can be attached to external heatsinks or liquid-cooled cold plates in order to keep the baseplate at or below its maximum operating temperature (typically 85 to 100°C).

Half Brick – DC-DC Converter

Full Brick – AC-DC Power Module

The maximum baseplate temperature is primarily determined by the maximum internal junction temperature of the semiconductors within the power bricks. The term “thermal management” refers to the designer’s challenge of cooling these power bricks by considering the many levels for heat transfers via conduction (direct contact between solids), convection (contact with air or a fluid) and thermal radiation (electromagnetic infrared energy), both internal and external to the power module.

Thermal impedance

The diagram above shows the series-connected thermal resistances that impede the flow of heat from one level to the next. These impedances need to be considered, beginning with the internal semiconductor’s junction temperature relative to its case, the thermoplastic module case and its metal baseplate, and ending with a mechanically attached heatsink that conducts away the heat from the baseplate to the surrounding ambient air via natural or forced air convection cooling. Heatsinks are designed to cross thermal barriers primarily by substantially increasing the surface area that comes in contact with the ambient air, thereby providing enhanced convection cooling. Because the mating surfaces of the power module’s baseplates and heatsinks are not perfectly flat, some type of thermally conductive interface material is required to fill the tiny voids. This interface material can range from a thin layer of thermal grease to a custom designed silicon pad.

AC-DC Power Module with Heatsink & Other Components

Selecting the proper size and shape of a heatsink and determining if forced air cooling is required are among the tradeoffs the designer needs to consider. This process begins with a detailed review of the power module’s specifications and knowledge of the end product’s heat loads, internal and external operating temperatures, space constraints, and available air flow sources, paths and restrictions.

The next step in this process is to determine the amount of power that will be lost (wasted) within the power module, based on its efficiency. This information for computing this is usually listed on the power module’s datasheet or installation manual, but it can also be determined by actual measurements of the input and output powers. For this example, we will use a typical AC-DC power module with a 48V/10.5A, 504W output rating, and a typical efficiency of 85% with a 120VAC input. By the way, the 85% efficiency rating is very good considering the fact that this module contains full-bridge rectification and active power factor correction AC front-end circuits as well as an integral DC to DC converter. In addition, this module has a maximum operating baseplate temperature, as measured at its center point, of 100°C.

Based on the above information, to compute the internal power dissipated (wasted heat); we can use the following formula:

Pd = (Pout / η) – Pout
Definitions & Calculation Example:
Pd : Internal Power Dissipated (W)

Pout : Output Power (504W)

η : Efficiency (85%)

Pd = (504W / 0.85) - 504W = 88.9W
To calculate the required baseplate to ambient air thermal resistance that would be needed for this application, the following formula would apply:
θba = Tb - Ta / Pd
Definitions & Calculation Example:
θba : Baseplate to Ambient Air Thermal Resistance (°C/W)

Tb : Baseplate Temperature (100°C)

Ta : Ambient Air Temperature (40°C)

Pd : Internal Power Dissipated (88.9W)
θba = 100°C - 40°C / 88.9W = 0.67°C/W

In this example, we would need a heatsink (with or without air flow) that provided a thermal resistance of 0.67°C/W. However, unless the heatsink includes a thermal interface material like thermal grease or a pad in its rating, we need to account for this additional thermal contact resistance (θbs), which can be on the order of 0.1°C/W. Therefore, the required thermal resistance of the heatsink itself, with the thermal interface material included, can be calculated per this formula and example:

θba-bs = θba – θbs

θba – θbs = 0.67°C/W - 0.1°C/W = 0.57°C/W

The next step in this process is to review specifications for potential heatsinks that have a thermal resistance of 0.57°C/W. In this case, the power module has three optional heatsinks to choose from as shown in the chart below.

The Y axis of this chart shows the thermal resistance between the heatsink and the air (°C/W) and the X axis shows the required airflow velocity for the three heatsinks. In this example, we need to find 0.57°C/W along the Y axis and then move to right along the X axis to where it intersects a heatsink curve. In this example, 0.57°C/W intersects with the HAF-15T heatsink curve at about the 1 m/s airflow velocity point. Therefore, for this application we would select the model HAF-15T heatsink and would have to provide forced air cooling with an air velocity of 1 m/s. To translate m/s (meters/second) into LFM (linear feet/second), use this general conversion factor: 1 m/s = 200 LFM. In this example, since 1 m/s = 200 LFM of forced-air velocity, the fan required for this application must provide 200 LFM.
Based on the above, we have now determined the requirements for cooling this power module with a heatsink, thermal compound and forced air flow. If we wanted to be more conservative and improve the MTBF of the module, we would recalculate the required heatsink with the assumption that we wanted to keep the baseplate temperature at 85°C.

