Using the power supply status subsystem to improve your test throughput

Continuing on my throughput theme here, one recommendation is to take advantage of the power supply’s status subsystem. Some power supply operations take notably longer than most to complete than others. Two notable examples:

• Initializing a triggered measurement
• Initializing a triggered output transient or output list event

When developing programs you can include long, fixed wait statements to make certain these operations have completed before proceeding. However, this can easily add many tens of milliseconds or more of unnecessary waiting, increasing overall test time.  A better way is to take advantage of the DC power supply’s status subsystem features that eliminate unnecessary waiting for these operations.

Triggered measurement and output sourcing events can substantially speed up testing by providing actions tightly synchronized with other test activities. But they do have some up-front set up overhead time needed for initializing them. Instead of using a fixed programming delay following an initialization operation it is better to take advantage of the Operation Status Group register in the status subsystem, which is illustrated in Figure 1.

Figure 1: Agilent N6700 series DC power system operation status group

The “WTG meas” bit (#3) or “WTG trans bit (#4) in the condition register can be monitored with a loop in the test program to see when they turn true. At the moment the measurement or output sourcing event is initiated and ready for a trigger the test program will then proceed with its execution without incurring any unnecessary additional waiting. This saves a considerable amount of time as illustrated in Figure 2.

Figure 2: Operation-complete wait time distribution

Instead of waiting for the full worst-case each and every time, the wait is now just the actual time. When repeated over and over for all DUTs being tested, the net result is the average of the actual wait time, which in most cases is just a small fraction of the worst case time! The net result can be many tens of milliseconds test time savings, making an improvement in test throughput.

The first five hints of my compendium “10 Hints for Improving Throughput with your Power Supply” can be viewed here: (click here to access).  For those reading our “Watt’s Up?” blog here are getting the opportunity to preview one of the remaining 5 hints yet to be released!

How fundamental features of power supplies impact your test throughput – Part 2

In part 1 of” How fundamental features of DC power supplies impact your test throughput” (click here to access) I shared definitions of some of the fundamental power supply features that impact test throughput, including:

• Command processing time
• Up-programming response time
• Down-programming response time

Another fundamental DC power supply feature impacting test throughput is its measurement time. There are actually two aspects to a DC power supply’s measurement time as depicted in Figure 1:

• Measurement settling time
• Measurement integration time

Figure 1: DC power supply measurement time

A good indicator of a DC power supply having a high performance measurement system is having programmable measurement integration time, or aperture time, often programmed in power line cycles (PLCs).  One reason for having a programmable integration time is for minimizing any 50 or 60 Hz AC line ripple getting into the DC measurement, by setting the time one or more multiples of a PLC.  Setting the time to 1 PLC provides good ripple rejection with relatively good throughput. When AC line ripple is not an issue the integration time can be set even smaller than 1 PLC, further reducing measurement time. When the DC power supply has a programmable measurement integration time it will no doubt also have a fast-responding measurement system as well, typically just milliseconds, to complement the higher achievable throughput with programmable measurement integration time.

In comparison basic DC power supplies commonly use a 100 millisecond fixed integration time to support AC ripple rejection for both 50 and 60 Hz line frequencies. They also have low bandwidth, slow-responding measurement systems, which can long time to settle after any step change in loading, before a valid measurement can be taken.

We have just introduced our Advanced Power System (APS) DC power supplies. This is a family of high-performance, high power (1 and 2 kW) DC power supplies designed to address the most demanding test challenges. These fundamental throughput-related features for APS are typically more than two orders of magnitude faster compared to more basic-performance DC power supplies, providing much better throughput in manufacturing test. A colleague of mine recently posted details of their introduction on his “General Purpose Electronic Test Equipment (GEPETE)” blog (click here to access) which I believe you will find of interest. Included in this introduction is a link on throughput that takes you to a series of application briefs I have written that go into more detail on improving test throughput with the DC power supply, which you may find very useful.

So how much test throughput improvement might you expect to see by switching from a basic-performance DC source to a high-performance DC source? Well, it really depends on how much the testing makes use of the DC power supply. If it only uses the power supply to provide a fixed DC bias to the device under test (DUT) that never changes for the duration of the test then it will not make a significant difference. More often than not however, a DUT is tested at several bias voltages with several current drain measurements taken for the various bias voltage settings and DUT operating modes. This can add up to a considerable amount of test time. In this case a high-performance DC power supply can more than pay for itself many times over due to improved test throughput.  To get an idea of the kind of difference a high-performance DC power supply can make I set up a representative benchmark test It compares the throughput performance one of our new APS DC power supplies to that of a more basic-performance power supply.  If you are interested in finding out how much difference it made, I made a video of this benchmark testing, entitled “Increasing Test Throughput with Advanced Power System” (click here to access). All I am going to say here is it is an impressive difference but you will need to watch the video to see how much difference!

