The need for power supply protection systems
A robust bench power supply like the QPX750SP with a wider than normal output power range enables test, development, and troubleshooting of systems that would overload, shut down, or damage standard bench power supplies. The QPX750SP can easily deliver 50 A of current at 12 VDC (600 W of DC power) and up to 9.4 A at a caution worthy 80 VDC. With a wider range of operational capability (750 W maximum output power) comes a wider ranger of hazards. 750 W of misapplied power can generate burn or even ignition levels of heat, pose a shock hazard, and destroy electrical and electronic components instantaneously. Borrowing from Spider-Man pop culture; with great power comes great responsibility. As long as the user always operates the supply correctly there is little reason to worry about hazards. However, my experience as a polytechnic educator, as a professional engineering technologist, and as a hobbyist, has taught me that it is foolish to assume any user will ever operate anything correctly. Even the most experienced and careful user makes assumptions that turn out to be incorrect, makes connections that turn out to be hazardous, or just screws up from time to time. For these reasons, any instrument capable of delivering high currents or high voltages needs to be designed with safety systems that shut off outputs if unexpected load currents or voltages occur during use.
The Aim-TTi QPX750SP has several safety systems designed to protect connected loads from damage due to excess voltage or current and to protect the instrument from out-of-spec stresses. I have tested a few of them under controlled conditions. This blog contains a summary of my findings related to two types of protection provided by the QPX750SP: over voltage protection (OVP) and over current protection (OCP). In general, I can say I found the OVP and OCP safety systems behaved very well from a hardware engineering point of view. That is, the safety systems generally activated as expected based on user provided trip point settings. What could be improved are the interfaces used to interact with the safety systems.
Current Protection Methods
There are two current protection methods available in the QPX750SP. Beyond knowing what each method of protection does, it is essential to read the user manual to know how to get the current protection behavior you are looking for. The choice of protection method depends on what sort of protection behavior the user needs. The two methods are outlined below, followed by my findings from bench testing each method.
The first line of current defense is provided by the Iset control. The user enters a numeric value for Iset via one of several methods (described in another blog post). The Iset value provided by the user establishes the upper limit of current the QPX750SP will allow into the load. In general use, the QPX750SP will provide the DC voltage set by the user via the Vset control at any current up to the Iset value.. If the load demands more than Iset amps of current, the QPX750SP will hold the output current to Iset and limit the output voltage to hold output current flow to Iset amps and no more. In this operating mode the output stays active during foldback; it does not shut off.
Over Current Protection (OCP)
The second type of current defense is provided by the Over Current Protection feature. This method shuts down the output of the supply when load current greater than the OCP set point is detected. Output voltage rapidly drops to 0 V and current flow to the load drops to 0 A. This protection method is more drastic than foldback current limiting in that the supply shuts off power to the load. To make OCP work, the OCP trip point must be set to a value less than Iset. Since foldback kicks in when load current exceeds Iset, OCP will never activate if it is set higher than Iset.
Exploration of QPX750SP current limiting protection
To generate a controlled load for the QPX750SP I connected it in a series circuit with a digitizing ammeter and a precision electronic load. A Keysight Source Measure Unit provided a precision programmable load. Current flow through the was digitized and logged by a Keysight 7.5 digit multimeter. QPX750SP output voltage was captured by a Tektronix MDO4104-3 oscilloscope. To explore current limiting the QPX750SP was configured as follows.
|Vset (output voltage)||5.000 VDC|
|Iset (output current)||2.20 ADC|
Output current setting was constrained by the sink capability of the . I hope it is safe to assume that results obtained at these lower current settings can be considered indicative of instrument behavior across the entire range of instrument capability.
With an Iset value of 2.20 ADC, any load that demands more than 2.20 ADC should cause the supply to enter current limiting. This means load current should not increase above Iset, even when the load request more current. When the supply enters current limiting mode output voltage should decrease to keep output current constant at 2.20 A. Note that OCP is set above Iset to prevent OCP from tripping the supply into a shutdown state. Remote sense makes use of the sense input terminals on the front panel. Wires were connected from the sense terminals to the output terminals of the QPX750SP before the ammeter.
The Keysight Precision Source Measure Unit was configured to produce a current sink linear ramp going from 0 A to -3.00 A then back to 0 A in 10 mA steps lasting 25.00 ms/step. A ramp with those specifications should activate current limiting during the portion of the ramp that exceeds 2.20 ADC. The will attempt to draw current above 2.20 A up to 3.00 A stepping 10 mA at a time. That is 800 mA over the Iset value (80 steps). At 25 ms/step there should be 2 s of limiting on the increasing portion of the ramp and another 2 s of limiting on the decreasing portion of the ramp. The QPX750SP should exit current limiting when load current goes below 2.20 A.
