Q1 (eafpres1): Who needs regulation?
The topic of battery powered low-powered circuits and relationship of same to IoT is a great one. As I like to say, all those nodes are “sensors” in some sense, and sensors go hand in hand with analog almost every time. So IoT is good for the analog industry.
However, here is a question only an ignorant manager type would ask:
If we are going to power our circuit ONLY with a battery, and we design it appropriately, why do we need voltage regulation at all? The discharge curve for our chosen battery is completely known, and we know what load we will be placing on it.
A1a (Bob Dobkin)
While battery characteristics maybe known, batteries change with time temperature and state of charge. Also as they age, characteristics change. Most popular battery types change 30% to 50% with state of charge. Many circuits will not tolerate this much change in voltage. Also, the battery impedance Changes with the state of charge. This can add another unknown into operating directly from the battery. Finally batteries or not short-circuit proof. Inserting a regulator between the battery and the load provide some protection.
A1b (Greg Waterfall)
You are correct. There are many processors and other analog circuits that can live with the wide range of voltage that will be supplied by a discharging battery.
However, LDOs not only provide voltage regulation, but they also act as filter for up stream power supply noise. In addition there are some low power applications that my benefit in efficiency if an LDO is use. In the case of battery powered sensors there are three places were you might want to use an LDO to regulate the voltage:
1. To provide a stable supply. Most analog circuit performance parameters will vary with power supply voltage, such as an OpAmp's input offset. Having a regulated supply will improve sensor accuracy.
2. To provide noise isolation for analog circuitry. A processor connected to the battery will induce noise onto the battery rail that may cause issues for your sensor measurements.
3. To improve efficiency. For very low power processors using a low Iq LDO to lower the supply voltage to the processor can improve efficiency because the processor power consumption is proportional to the supply voltage squared.
A1b reply to Greg Waterfall by etnapowers:
@Greg, really nice post, thank you. regarding the point 1 that you described, I recall that I've worked on a DC DC switching voltage regulators which integrates an LDO able to provide a stable voltage, being feed by the output of the switcher. It's a fashionable solution to integrate a DC DC converter and a LDO in the same chip.
Q1 Comment by eafpres1
@Bob–you are definitely correct that there are many variables which affect the battery. Perhaps in some cases, if the sensors were already there, you could get away with it. What I'm saying is that, let's look at a mobile phone (just an example). There is one or more temperature sensors. There is some sort of fuel gauge. There is a clock/timer. So in theory you could keep track of the ageing as well as state of charge and the environment, without adding anything new to the hardware. Now, in the real case of a phone, we are going to recharge it vs. swapping batteries (usually), so that dictates a power supply and regulation and all that goes with it. But some sensor packages may also have all the needed data (for other reasons) and could do a pretty good job of knowing what to expect from the battery.
All this is hypothetical. For IoT the cost, size, etc. are very critical. If it could be done “good enough” w/out some parts, that's how it will be done.
Q1 comment by Bob Dobkin
Cell phones are full of switching regulators and linear regulators. One of the problems with batteries is they are not as consistent as we would like them to be. The electrochemical system in a battery has all kinds of aberrations as it ages and goes through many charge cycles
Follow-on question re: Q1 from eafpres1: Bob, of course you are right about batteries. Another question: In my fantasy world of low voltage low power IoT nodes operating only on batteries (no charging), does the regulator help with ESD protection? ESD is a major issue and as IoT sensors are attached to who knows what, I suspect static discharges to hit lots of these devices. Just imagine the recently hot topic of “wearables”; you put something in clothing and it will get zapped.
Reply by Bob Dobkin to eafpres1 follow-on question:
I think it depends on where the ESD strike occurs. All linear regulators should be able to take an ESD strike on the input or output pin. Also, the regulator can absorb An ESD strike if the sensor is clamped to the power supply. Batteries can absorb ESD as well but may allow more voltage excursion for very fast transient.