How fundamental features of power supplies impact your test throughput – Part 1

When it comes to manufacturing of electronic products, reducing test time to improve throughput is virtually always a top priority, because “time is money” as the old saying goes! Usually most all of the attention may be placed on reducing the test time of the banner aspects of the product, such as the RF performance of a wireless device, for example. However, the choice of the DC system power supply can also have a huge impact on your test time and throughput during manufacturing. You may find the lowest cost, more basic-performance DC power supply that meets your immediate needs end up costing you the difference in price many, many times over of that of a higher-performance DC power supply having better throughput performance in the long run!

The DC power supply can incorporate a number of advanced features, such as elaborate triggering and sequencing systems, which will allow you restructure your testing to optimize throughput. However, even fundamental throughput-related features of the power supply can also have a large impact on your test time, including:

• Command processing time
• Output up-programming time
• Output down-programming time
• Measurement time

Figure 1 illustrates what the command processing and up-programming times are for a DC power supply. The command processing time is the time from when the command is first received to the point where the power supply starts acting on it. In this case it is when power supply’s output starts to change. The up-programming response time is the time the power supply takes for the output to rise and settle within a small band around the final output level, after processing the command instructing it to change its output level.

Figure 1: Power supply command processing and up-programming response times

The down-programming response time is like the up-programming response time except that the power supply is instead being programmed to a lower level. However, you need to look at down-programming independently as short up-programming time does not necessarily guarantee comparably short down-programming time. More basic performance DC power supplies usually lack an active down-programmer circuit that quickly brings down the output. In this case the down-programming response time can be very dependent on how much load the DUT presents to the power supply’s output.

How much difference is there in performance between more basic performance and higher performance DC power supplies on these throughput-related features? It can be considerable; over several orders of magnitude difference. As one example, command processing time can range from up to 100’s of milliseconds for entry-level power supplies to under 1 millisecond for high performance power supplies.

Another fundamental throughput-related feature of a DC power supply is its measurement time. There are a couple of aspects to consider here as well, which I will elaborate on in part 2 of this series on how fundamental features of power supplies impact your test throughput, in an upcoming posting here on “Watt’s Up?” along with tying it all together to show how they affect actual test throughput!

What is Command Processing Time?

Hello everybody,

We have a new intern here at Agilent Power & Energy HQ named Patrick.  Gary, Patrick, and I have been having a philosophical debate on what the term command processing time means.  This is a very important number for many of our customers since it tells them what kind of throughput they can get out of our test equipment.  A fast command processing time allows you to reduce your test times and therefore increase your throughput.  The question that we have been debating is:  what is command processing time and how can we measure it?  We have been discussing three scenarios.   Let’s go through them.

The first option is the amount of time that it takes the processor to take one command off the bus so that it can get to the next command.  This tells you how quickly you can send commands to the instrument.  The only issue with this is that some instruments have a buffer so it is not actually “processing” the command, just bringing it into the buffer and letting you send the next command.  Obviously this is useful but it really does not address the throughput question.  This is pretty easy to test by sending a command in a loop and timing it.  You record the time before the command is sent and the time after the loop and then divide by the number of loops you executed.  This would yield a pretty good approximation of the time.

The second option is the amount of time from when the instrument receives a command until it starts performing the action.  I believe that this is what we list in our manuals for the Command Processing Time Supplemental Characteristic.  This does address the throughput issue.  This is also easy for us at Agilent to measure.  We have a breakout for GPIB that allows us to monitor the attention line.  The test that we did was send a VOLT 5 to the instrument.  We looked at the GPIB attention line.  The time from when the attention line toggles until the power supply starts slewing the voltage up would be our command processing time (measured with an always awesome Agilent Oscilloscope).  This is what I consider to be the command processing time.

The third option includes what I spoke about in the last paragraph but also includes the slewing of the voltage.  The processing time would be the time that it takes to take the command and complete all the actions associated with it (for example settling at 5 volts after being sent a VOLT 5 command).  I do not think that this is a bad option but we have a Supplemental Characteristic for voltage rise time that addresses the slewing of the voltage.   The test method would be the same as above using an oscilloscope but watching for where the voltage settles at five volts.

What do you, our readers and customers think the correct interpretation of command processing time is?  Also, please stay tuned for a future installment where we try to figure out what the quickest interface is: LAN, USB, or GPIB.