Tuesday, May 8, 2012
Assuming the leakage current being drawn by the DUT, as well as any bypass capacitors on the fixture, have fully stabilized, the key thing with selecting the correct measurement integration time is getting an acceptable level of measurement repeatability. Some experimentation is useful in determining the minimum required amount of time. The primary problem with leakage current measurement is one of AC noise sources present in the test set up. With DC leakage current being just a few micro amps or less these noises are significant. Higher level currents can be usually measured much more quickly as the AC noises are relatively negligible in comparison. There are a variety of potential noise sources, including radiated and conducted from external sources, including the AC line, and internal noise sources, such as the AC ripple voltage from the DC source’s output. This is illustrated in Figure 1 below. Noise currents directly add to the DC leakage current while noise voltages become corresponding noise currents related by the DUT and test fixture load impedance.
Figure 1: Some noise sources affecting DUT current measurement time
Using a longer measurement time integrates out the peak-to-peak random deviations in the DC leakage current to provide a consistently more repeatable DC measurement result, but at the expense of increasing overall device test time. Measurement repeatability should be based on a statistical confidence level, which I will do into more detail further on. Using a measurement integration time of exactly one power line cycle (1 PLC) of 20 milliseconds (for 50 Hz) or 16.7 milliseconds (for 60 Hz) cancels out AC line frequency noises. Many times a default time of 100 milliseconds is used as it is an integer multiple of both 20 and 16.7 milliseconds. This is fine if overall DUT test time is relatively long but generally not acceptable when total test time is just a couple of seconds, as is the case with most components. As a minimum, setting the measurement integration time to 1 PLC is usually the prudent thing to do when short overall DUT test time is paramount.
Reducing leakage current test time below 1 PLC means reducing any AC line frequency noises to a sufficiently low level such that they are relatively negligible compared to higher frequency noises, like possibly the DC source’s wideband output ripple noise voltage and current. Proper grounding, shielding, and cancellation techniques can greatly reduce noise pickup. Paying attention to the choice and size of bypass capacitors used on the test fixture is also important. A larger-than-necessary bypass capacitor can increase measured noise current when the measuring is taking place before the capacitor, which is many times the case. Establishing the requirement minimum integration time is done by setting a setting an acceptable statistical confidence level and then running a trial with a large number of measurements plotted in a histogram to assure that they fall within this confidence level for a given measurement integration time. If they did not then the measurement integration time would need to be increased. As an example I ran a series of trials to determine what the acceptable minimum required integration time was for achieving 10% repeatability with 95% confidence for a 2 micro amp leakage current. AC line noises were relatively negligible. As shown in Figure 2, when a large series of measurements were taken and plotted in a histogram, 95% of the values fell within +/- 9.5% of the mean for a measurement integration time of 1.06 milliseconds.
Figure 2: 2 Leakage current measurement repeatability histogram example
Leakage current measurements by nature take longer to measure due to their extremely low levels. Careful attention to minimizing noise and establishing the minimum required measurement integration time contributes toward improving the test throughput of components that take just seconds to test.
Thursday, April 19, 2012
I settled on a high performance, switching-based DC source on having a low common mode noise characteristic of 10 mA p-p and 1 mA RMS over a 20 Hz to 20 MHz measurement bandwidth. To properly make this measurement the general consensus here is a wide band current probe and oscilloscope is the preferred solution for peak to peak noise, and a wide band current probe and wide band RMS voltmeter is the preferred solution for RMS noise. As the wide band RMS voltmeters are pretty scarce here I relied on the oscilloscope for both values for the time being. The advantages of current probes for this testing are they provide isolation and have very low insertion impedance.
I located group’s trusty active current probe and oscilloscope. The low signal level I intended to measure dictated using the most sensitive range providing 10 mA/div (with oscilloscope set to 10 mV/div).
One area of difficulty to anticipate with modern digital oscilloscopes is there are a lot of acquisition settings to contend with, all having a major impact on the actual reading. After sorting all of these out I finally got a base line reading with my DC source turned off, shown in Figure 1.