A word or caution should to be injected here. After going through the thermal calculations and selecting a heatsink, air flow, etc., the next step is to confirm the “paper-design” by running actual tests on a sample unit. The tricky part is to get access to the center point of the power module’s baseplate so you can measure the temperature at that point while the module is operating under load. One way to do this is to drill a hole in the center of the heatsink so the leads from a thermocouple can be mounted on the module’s baseplate and routed to your temperature measurement device.
In summary, we have shown how to determine the correct heatsink for power module applications. As the efficiencies of these devices improve the need for cooling will reduce, but the designer should always be aware of the heating effects from not only from the power module, but also from nearby devices. Therefore, it’s always best to run actual thermal tests with thermocouples attached to the power module and inside the end product to insure the design will be as reliable as possible.

Power Supply "Remote Sense" Mistakes and Remedies

Most medium to high power AC-DC power supplies and some DC-DC converters include "Remote Sense" connection points (+ and - Sense) that are used to tightly regulate the supply's output voltage at the load. Since the output cables that connect a power supply's output to its load have some resistance, as current flow increases, so will the voltage drop across the cables (I x R = Voltage Drop). Moreover, since it's best to regulate the voltage directly at the load, the use of the two Remote Sense wires connected from the supply to the load will compensate for these unwanted voltage drops. Refer to Fig. 1 which shows the typical connections when the Remote Sense function is used.

Fig. 1: Power Supply with Twisted "Remote Sense" Wires Connected to the Load

Typical "Remote Sense" Problems & Remedies

1. Most remote sensing circuits are capable of compensating for from 0.25V to 0.75V of voltage-drops across the output cables. However, to be sure, always check your power supply's instruction manual to determine its maximum remote sense compensating range. If the voltage drop across the output cables exceeds the compensating range of the remote sense circuits, the voltage at the load will no longer be regulated. This problem can be remedied by either reducing the length of the output cables or increasing the size (heavier wire gauge) of the output cable's to reduce the excessive voltage drop. Voltage drops across the output cables should be minimized since this is a source of wasted power. For example, with just a 0.5V cable drop with a 100A load, the lost power amounts to 50W in each cable or 100W total.
2. The remote sense function automatically increases the output voltage at the output terminals of the supply to compensate for any unwanted voltage drop in the output cables with heavy load currents. Likewise, the remote sense function decreases the output voltage of the supply when the required load current is reduced. In some applications, the power supply's output needs to be adjusted by the user to voltage higher than its nominal (e.g. 5V nominal, adjusted to 5.5V). Always adjust the power supply's output while measuring the voltage at the load. In addition, care should be taken to assure that under full load that the remote sense function does not push the Vout to a higher voltage that could possibly trip the OVP set-point and shutdown the supply. Therefore, always read the power supply's instruction manual to be aware of the supply's adjustment range and OVP set-point.
3. The remote sense leads carry very little current so light gauge wires can be used. However, steps should be taken to ensure that the remote sense wires do not pick up radiated noise by either twisting the + and - Sense wires together and/or by shielding the wires from the noise (refer to Fig 1). It is best to use different colored sense wires (e.g., black and red) so that after they are twisted it is easy to determine which wire is the + and – Sense.
4. Refer to Fig. 2 below for a simplified schematic of a power supply's remote sense circuits. It is important to observe the correct polarities, i.e., the +Sense wire should connect at the load near the +Vload connection and the –Sense wire should connect at the load near to the – Vload connection. If by mistake the remote sense wires are crossed-connected (+Sense to –Vload and – Sense to +Vload) current will flow in the Sense lines and burn out the internal Rsense resistors, causing a malfunction of the supply. Typically, these internal Rsense resistors are around 10 to 100 Ohms with a maximum rating of 0.5W.

Fig. 2: Simplified Schematic of Remote Sense Circuit with External Output & Sense Wires

5. We have seen applications where the user has installed a switch or fuse in series with one or both output wires. This can cause a serious problem if the remote sense lines remain connected to the load, because if the output cable switch or fuse opens, current will flow in the sense lines and cause the internal Rsense resistors to burn up. System debugging can cause similar problems, for example, where the power and sense cables are located on separate connectors and if by error, only the power cable connector is disconnected. 
6. There are applications where the user may not want to use the remote sense feature. In these cases, the remote sense lines should not be left open for optimum load regulation; instead, a local sense configuration must be used. Referring to Fig. 3, to use a local sense set up the + and -Sense lines should be connected to either their corresponding local sense (LS) terminals, which are provided on many power supplies, or connected to the corresponding +Vout and –Vout terminals. Most power supplies are shipped from the factory with these "Local Sense" jumpers installed on the power supply (see photos below).

Fig. 3: Schematic of Power Supply with "Local Sense" Jumpers Installed

Photo of Power Supply with Local Sense Wires Connected (see Red & Black jumper wires)

Photo of PSU with Sense Screw Terminals Connected to Output Screw Terminals with Metal Jumpers

In summary, the "Remote Sense" feature automatically compensates for unwanted output cable drops, which vary as the output current increases and decreases. This feature is advantageous to the user, but is subject to mistakes that should be avoided to insure the proper operation of the power supply and the end-product.

