As I was preparing to set up an experiment on viewing analog and digital signals associated with various components on a recapped PSU, I was interrupted by a friend who asked “Was machts du?” (What are you doing?). That was when I realized that I may not have a clue as to what is going on! I had hoped to look at the signals one can observe by experiment, compare them to the ideal case of what they ought to be and try to understand the cause behind discrepancies if they arise. A tall order, but this is something I’ve always wanted to do.
However, the pause made me think about PSU reviews in general, where one measures output signal quality – popularly called “ripple” – and draw conclusions about the quality of the unit being reviewed. Several authors around the interweb are making such measurements, but the question I have is “Do we have the complete picture? How comprehensive are such measurements? Would just a single measurement configuration suffice?”
These are some of the questions I’ve attempted to tackle. I may not have all the answers, but I do hope that this article offers a bit more assistance to those reviewing PSU’s by asking the right questions.
An SMPS offers a huge variety of signals. A good design would seek to localize disparate classes of signals. The essential measurement to characterize a PSU or a subset among the components of a PSU is current.
In a PSU, current is the more fundamental quantity, whereas voltage is a consequential entity. What you see on an O-Scope display is this consequential entity (the O-Scope has an in-built scaled I-V amplifier), which, may or may not, depict the true picture. We can have currents ranging from mA’s to ten’s of A DC or AC, with a frequency band ranging from DC to hundreds of kHz. This presents a unique problem.
The issue with having such a wide frequency band is that one cannot use a simple current-shunt configuration (like your multi-meter or merely tacking on BNC cables would) to extract information through the entire bandwidth of frequencies. So, our O-Scope output shows no “ripple”. What does this mean? Not much, I believe.
In my opinion, this information is useless without knowing beforehand what frequency band is being targeted as dictated by the measurement setup. Remember that the O-Scope has a very wide bandwidth and is not part of the problem. The problem lies in how the data is collected, or more specifically, the probes being used.
It is pointless to portray a unit as having zero 1 kHz or 60 Hz ripple when your probes attenuate those signals! A simple prescription to void such a pitfall would be to figure these out:
- What kind of waveform are you measuring? (Pulse, sine, DC, sawtooth or a combination of these?)
- What is the approximate expected frequency range of the current being measured?
- What range do you wish to target?
- Do your current probes match your target?
- Do you need to design a proper terminating network?
In the next section, we will take a brief look at some of the practicalities of this measurement issue.
Non-inductive current shunts are 4 terminal resistive devices inserted in series with the line in which the current is being measured. In other words, it is just a piece of wire with four ports. Ideally it should have very low resistance and almost zero inductance, but that is not possible practically. So, they are generally unsuitable for kHz level measurements and pulsed wave shapes.
A lot depends on the type of wire being used, its length and the current being measured. The moment the inductance of the wire becomes comparable to its resistance, we are going to have a problem and the voltage wave shape will get distorted. This problem is particularly acute when the currents are large (>10A).
If you do wish to use a piece of wire, twist it around to form a double helix, insert it in the line and connect the coax BNC cable in parallel with it. For higher currents, it may be a good idea to use a sandwich arrangement made by a folded sheet of metal.
These devices can be used over a very wide frequency range, say from 0.1 Hz to 1 MHz. Most commercial probes, such as the Textronix CT series, have a built-in current transformer in them. I’d say these commercial probes are a must have in any reviewer’s armory. They can be designed in-house if the nature of the signal being measured is known.
The design specifics are beyond the scope of this article. They are the basis of both high frequency and low frequency AC current probes. The basic design is just an LR parallel circuit but differences include different choice of magnetic core, number of windings and the thickness of the wires used.
When a conductor is introduced in a magnetic field, the electrons in the valance band feel a force and arrange themselves in such a way that the energy of the field+conductor is a minimum. This manifests as a measurable potential difference or voltage and is called the Hall effect. These probes are great for characterizing EMI (Electromagnetic Interference) fields.
While it is OK to use a non-inductive probe if you wish to measure low currents at low frequencies, in practice it is a good idea to provide proper termination to the transmission line, ie your BNC cable. Most SMPS units are accompanied by “common mode noise”. So it is essential to provide a path for the extraneous signals to travel to ground. A simple 47uF tantalum capacitor between the BNC signal tip and the ground would suffice.
It is always good to see that there is an increase in technical awareness among computer enthusiasts. Indeed, several sites have come to the forefront with professional grade reviews. But then there is always scope to learn. I hope this article offers a different perspective on the methods of characterizing output signal quality in traditional PSU reviews and augments the hard work put in by PSU reviewers.