Figure 1: Common mode noise current base-line reading
My base-line reading presented a bit of a problem. With 1 mV corresponding to 1 mA my 2.5 mA p-p / 0.782 mA RMS base-line values were a bit high in comparison to my expected target values. It would be nicer for this noise floor to be at about 10X smaller so that I don’t have to really factor it out. Resorting to the old trick of looping the wire through the current probe 5 times gave me a 5X larger signal without changing the base-line noise floor. The oscilloscope was now displaying 2 mA /div, with 1 mV corresponding to 0.2 mA. In other words my base-line is now 0.5 mA p-p / 0.156 mA RMS. The penalty for doing this is of course more insertion impedance. Now I was all set to measure the actual common mode noise current. Figure 2 shows the common mode noise current measurement with the DC source on.
Figure 2: Common mode noise current measurement
Things to pay attention to include checking the current on both + and – leads individually to earth ground and load the output with an isolated load (i.e. a power resistor). Full load most often brings on worst case values. Based on the 0.2 conversion ratio I’m now seeing 8 mA p-p and 1.12 mA RMS, including the baseline noise. I am reasonably in the range of the expected values and having a credible measurement!
I decided to compare this approach to making a 50 ohm terminated direct connection. This set up is depicted in Figure 3 below.
Figure 3: 50 ohm terminated directly connected common mode noise current measurement
I knew insertion impedance was considerably more with this approach so I tried both 10 ohm and 100 ohm shunt values to see what kind of readings I would end up with. Table 1 summarizes the results for the directly connected measurement approach.
Table 1: 50 ohm terminated directly connected common mode noise current results
Clearly the common mode noise current results were nowhere near what I obtained with using a current probe, being much lower, and also highly dependent on the shunt resistor value. Why is that? Looking more closely at the results, the voltage values are relatively constant for both shunt resistor cases. Beyond a certain level of increasing shunt resistance the common mode noise behaves more as a voltage than a current. For this particular DC source the common mode voltage level is extremely low, just a few millivolts.
Not entirely content with the results I was getting I located a different high performance DC source that also incorporated switching topology. No actual specifications or supplemental characteristics had been given for it. When tested it exhibited considerably higher common mode noise than the first DC source. The results are shown in table 2 below.
Table 2: 50 ohm terminated directly connected common mode noise current results, 2nd DC source
With both voltage and current results changing for these two test conditions the common mode noise is exhibiting somewhere between being a noise current versus being a noise voltage. I had hoped to see what the results would be using the current probe but it seemed to have walked away when I needed it!
Making good common mode current noise measurements requires paying a lot of attention to the choice of equipment, equipment settings, test set up, and DUT operating conditions. I still have bit more to investigate but at least I have a much better understanding as to what matters. Maybe in a future posting I can provide what could be deemed as the “golden set up”! To get results that correlate reasonably with any stated values will likely require a set up that exhibits minimal insertion impedance across the entire frequency spectrum. Making directly coupled measurements without the use of a current probe will prove challenging except maybe for DC sources having rather high levels of common mode noise currents
The underlying concern here of course is what is what will be the impact to the DUT due to any common mode noise current from the test system’s DC source. Generally that is any common mode noise current ends up becoming differential mode noise voltage on the DUT’s power input due to impedance imbalances. But one thing I found from my testing is that the common mode noise is not purely a current with relatively unlimited compliance voltage but somewhere between being a noise voltage and noise current, depending on loading conditions. For the first DC source, with what appears to be only a few millivolts behind the current it is unlikely that it would create any issues for even the most sensitive DUTs. For the second DC source however, having 100’s of millivolts behind its current, could potentially lead to unwanted differential voltage noise on the DUT. Further investigation is in order!
Tuesday, December 6, 2011
In the first three parts of this post we looked at the topologies and merits of linear DC power supplies, traditional and high-performance switching DC power supplies, and common mode noise current considerations of each. So now in this final part we have reached a point where we can hopefully make an informed comparison and choice. Tables 1 and 2 summarize several key qualitative and quantitative aspects of all three DC power supply types, based on what we have learned.