The ramp test was captured with a digitizing ammeter and an oscilloscope, and with Aim-TTi's Test Bridge application. A screen shot of the Test Bridge result is shown below.
Note that the minimum sample interval in Test Bridge is 250 ms. This limits the resolution of the strip chart, nevertheless, the general shape of the current limit behavior is clearly visible. The green trace in Graph 2 represents load current. The SMU attempted to ramp up to 3A of load current draw, but the QPX750SP limited the load current at 2.2 A. To do so it dropped output voltage well below 5 VDC as seen in the red trace in Graph 1.
Test Bridge allows data logs to be stored in a (rather odd) tab separated value (TSV) format which can be imported into Excel by going through a few configuration steps. An Excel generated graph of the current limiting test derived from the TSV file is shown below.
Relatively low sample rates and limited sample resolution cause the current slopes to be rather irregular. Output voltage drops to 1.1 V during foldback according to the samples recorded by Test Bridge. During limiting current holds at about 2.2 A. These two values can be used with Ohm's law (R = V/I) to calculate the load resistance during limiting. With these values of voltage and current, load resistance calculates as 50 mOhms. The QPX750SP calculates and displays load resistance. I wrote down the displayed load resistance during the 4 s of current limiting action. It was 0.5034 Ohms. For a closer look at limiting behavior take a look at the digitized Keysight ammeter log. The was set to sample at a 1 kHz rate for 18 s with a trigger on positive slope at 10.0 mA of current flow.
With a 1 ms sample rate much more detail is captured. The current ramp is revealed to be linear going up and down. Also, greater detail emerges at the entry and exits points for current limiting. In this particular test current limiting kicks in rapidly at 2.32 A. In terms of time delay, this represents about 300 ms from the point where current demand exceeded the Iset value of 2.2 A. I could not find a specification for activation time for switching from CV to CC mode, so I'll treat this as a discovered specification subject to further refinement with additional testing. During current limiting output is held to an average of 2.2 A as set by the Iset value on the supply. During limiting the supply switches from CV mode to CC mode. In this current protection method the supply does not switch off. Current continues to flow through the circuit to the load, however output voltage drops as necessary to hold current flow to the Iset limit. A screen capture of the oscilloscope trace of output voltage is shown below.
Of note in the image above are the steep edges upon entry into and exit out of foldback current limiting in contrast to the the more shallow slopes suggested by the Test Bridge results. The width of the limiting action is measured by the oscilloscope to be 4.244 s, which is close to the 4 s estimated previously. Also, the output voltage drops to 1.032 V as measured by the oscilloscope. Running the oscilloscope measured output voltage and the digitized ammeter current values through Ohm's law gives a calculated load resistance of about 47 mOhms during limiting, which is reasonably close to the 50 mOhm value estimated earlier.
Two take aways from this experiment:
- Current limiting works as expected in the QPX750SP.
- Test Bridge can be used to trace voltage and current behavior, but don't expect high resolution on the time or magnitude axes.
Current limiting works well in situations where the user wants to limit, but not shut off, current flow to a load. In situations where excess current flow may cause damage to a circuit, it may be better to use a protection method that does shut off the output of the power supply, such as Over Current Protection. In the next section I take a look at how the QPX750SP OCP feature performs.
Exploration of QPX750SP Over Current Protection (OCP)
To explore OCP behavior I used the same test equipment set up as described for the Current Limiting test. In this case I set up the to produce a current pulse rather than a ramp. My reasoning for using a pulse emerges from my experience of circuit faults that tend to draw a lot of current. Many are caused by the user accidentally shorting traces with a slip of a test lead, or setting up circuit configurations that result in heaps of unintended current flow. Faults of this nature tend to cause sudden pulse-like increases in current flow.
For an example from the slipped-test-lead category I recall one expensive day from my teaching career where a class of electronics students were taking measurements in a power amplifier. They were measuring the output signal from an audio power amplifier IC that was constructed with fairly tight pin pitch. The output pin on the power amplifier package was adjacent to the VCC supply rail. In a matter of one half hour I heard the distinctive "pop" sound made by a 'scope probe shorting the output pin to the VCC pin, followed immediately by a wisp of smoke and a sheepish student approaching me to confess they needed a new amplifier to complete the lab. This all happened in spite of a pre-lab lesson on the importance of using caution around high voltage by attaching probes with power off, then powering up to make a measurement, then powering down to remove the probe.
