My Account
Shopping Cart 0 item(s)
View Cart
Facebook Google Plus LinkedIn

Testing a Selenium Rectifier

Posted by Peter Hiscocks on 10/25/2016

Before the invention of the silicon diode, designers had the choice of germanium point contact diodes for low currents, copper-oxide rectifiers in instruments, and selenium rectifiers for power applications. The selenium rectifier replaced the vacuum tube rectifier in the period 1950 onward and was in use until the 70's. Selenium rectifiers are recognizable by the cooling fins, which are often painted green or orange.

We were given a three-phase selenium rectifier, taken from an ancient Honda 150 motorcycle. The alternator is wired to produce 3 sine wave outputs, each 120 degrees apart. The rectifier consists of three diodes with one common terminal, so that the output is a reasonable approximation of direct current.

The owner of the bike had replaced the selenium rectifier with a modern three-phase silicon diode array, a much smaller and probably more efficient device.  Was the selenium rectifier still good?  We measured the forward and reverse resistance with a multimeter, and the results were inconclusive.

The Wikipedia page on Selenium Rectifier mentions that a 'forming current' may be required after a long period of disuse, and long period of disuse applies in this case: it had been out of service 40 years or more.  The forming current is a minimum current to regenerate the proper operation. So we connected it to a CTR-101 curve tracer to run the forward characteristic curve. 

Unlike a multimeter test, the curve tracer exercises the device under test over a range of voltage and currents, so it's a much more informative measurement.

As you can see from the graphs, the forward characteristic is a classic diode curve with a threshold about 0.2 volts, increasing to 0.8 volts at 1 amp forward current.  Switching on the measurement cursors, the forward resistance over the linear region is 0.37 ohms.

The reverse characteristic shows a reverse current of 800uA or so up to a reverse voltage of 30 volts, also very respectable.

This is the measurement of one of the three diodes: the other two have a similar characteristic.

So yes, the selenium rectifier is still functional. And the Honda 150 engine also runs just fine, and is being used to drive a home-constructed sawmill.

Yes, we test IGBTs

Posted by Peter Hiscocks on 10/6/2016 to Software
We were asked the other day whether our curve tracer CTR-101 can test IGBT devices.

What's an Insulated Gate Bipolar Tranistor (IGBT)? If you live in a world of small analog signals (that's us), then you probably haven't used an IGBT. But if you work in high power applications such as variable speed drives for induction motors or welding controllers, you've probably used them.

The IGBT is one of those hybrid devices that is 'the best of both worlds'. It combines the high input impedance of the MOSFET with the low saturation voltage of the bipolar transistor.  The IGBT is used almost exclusively as a switching device, so in the ON state, the power dissipation is proportional to the saturation voltage.  Smaller saturation voltage results in less power dissipation and simplified cooling requirements.

We ordered some of the IGBT model IRG4PF50 to test on the curve tracer. The specifications are impressive: 900 volt breakdown voltage, 51 amps current, 200 watts, all for $6 from Digikey.  Even allowing for the limitations of heatsinks and thermal resistance (which always make the actual power less than shown on the spec sheet), this is an impressive device.

The CTR-101 can measure up to 30 volts or so at a test current of 1 ampere: it's intended for small signal devices.  But we can test the behaviour in the small signal region.  This would be useful, for example, if you needed to match one or more units.

The figure shows the results: on the N-MOSFET setting of the curve tracer, we get a family of curves similar to an enhancement mode MOSFET.  There is certainly enough information here to determine the gate threshold voltage and match device characteristics.  So yes, the CTR-101 can measure the characteristics of an IGBT.

A transfer characteristic (drain current vs gate-source voltage) might also  e useful: if you have some interest in that, give us a shout and we'll add that feature to the software.

Peter Hiscocks
October 2016


Negative Tracking Power Supply

Posted by Peter Hiscocks on 6/27/2016 to Technical Papers
It is not uncommon for a circuit to have available a positive power supply, but require a negative supply. For example, using op-amps to process an alternating voltage signal (one that goes both positive and negative) may require a negative power supply. 

There are charge pump integrated circuits for this purpose: a switching circuit charges discharges a group of capacitors such that the output capacitor is charged negative with respect to ground.  This is useful for small currents, in the order of a few milliamps, which may be sufficient for powering an op-amp or two.  But charge pumps are very limited in output current capability - and some have high internal resistance, which results in poor regulation.

For higher currents, the energy storage device must be an inductor. Current is dumped from the positive supply through a switch into the inductor. Current ramps up in the inductor, storing magnetic energy.

Then the switch opens. Current cannot change instantaneously in an inductor, so the current continues in the same fashion. A diode directs this currentinto a conductor into a capacitor, in such a manner that the capacitor charges up to a negative voltage.

There are many switching regulator integrated circuits that can be configured to generate a negative voltage in this fashion. Such a circuit can produce a few hundred milliamperes of current.