Table 1: Qualitative comparison of DC power supply topologies
Table 2: Quantitative comparison of DC power supply topologies
So what DC power supply topology is the best choice for your next test system? In the past it usually ended up having to be a linear topology to meet performance requirements in most all but very high power, lower performance test situations. However, high-performance switching DC power supplies have nowadays for the most part closed the performance gap with linear DC power supplies. And, at higher power, the favorable choice may come down to selecting between several different switching DC power supplies only, due to their cost, size, and availability. So the answer is you need to make a choice based on how well the power supply meets your performance, space, and cost requirements, rather than basing the choice on its topology. Except for the most demanding low power test applications, like those needing the performance of a source measure unit (SMU), chances are much higher these days that the next DC power supply you select for your next test system you will be a switcher (and you possibly may not even realize it). What has been your experience?
Tuesday, November 29, 2011
Common mode noise current is a fact of life that manifests itself in many ways in test systems. There are several mechanisms that couple unwanted common mode noise currents into ground loops. An excellent overview on this is given in a two part post on the General Purpose Test Equipment (GPETE) blog “Ground Loops and Other Spurious Coupling Mechanisms and How to Prevent Them” (click here). However this is also an important consideration with our choice of a DC system power supply for testing as they are a source of common mode noise current. This is one area where linear DC power supplies still outperform switching DC power supplies. This can become a concern in some highly noise-sensitive test applications. As shown in Figure 1 the common mode noise current ICM is a noise signal that flows out of both output leads and returns through earth. By nature it is considered to be a current signal due to its relatively high associated impedance, ZCM.
Figure 1: Common Mode Noise Current and Path
Common mode noise current is often much greater in traditional switching DC power supplies. High voltage slewing (dv/dt) of the switching transistors capacitively couples through to the output, in extreme cases generating up to hundreds of milliamps pk-pk of high frequency current. In comparison, properly designed linear DC power supplies usually generate only microamps pk-pk of common mode noise current. It is worth noting even a linear DC power supply is still capable of generating several milliamps pk-pk of common mode noise current, if not properly designed. High-performance switching DC power supplies are much closer to the performance of a linear. They are designed to have low common mode noise current, typically just a few milliamps.
Common mode noise current can become a problem when it shows up as high frequency voltage spikes superimposed on the DC output voltage. This depends on the magnitude of current and imbalance in impedances in the path to the DUT. If large enough, this can become more troublesome than the differential mode noise voltage present. Generally, the microamp level of a linear DC power supply is negligible, while hundreds of milliamps from a traditional switching DC power supply may be cause for concern. Because common mode noise current is often misunderstood or overlooked, one may be left with a false impression that all switching DC power supplies are simply unsuitable for test, based on a bad experience with using one, not being aware that its high common mode noise current was actually the underlying cause.
In practice, at typical levels, common mode noise current often turns out not to be an issue. First, many applications are relatively insensitive to this noise. For example, equipment in telecommunications and digital information systems are powered by traditional switching DC power supplies in actual use and are reasonable immune to it. Second, where common mode noise current is more critical, the much lower levels from today’s high-performance switching DC power supplies makes it a non-issue in all but the most noise sensitive applications.
In those cases where common mode noise current proves to be a problem, as with some extremely sensitive analog circuitry, adding filtering can be a good solution. You can then take advantage of the benefits a switching DC power supply has to offer. A high-performance switching DC power supply having reasonably low common mode current can usually be made to work without much effort in extremely noise-sensitive applications, using appropriate filtering, capable of attenuating the high frequency content present in the common mode noise current. Such filtering can also prove effective on other high frequency noises, including AC line EMI and ground loop pickup. These other noises may be present regardless of the power supply topology.
Coming up next is the fourth and final part where we make our overall comparison and come to a conclusion on which power supply topology is best suited for test.
1. Taking The Mystery Out Of Switching-Power-Supply Noise Understanding the source of unspecified noise currents and how to measure them can save your sanity
By Craig Maier, Hewlett Packard Co. © 1991 Penton Publishing, Inc.
Wednesday, November 23, 2011
In part 1 we looked at the topology and merits of a linear DC power supply. To be fair we now have to give equal time to discuss the topology and merits of a switching DC system power supply, to make a more informed choice of what will better suit our needs for powering up and testing our devices.
Traditional switching DC power supply topology
The basic traditional switching power supply depicted in Figure 2 is a bit more complex compared to a linear power supply:
1. The AC line voltage is rectified and then filtered to provide an unregulated high voltage DC rail to power the following DC-to-DC inverter circuit.