For an example from the "heaps of unintended current flow" category I recall a situation where I was developing a motor controller using an IC based H-bridge. One coding error resulted in the H-bridge being configured in a way that allowed Niagara Falls worth of current to flow through the IC. In a matter of seconds I saw the wire frame inside the chip glow red hot with attendant smoke pouring forth along with tar like black plastic dripping off the package. Very alarming but also super impressive.
So, for this experiment I set up a controlled current draw of 3 A in the shape of a 500 ms pulse. The QPX750SP was configured as follows.
|Vset (output voltage)||5.000 VDC|
|Iset (output current)||2.50 ADC|
An OCP value of 2.20 ADC combined with an Iset value of 2.50 ADC may seem counterintuitive. This seems to suggest that I want to allow up to 2.5 A of current to flow in the load, but to also shut down if more than 2.2 A of current flows into the load. In this case I do want the supply to shut down if more than 2.2 A of current flows into the load and I do not want current foldback to activate. I refer to a Note from the QPX750SP operation manual.
When I tried this test a few things happened that drew my attention. In general, the test worked, in that the supply did shut off the output and it did report an OCP trip. However, the OCP trip was only reported locally on the supply display, not on the Test Bridge application, and the shut down behavior was a combination of OCP trip and current limiting. Let me explain.
The images below show the QPX750SP front panel after the 3 A current draw pulse was applied and the Test bridge data log of the event.
In the image above, the QPX750SP has detected an OCP event and shut off the output terminals. The displayed voltages and current are grey, meaning they represent the set values, not the output values. The red OCP message beside the current value indicates an Over Current Protection event has caused the output to be shut down. The image below shows the Test Bridge state after the OCP event. Notice that the voltage and current readings are grayed out, but there is no indication that an OCP event has occurred. Also note the strip chart data logs of voltage and current. They clearly do not appear to be accurate. Voltage starts to drop as current starts to rise. Maximum current peaks at a value slightly above 600 mA.
Now let us take a look at what the digitizing ammeter and oscilloscope captured during this event.
Load current snaps up to 3 A when the SMU load pulse starts. Because this level of load current exceeds the OCP trip point (which was set to 2.20 A), the QPX750SP determines, after about 35 ms, that a trip condition has occurred. Next, it appears the supply enters a current limiting mode where current is limited briefly to the Iset value of 2.50 A. Then, about 250 ms into the event, the QPX750SP disconnects the output and load current drops to 0 A. I was not expecting the brief journey through limiting on the way to total output cutoff. Upon reflection, I don't think this is an issue. If Iset is at an acceptable level for the circuit, then dwelling briefly at this level before turning off the output seems acceptable. I ran the test several times in a row, capturing the voltage behavior on an oscilloscope. The result of four consecutive trials is shown below.
Over four consecutive trials the time the QPX750SP needed to recognize an OCP violation stayed the same at about 35 ms. The specification says response time on OCP events is typically 100 ms with a 100 mA resolution and 0.5% +/-0.2 A accuracy. In all four trials the QPX750SP appears to journey through current limiting before shutting off completely. The amount of time spent in current limiting appears to be variable as seen by the four different drop points to 0 V output. The drop points occurred at 74 ms, 84 ms, 191 ms and 252 ms as measured with cursors on the oscilloscope.
My take-aways from this test are:
- OCP works on the QPX750SP. The output is shut off when an OCP event is detected.
- As noted in the current limit tests, Test Bridge can be used to trace voltage and current behavior, but don't expect high resolution on the time or magnitude axes.
- The QPX750SP dwells briefly in current limiting before shutting off the output.
What happens if the current pulse is made shorter, say down to 50 ms? This would have the high current event end sometime during the limiting portion of the shut down sequence. The answer is that odd things happen. First, though the output is indeed disconnected, the front panel of the QPX750SP reports an Over Voltage Protection (OVP) event rather than an OCP event, and this time, Test Bridge reports an OCP event. That is odd. The oscilloscope trace shows why an OVP may have been reported..
When the over current event ends at 50 ms, the output voltage spikes up beyond the 6.00 V OVP trip point, thereby activating an OVP event shutdown. When I changed the OVP trip point to 10 V and initiated a 3 A load pulse, even stranger things happened. The voltage trace shows that an OCP trip event is initiated, but at 50 ms when the pulse ends, the supply changes its mind and returns to normal output voltage, as if nothing happened at all. It seems it is necessary for the QPX750SP to dwell in current limiting for awhile to confirm the validity of an OCP trip event before it commits to shutting off the output.
You may have anticipated my next test. I reduced the load current pulse to 25 ms, a duration that situates the pulse less than the observed detection time of the OCP system. Indeed, as might be expected, nothing happened at all. The pulse does not seems to have been detected. The ammeter captured a 25 ms pulse of load current at 3 A, but the voltage trace remained stable, apart from brief inductive glitches at the leading and trailing edges of the pulse. The lesson here is that OCP events have to be long enough to fully activate the protection circuits.