What does SELV mean for Power Supplies?

SELV stands for Safety Extra Low Voltage. Some AC-DC power supply installation manuals contain warnings concerning SELV. For example, there may be a warning about connecting two outputs in series because the resulting higher voltage may exceed the defined SELV safe level, which is less than or equal to 60VDC. In addition, there may be warnings about protecting the output terminals and other accessible conductors in the power supply with covers to prevent them from being touched by operating personnel or accidently shorted by a dropped tool, etc.

UL 60950-1 states that a SELV circuit is a “secondary circuit which is so designed and protected that under normal and single fault conditions, its voltages do not exceed a safe value.” A “secondary circuit” has no direct connection to the primary power (AC mains) and derives its power via a transformer, converter or equivalent isolation device.

Most switchmode low voltage AC-DC power supplies with outputs up to 48VDC meet the SELV requirements. With a 48V output the OVP setting can be up to 120% of nominal, which would allow the output to reach 57.6V before the power supply shuts down; this would still conform to the maximum 60VDC for SELV power.

In addition, an SELV output is achieved through electrical isolation with double or reinforced insulation between the primary and secondary side of the transformers. Moreover, to meet SELV specifications, the voltage between any two accessible parts/conductors or between a single accessible part/conductor and earth must not exceed a safe value, which is defined as 42.4 VAC peak or 60VDC for no longer than 200 ms during normal operation. Under a single fault condition, these limits are allowed to go higher to 71VAC peak or 120VDC for no longer than 20 ms.

Don’t be surprised if you find other electrical specs that define SELV differently. The above definitions/descriptions refer to SELV as defined by UL 60950-1 and other associated specs regarding low voltage power supplies.

Power Supply Rise and Fall Output Characteristics

A common question asked by our customers is “what are the power supply’s rise and fall output characteristics?” Our usual answer “it depends” sometimes raises an eyebrow. Let’s have a look at why.
As you know, there are two ways a power supply can be turned-on and off. One method is to apply or remove the AC input power via a switch or circuit breaker to the supply. Another method is use the Remote On/Off control of the supply, if the unit has this feature. Let’s examine both methods.

The curves below (Fig. 1) show the typical delay between when the AC input power (Vin) is applied and the power supply’s output reaches its rated voltage of 12V under full load conditions. As can be seen (in this example) with a low AC input of 85VAC the output turn-on delay is slightly more than with higher input voltages. However, in this example, the typical turn-on delay is about 250ms and the output rise time is about 25ms.

Conditions:

Vin	: 85VAC (A)
	: 115VAC (B)
	: 230VAC (C)
	: 264VAC (D)
lout	: 100%
Ta	: 25°C

                                                                

Fig. 1

The next curves (Fig. 2) show the same power supply’s output (with 100% load) when the AC input is removed. The time delay from when the AC power is removed (or lost) is referred to as the Hold-Up time spec. For this supply, the specified minimum hold-up time is 16ms. As shown below, the measured hold-up time is about 30ms, which meets the spec. The output fall time is approximately 10ms, when measured from 90% to 10%. Note that the energy is pulled from the power supply very quickly due to the heavy load on its output.

Fig. 2

By comparison, without a load (zero load), the curves below (Fig. 3) show that the Hold-Up time of the power supply increases substantially to about 1.8 seconds, and the output drops to zero in about 6 seconds!

Fig. 3

Below is a graph (Fig. 4) for this power supply that shows how minimum Hold-Up time varies as the load changes between 10% and 100% (maximum load).

Fig. 4

Next, we shall review how the remote on/off control (if used) can affect the power supply’s output rise/fall, and delay times.

Conditions:

Vin	: 115VAC (A)
lout	: 100%
Ta	: 25°C

Fig. 5

In the above curves (Fig. 5), the remote On/Off control is the bottom trace and above that is the power supply’s DC OK signal level. This situation assumes the AC input in always on. As can be seen, from the time the Remote On is activated (goes from high to low logic level); it takes the output about 150ms to reach its full rated voltage. And, the DC OK signal changes state when the output level is about 75% of its rated voltage, which in this case is about 125ms after the Remote On signal is activated.

The next set of curves (Fig. 6) show the reverse situation, where the Remote On/Off signal is used to turn the power supply off.

Fig. 6

There is about 25ms delay after the Remote On/Off changes state to when the Vout reduces to approximately 75% its nominal. In addition, at that time the DC OK signal changes state. The output fall time, from 90% to 10% of nominal, is approximately 15ms.

From the above curves and explanations, I believe the reader can see why our answer at the beginning was "it depends." In these types of measurements, delay times are in many cases more important than rise and fall times. Other considerations include the AC input voltage, the load, operating temperatures, measurement criteria and the design specifications for the power supply.