2. Power transistors switching at 10’s to 100’s of kHz impose a high voltage, high frequency AC pulse waveform on the transformer primary (input).
3. The AC pulse voltage is scaled by the transformer turns ratio to a value consistent with the required DC output voltage.
4. This transformer secondary (output) AC voltage is rectified into a pulsed DC voltage.
5. An LC (inductor-capacitor) output filter averages the pulsed voltage into a continuous DC voltage at the power supply’s output.
6. As with a linear power supply, an error amplifier compares the DC output voltage against a reference to regulate the output at the desired setting.
7. A modulator circuit converts the error amplifier signal into a high frequency, pulse width modulated waveform to drive the switching power transistors.
Figure 2: Basic switching DC power supply circuit
In spite of being more complex the key thing is its much higher operating frequency, several orders of magnitude over that of a linear power supply, greatly reduces the size of the magnetic and filtering components. As a result traditional switching DC power supplies have some inherent advantages:
• High power conversion efficiency of typically 85%, relatively independent of output voltage setting.
• Small size and lightweight, especially at higher power.
• Cost effective, especially at higher power.
Traditional switching DC power supplies also have some typical disadvantages:
• High output noise and ripple voltage
• High common mode noise current
• Slow transient response to AC line and DC output load changes.
High-performance switching DC power supplies lessen the gap
Traditional switching DC power supply performance is largely a result of optimizing well established switching topologies for cost, efficiency and size, exactly the areas where linear DC power supplies suffer. Performance generally had been a secondary consideration for switching DC power supplies. However, things have now improved to better address the high-performance needs for electronics testing. Incorporating more advanced switching topologies, careful design, and better filtering, high-performance switching DC power supplies compare favorably with linear DC power supplies on most aspects, while still retaining most of the advantages of switchers.
So our choice on whether to use a linear or switching power supply has now gotten a bit more difficult! One area that still differentiates these DC power supply topologies is common mode current noise, worthy of its own discussion, which is exactly what I will do in part 3, coming up next!
Tuesday, November 15, 2011
To kick things off I thought it would be helpful to start with a short series of posts discussing something fundamental we’re often faced with; that is making the choice of whether to use a switching or linear DC power supply to power up our devices under test. In part 1 here I’ll begin my discussion with the topology and merits of linear DC power supplies, as I have heard countless times from others that only a linear power supply will do for their testing, principally due to its low output noise. Of course we do not want to take the chance of having power supply noise affect our devices’ test results. While I agree a linear DC power supply is bound to have very low noise, a well-designed switching DC power supply can have surprisingly good performance. So the choice may not be as simple anymore. The good thing here however is this may give us a lot more to choose from, something that may better meet our overall needs, including size and cost, among other things.
Linear DC Power Supply Topology
A linear DC power supply as depicted in Figure 1 is relatively simple in concept and in basic implementation:
- A transformer scales the AC line voltage to a value consistent with the required maximum DC output voltage level.
- The AC voltage is then rectified into DC voltage.
- Large electrolytic capacitors filter much of the AC ripple voltage superimposed on the unregulated DC voltage.
- Series-pass power transistors control the difference between the unregulated DC rail voltage and the regulated DC output voltage. There always needs to be some voltage across the series pass transistors for proper regulation.
- An error amplifier compares the output voltage to a reference voltage to regulate the output at the desired setting.
- Finally, an output filter capacitor further reduces AC output noise and ripple, and lowers output impedance, for a more ideal voltage source characteristic.
Figure 1: Basic Linear DC Power Supply Topology
Linear DC power supply design is well established with only incremental gains now being made in efficiency and thermal management, for the most part. Its straightforward configuration, properly implemented, has some inherent advantages:
- Fast output transient response to AC line and output load changes
- Low output noise and ripple voltage, and primarily having low frequency spectral content
- Very low common mode noise current
- Cost competitive at lower output power levels (under about 500 watts)
It also has a few inherent disadvantages:
- Low power efficiency, typically no better than 60% at full output voltage and decreases with lower output voltage settings
- Relatively large physical size and weight
- High cost at higher power (above about 500 watts)
So it sounds like a linear power supply has to be the hands-down winner especially for low power applications. Or not? To make a more-informed choice we need to look at the topology and merits of a switching power supply, which I will be doing in part 2!