I like the regulator in the middle for over current protection. Some low impedance batteries can put out huge currents under short-circuit. Accidental shorts then can fuse wires
Q2 A reader question
Here is a question from a reader this week that could not attend this session:
We have DC-DC switching regulators so that we want to get very high efficiency as compared to LDO's.
Now, in order to improve the PSRR we introduce the LDO back taking the hit on the efficiency? The total efficiency has taken a toll?
A2a (Bob Dobkin)
There are many linear regulators that operate with less than 0.5V across them. This can minimize the hit on efficiency. If your output voltage is low sometimes the efficiency of an LDO can approach a switcher. Another trade-off is to use additional filter circuit on the output of the switching regulator.
A2b (Greg Waterfall)
Yes, the over all system power efficiency has taken a hit to the tune of the current running through the LDO times the input output voltage differential.
However, many new LDOs can be operated at very low drop out, even as low as 300mV at very high currents. In the case of an LDO with 3.3V input and 3.0V output this is only a 9% efficiency hit. Which, in many cases is worth the improved system performance provided by using and LDO.
A2b comment to Greg Waterfall by amrutah:
Greg: Thanks, I now understand the intention behind cascading the Switcher with a LDO and a optimized solution of a regulator.
Just to add: to think of a regulator (dc-dc + LDO) of having 2 power FETs in series to support the same high current kept me thinking.
General discussion: “Linears are getting better”
Bob Dobkin: Even with all the switchers, and their popularity, linears still have a place. Today's linear regulators are really high-performance. In addition to a regulator loop they include protection and monitoring functions.
Regulators are available that report current and temperature as well as allow the user to adjust the current limit. Supply currents down to a few microamps are available allowing for them to be continuously operated on a battery.
One of the most difficult things in designing linear regulators is protection from abnormal use an overload. Many of today's regulators can withstand reverse voltage on the input or output without blowing up. Careful design of the power transistor gives a large safe operating area and thermal protection prevents the device from overheating.
Older devices (cheap devices) leave you stuck with high supply current, large dropout voltages, and unknown Overload and abuse characteristics. So new linear regulators are advancing their usefulness and still have a place over switchers.
Greg Waterfall: To add to Bob's comments, with improved dropout and higher bandwidth PSR, newer LDOs can often replace bulk passive Pi filters. Providing smaller board space, easer system design, improved noise performance, and sometimes improved cost.
General discussion about “Dissipating power”
Bob Dobkin: Of course one of the biggest problems with linear regulators is getting rid of the heat. The first thing that comes to mind is a heat sink or or a heat sink and fan. But there're other options:
Resistors can be spread around the board to distribute wasted power. The resistors are in series with the input of the regulator, and designed such that at minimum input voltage and Max output current there is still sufficient voltage across the regulator for operation. This keeps the heat from being localized and can allow the board to be a heat sink. Also several regulators running at low power can be paralleled to spread the heat around the board. There are aluminum PC boards designed especially for heat dissipation.
These options are good for dissipating several watts. In hours hi -10 to 20 W A good size heat sink is needed.
Q3 Steve Taranovich
Linear Reg with current source vs. voltage source
What are the advantages and disadvantages of a Linear Regulator with a current source vs. voltage source internally? I know Bob Dobkin's team has developed a voltage controlled current source inside a new Linear Regulator
A2a (Bob Dobkin)
If you are doing a switching regulator, you must have a big output capacitor on it. This means that the AC impedance is low even though the feedback loop maybe controlling current. A linear regulator can provide A high AC output impedance if it does not require an output capacitor.
A2b (Greg Waterfall)
Linear Regulators can also be used to make simply implement many other valuable circuit functions such as high power current sources. Once of my favorite references is the 1980 National Semiconductor Voltage Regulator Handbook.
Comments on Protection
Bob Dobkin: Circuit and system protection are one of the biggest reasons to choose certain types of regulators. If the regulator has good protection features you will find it on the datasheet. If the regulator has poor protection features it will be notably absent from the datasheet. So many companies now make low-cost linear regulators-All with the same part number-that you cannot fully trust the datasheet from secondary manufacturers.