I needed a negative voltage supply that would track the positive supply over a range of 0 to +15 volts, so that the negative supply would produce 0 to -15 volts.  The target output current was 150mA.  Using the circuit of reference (1) as a starting point, I developed the circuit shown in figure 1.  It turns out that the CMOS version of the venerable 555 timer, the LMC555, can operate down over a supply range of +1 to +15 volts.  The output drives a PNP switching transistor, which pumps the 150uH inductor.  When Q1 switches off, the top end of the inductor goes negative, charging the output 10F capacitor to a negative voltage.

The two 100k resistors form a voltage divider between the two supplies. For tracking to occur, this voltage should be at zero volts, so it is used as a feedback signal through Q2 and Q3 to the control terminal of the timer IC.

It's really useful to be able to monitor the current in the inductor but there is no easy way to do that directly. I put a 1 ohm resistor in the return line to the input supply and monitored the voltage across that with the oscilloscope. That shows the current in the inductor when it is charging, so it's possible to determine the peak current -- which is the quantity of most interest.

It turns out (see reference 2) that the peak current in the inductor and switching transistor is at least 4 times the average output current. So, for 150mA output current, the switch must be able to handle 600mA. This is a consequence of the inductor waveform, which is a triangle. I decided that a circuit based on square waves would have a more favourable peak-average ratio, and so I did not pursue this circuit any further. But it might be useful for moderate output currents where a tracking supply is useful.

A caveat: the circuit of figure  is not fully debugged. It needs to be tested under a range of currents and voltages. It's likely that under certain conditions the system will be unstable and will require compensation. 

But it might be a starting point.


References

1. 555 as switching regulator supplies negative voltage	
S.L.Black							
http://www.epanorama.net/sff/Power

2. Switching Power Supply Design, 2nd Edition		
Abraham I. Pressman						
McGraw Hill

Analog Design Not Dead

Posted by Peter Hiscocks on 6/27/2016 to Resources
I'm not much of a party person but there are occasions when one's presence is required - and some parties can even be fun if you keep an open mind about it.  (Just beware of an open bar on an empty stomach, that can be deadly.) At one party, I asked one of the guests what he did: 'I'm an astrophysicist' he said.  I guess I looked astonished because he added'Well, someone has to do it.' True enough.  (See the movie Star Men for a nice take on that.)

Anyway, if someone asks me at a party what kind of electronics I do I may say 'analog circuit design', and then the response might be 'I thought analog had been replaced by digital'.  Well, no.

Those comments have come back to me as I've been sweating over various analog circuits for the last few weeks.  The PSM-101 power supply has a major analog circuit, the scopes have preamps and signal generators, the curve tracer is full of analog circuits.  One of our products in development has 18 op-amps and a bunch of discrete transistors.  Op-amps have greatly simplified analog circuit design.  I'm old enough to remember when the one of the first integrated circuit op-amps (the Fairchild uA709) arrived at our workplace.  It came in a jewel box and cost about $100.  Everyone was terrified of using it.  That's changed over the years: one of our op-amp favorites, the TL074, is under a buck for a quad op-amp package.  But analog circuit design is more than plugging an op-amp and some discrete components into a circuit board.  We struggled for over an hour one evening to find the source of a mysterious offset, which turned out to be op-amp bias current (a rookie mistake on my part).

Microprocessor systems are easier to get working, and to do something significant with relatively modest circuit design effort and some programming. So the world is full of Arduinos and other Fruity Computers doing wonderful things, much of it accomplished by beginners - which is great stuff. (Beware, however, computer coders: big programs are not just bigger small programs: they are qualitatively different. You need to plan and design them.) But if you want to be a designer of electronic systems, you need to know both: digital *and* analog. For one thing, you need to be able to choose: sometimes one op-amp can replace hundreds of lines of computer code -- and work better. Other times, you need the flexibility of software to reduce the demands on the analog circuitry.

It takes some time to become an analog circuit designer. Analog design is a nice mix of theory and practice.  There is always new stuff to learn, which is one of the attractions.  It helps to like tinkering with circuits (the best way to learn) and to be endlessly curious.

That said, this is a great time to do analog circuit design: loads of information on the web, readily accessible parts catalogs and suppliers, free circuit simulators, inexpensive components, and terrific instruments for your workbench.  Go for it.

Waveform Averaging

Posted by Peter Hiscocks on 3/23/2016 to Technical Papers

Click here to read our latest application note on waveform averaging in the CircuitGear software.  Waveform averaging can be very useful in extracting a signal from noise as long as a solid trigger signal is available.  The application note shows how a software contribution from a Syscomp customer can significantly improve the signal-to-noise ratio using waveform averaging.

On Power Supplies

Posted by Peter Hiscocks on 1/13/2016 to Products

Are you starting to learn electronics? A power supply is an excellent starter project: the basics are very straightforward, and a power supply is reasonable to built and troubleshoot.  Even better, as your skills advance, there are some interesting - and not so obvious subtleties.