Over Voltage Protection (OVP)
Turning attention now to the OVP behavior of the QPX750SP, my plan was to use load current pulses to create transient voltage spikes on the output of the supply in accordance with the equation vL=−L di . Parasitic inductance in the wiring (L) combined with fast edges (low dt) and moderately high currents (high di) should generate voltage transients (vL) on the leading and trailing edge of the load pulse. Negative transients will be generated on the leading edge of the current pulse with positive transients generated on the trailing edge. My first experiment will be to determine if the transients are of sufficient magnitude to trip the OVP circuits in the supply. A trial run with no added inductance in the leads produced the following result.
In the image above you can see where the oscilloscope triggered on the negative going transient at the start of the load pulse. 500 ms later there is a positive going transient on the trailing edge of the load pulse. The positive going pulse is the one that I hope to use to trip the OVP circuit. The zoomed portion of the trace on the bottom shows that the positive transient has a peak value of 567.3 mV with no additional inductance in the circuit. This isn't huge, but high enough to explore a few OVP trip points between 5.1 V and 5.6 V. The specification for the OVP system stated in the operating manual is given as:
I will be looking for noticeable changes in trip point with 100 mV increments of the OVP setting. I don't expect the trip points to be exactly 100 mV apart because of the 0.2% +/- 0.2 V accuracy spec. I will also be looking to see if OVP engages within 100 us of a violation. The test set up for the trail runs will exclude the digitizing ammeter. The SMU will provide programmable load current pulses and the Tektronix MDO4104-3 oscilloscope will capture output voltage behavior. The QPX750SP will be configured as follows.
|Vset (output voltage)||5.000 VDC|
|Iset (output current)||3.50 ADC|
The result of running 3 A pulses through the QPX750SP at five different OVP settings is shown below. Measurement data and analysis is below the oscilloscope image.
|OVP voltage (V)||Peak V over 5.0 VDC (mV)||Time to activate clipping (us)||Time to initiate shut down (us)|
|5.6||588||330||no shut down|
The oscilloscope image shows four instances where OVP was activated and one instance where OVP was not activated. The instance where OVP was not activated shows the transient voltage rise then fall back to a steady state DC voltage. This instance was captured when OVP was set to 5.6 VDC. Any OVP voltage setting I entered above 5.6 VDC resulted in the same trace behavior (I went as high as OVP = 10 V). The peak voltage hit 5.588 VDC on this trial run, but it did not trip the OVP circuit. All of the remaining instances did record OVP activation. This can be seen in the transients falling at a steep slope off the bottom of the screen. The lowest amplitude trace represents an OVP setting of 5.1 VDC. followed in increasing amplitude by 5.3, 5.4, and 5.5 VDC OVP trip points. From these trials I can see that 100 mV changes in OVP trip points do make a measurable difference in OVP behavior, though with a 5 VDC output, the difference between 5.3, 5.4, and 5.5 V OVP settings seems to be more like 40 to 44 mV rather than 100 mV. The more difficult behavior to assess is response time. The shape of the transient pulses suggests to me that there may be a short interval of clamping or clipping followed by shut down of the output. I measured the time from the rising edge of the transient to the point where what might be clipping seems to occur (where the transient flattens before falling). This response time varied from about 100 us to 340 us. The actual output shut down occurred between 320 us and 580 us after the start of the transient. The specification declares response time is typically 100us. Is response time the time required to notice an OVP condition and clamp the output, or is it the time to shut down the output after detection an OVP condition? Either way, when tested under the conditions described here the only instance that met a 100 us criteria was the time to clamp/clip the transient when OVP was set to 5.1 VDC.
What I can say is that OVP works in that the output does shut down when terminal voltage exceeds a set point. I can also say that 100 mV changes in OVP set point so make small changes in OVP trip behavior. Time to respond by shutting off the output increases as OVP set point increases, at least in this limited set of trials. There is a bug in the Test Bridge application that I noticed during these trials. Anytime the QPX750SP reported an OVP trip on its front panel, Test Bridge reported an OCP trip for the same event.
Protection systems are very nice to have in any power supply. They are very much necessary in a high power instrument like the QPX750SP. My explorations have confirmed for me that the OVP and OCP systems work as expected in the QPX750SP, although there is some wiggle room in interpretation of the performance specifications. I must emphasize that merely having good protection systems in a power supply does not mean the user can use the instrument free of care. The protection systems are not artificially intelligent (yet). They still require the user to activate them and input parameters that are appropriate for each use case.