A linear regulator is a thermal and electrical system. The thermal system must work properly for the regulator to be protected. Also the power transistor must be big enough to withstand safe area power dissipation. If a regulator is going to be used where protection is necessary and power dissipation is high, it is a good idea to short-circuit test the low-cost regulators.
Q re:Protection by DaeJ: In protection viewpoint, which one is better recommend for power electronic application circuit either Linear or Switcher?
A re:Protection by Bob Dobkin: For protection, I linear is much faster since the energy storage is less. With a long-term overload the linear will go into thermal limiting giving you a hotspot. The switcher will continue switching in overload and not dissipate much power
Comments on LDOs vs. Switchers
Greg Waterfall: One application we often see engineers choosing an LDO over a switcher are in processor applications were the core voltage needs an accurately regulated supply and has large load transients.
Newer LDOs can provide a very well regulated supply even during these large current transient. Most switchers are not fast enough for these <1us transients and must use more output capacitance to overcome this.
Bob Dobkin: That's a good point Greg. Linear's are fast.
Also, low-voltage Linear's can be almost as efficient as switchers. For example going from 1.2V to 0.9V, the linear is probably as efficient as a switching regulator.
Greg Waterfall: Another power saving application for LDOs is in standby circuits. Many larger systems have standby states to save power. These systems are often waiting for a simple user input, such as pushing a button on a remote.
During this time only a low power uC may be running. Most switch regulators have dramatically decreased efficiency at low output current because they consume power when they are switching. An LDO may be a better choice for these applications.
Q4 (Steve Taranovich):
@Greg—Tell us a little about your power filter capable of sourcing a 1-A load suitable for quiet supply solutions. It provides voltage regulation across the input and output terminals with high efficiency (low insertion loss), and power-supply rejection. I guess it is based upon the Linear Regulator?
A4 (Greg Waterfall)
Yes. Our power filter is designed to regulate the voltage from the input to the output instead of the output voltage to ground like a typical LDO. It is based mostly on a LDO design, but this slight change in topology allows us to optimize the PSR bandwidth.
In addition, since we are regulating the input/output voltage delta, this allows us to divide down the internal reference vs. gaining it up like a typical LDO giving us improved noise performance. As a supply rail the device provides one of the lowest noise spectrums available.
Q4a by eafpres1:
@Greg–so your device essentially tracks the input voltage?
A4a by Greg Waterfall:
Yes, but it also includes a “filter” capacitor to set a low frequency zero, that determines how fast you track the input. You can essential choose a capacitor that would prevent the power filter from tracking anything below 100Hz for example.
Q5 (Steve Taranovich)
Linear regulators in Energy scavenging
What are the key advantages of a Linear regulator in an energy scavenging design? An obvious one is low power, but what other characteristics are important?
A5 (Greg Waterfall)
One issue with switchers in energy scavenging application that are using vibration as the energy source the vibrations can induce voltage in the switching regulator's inductor, which may cause regulation and noise issues in the rest of the system.
Steve/Rob, I missed the chat, sorry for that.
The LDO you mention has only 9% efficiency hit, but it may be using the VIN from a switching regulator which might be having a efficiency of 80%. The overall efficiency w.r.t. to Battery is crunched?
All I wanted to know is getting rid of LDO to improve efficiency was okay, but to counter the problem of PSRR we are again bringing the LDO back. Getting rid of DC-DC is also not good as discussed earlier.
A6a (Greg Waterfall)
If the choice is between the DC-DC and the LDO, clearly with higher Vin and high output current, the DC-DC will win on efficiency. The optimized design for efficiency and noise is to use the DC-DC to convert the power rail to something close to the required system voltage and then use the LDO for the last 0.3V to 0.5V to get the benefits of low noise.
@Amrutah the efficiency of LDO is limited by quiescent current and I/O voltages,to get a high efficiency drop out voltages and current must be minimized, and voltages difference between input and output must also minimized.