Integrating Sphere for Luminance Calibration

Posted by Peter Hiscocks on 1/13/2016 to Technical Papers

One of our projects is to measure the luminance (brightness) of various sources of light pollution, using a digital camera.  To do that, you need a calibrated source of luminance.  It turns out that a low-cost purpose-made light integrating sphere can provide this.  The linked paper describes the integrating sphere and provides some background information and theory.

Introduction to Electronics

Posted by Peter Hiscocks on 8/20/2015 to Technical Papers
Here is a presentation that Syscomp did for a group at the University of Toronto. Students are assumed to have *no* background in electronics. The objectives are to provide an introduction to circuit building and electronic measuring instruments with immediate results (an illuminated LED!), working up to a light-controlled warbler sound generator.

We also have this material available as a slide presentation. Contact us (support@syscompdesign.com) to obtain a copy in Open Office Presentation or Microsoft Powerpoint format.

Syscomp Software on the Raspberry Pi 2

Posted by Peter Hiscocks on 8/20/2015 to Software
Recently, we have been evaluating the Raspberry Pi 2 computer. For those who have been visiting a distant part of the galaxy for the last few months, the Raspberry Pi 2 (RP-2) is a single-board computer, about the size of a business card, that runs Raspian, a Linux variant that is very similar to Debian linux. 

For example, one can download and install software using an internet connection with the standard command 'sudo apt-get install name-of-package'. The boot sequence looks very similar to the sequence we used to see, years ago, on a very expensive Sun Workstation.


I'm *very* impressed by this device. The network connection and video worked immediately.  In a few minutes, I was able to configure a new password, change the keyboard configuration to Dvorak, and install a bunch of useful programs - such as the Joe editor.

There's been some interest in running Syscomp software on this platform, so we investigated that possibility.

Syscomp software is written in the Tcl language, and happily Tcl/Tk 8.5.11 is included in Raspian.  So the majority of the code runs correctly.  We use some 'packages' that are written in Tcl/Tk, and so they should (and do) run correctly.  We also use some packages that were written in C and compiled for a specific architecture.  In the past, we have taken care of that by bundling them with a platform-specific executable.  But the compiled packages won't run on the Pi unless they are re-compiled.

It turns out that those compiled, platform-specific packages can be replaced in various ways.  For example, we converted all the .png formatted images to the .gif version, using the Imagemagick 'convert' command.  That eliminated the requirement for the Img package.  The Tktable package is used in the curve tracer software: that was replaced by the standard Tcl text widget.  The directories 'Bwidget18Lin' and 'MathLin' (both of which are written in Tcl) were obtained from the DSO-101 software and moved into the software directory.

When these changes are made, code can be run on *any* platform that has Tcl/Tk available, using the command 'wish main.tcl'. That includes the Raspberry Pi.

Here's a screenshot of the Curve Tracer software running on the Pi.


We've also verified that these modifications allow the CircuitGear CGR-201 and Waveform Generator WGM-201 to run on the Pi.

The next release of the software for our various instruments will be modified to eliminate the requirement for platform-specific, compiled packages - so it should run from the wish command on the Linux, Windows, Mac *and* Raspberry Pi 2 platforms.

If you want to port our software to the RP-2 and can't wait for the next release of the software, get in touch and we'll send you a zipfile of the RP-2 version.

Incidentally, the RP-2 is powerful enough to run the CGR-101 software, which is *very* demanding.  Earlier versions of the Pi are too slow to be useable for the CircuitGear software, but they may be capable of running software for less demanding instruments such as the Curve Tracer and Multimeter.

No More Sourceforge

Posted by Syscomp Team on 7/29/2015 to Software
In the past, we have occasionally posted our source code to the SourceForge repository. We've decided to stop using SourceForge, for the reasons shown in this story on Slashdot:


In essence, SourceForge were modifying hosted code to add advertising to the code. This has caught the attention of the open source community: there are currently hundreds of comments on this story. The source code for our products is always available on our website, and that's the official source. We discourage using source code from other sites. Yes, you can read the code to ensure it's not virused or loaded with advertizing, but that's tedious and it's not necessary if you get it from the Syscomp website.

 Technical Papers
 Software
 Resources
 Products
 Accessories

 Testing a Selenium Rectifier
 Yes, we test IGBTs
 Negative Tracking Power Supply
 Analog Design Not Dead
 Waveform Averaging

 October 2016
 June 2016
 March 2016
 January 2016
 August 2015
 July 2015
 May 2015
 February 2015
 January 2015
 October 2014
 September 2014
 June 2014
 March 2014
 February 2014
 September 2013
 July 2013
 April 2012
 February 2012
 January 2012
 October 2011
 September 2011
 June 2011
 March 2011
 January 2011
 December 2010
 October 2010
 September 2010
 May 2010
 April 2010
 December 2009
 November 2009
 April 2009
 March 2009
 December 2008
 October 2008
 September 2008
 June 2008
 May 2008
 March 2008
 February 2008
 September 2007
 August 2007
 July 2007
 June 2007
 October 2006
 September 2006
 July 2006
 May 2006
 March 2006