circuitcellar.com
Magazine Support   Digital Library   Products & Services   Suppliers Directory 
 
 





 

Issue 149 December 2002
Quad Bench Power Supply


by Brian Millier

Start The Analog CoreThe Zetex ZXCT1009 An Ideal IsolatorMCU and User InterfaceFirmware Sources and PDF

AN IDEAL ISOLATOR

After spending years servicing and designing electronics devices, I have to say that I’m as impressed with some of the amazing things that were done with vacuum tube circuits back in the old days, as I am with some of the modern, miniature ICs that are available today.

For this project, though, I pampered myself with state-of-the-art devices rather than depending on clever, but more involved, circuits using conventional devices. I’ve already described the Zetex current monitor, which is one example of this. I continued with this trend in choosing the isolation technique for the floating power supplies.

The digital control and monitoring signals for the two floating supplies have to be electrically isolated from the ground-referenced MCU circuit. Thanks to the clever design of Microchip’s SPI digital potentiometer and SPI ADC, each power supply needed only four control signals: three outputs from the MCU and one input.

My first inclination was to use optoisolator chips. I had just finished another project using optoisolators to interface the same Microchip SPI ADCs. In that project, meeting the ADC’s SPI timing considerations given the rather slow response of the optoisolators was a bit tricky, although possible.

Luckily, Jeff Bachiochi had just written a column about isolation in which he outlined a novel line of isolators made by Nonvolatile Electronics ("You’re Not Alone—Dealing with Isolation," Circuit Cellar 142). Rather than using an optical method to achieve galvanic isolation, these isolators use magnetism. Although pulse transformers have been around for ages and can perform isolation using magnetism, they are comparatively bulky, expensive, and don’t pass DC levels.

The IsoLoop isolators, on the other hand, use GMR or giant magnetoresistive devices to sense the magnetic field change produced by an excitation coil, which is nearby but electrically isolated. The change in resistance of the magnetic thin film layer is used, along with other on-chip circuitry, to implement the isolation function of the device. The IsoLoop devices actually differentiate the input signal, and send only short magnetic pulses through their excitation coils during input signal transitions. The resulting resistance changes in the magnetic thin film layer—configured in a Wheatstone bridge—are measured, and the resulting output signal is actually the output of an on-chip latch device.

Don’t be fooled by the use of the term "giant" in GMR; these devices are tiny. Typically, four isolators will fit into a 16-pin wide SOIC package. The wide package is needed, presumably, to allow the devices to withstand the 2500 VRMS at which they are rated.

With regard to the packaging, I was impressed with NVE’s decision to produce several different device configurations. They sell the normal quad devices with all four channels configured in the same direction (IL715); however, they also sell quad devices containing two channels in each direction (IL716). My favorite, the IL717, has three channels in one direction and the remainder going in the other direction. This configuration is perfect for SPI device isolation, which needs a Chip Select, Clock, and Data Out lines coming out from the MCU and a Data In line going back into the MCU.

Given the modest voltage isolation I needed for this supply, I could have used a quad optical isolator and wired up one section "backwards," so to speak, but the PCB layout would have been much less neat. In cases where input and output signals have to be isolated and substantial voltage isolation is required, the only way to achieve this—apart from using separate devices—is to use an appropriately configured device like those in this IsoLoop family.

I’ve actually saved the best part for last: these IsoLoop devices are fast! The IL700 family exhibits a 100-Mbps data rate. In addition, it has only 2-ns pulse width distortion and 10-ns pulse delay.

Unlike optoisolators, which require LED drive voltage/current and often don’t provide logic-level output signals, the IsoLoop devices work directly with 3.3- or 5-V logic devices including MCUs. Although an optical isolator requires a steady drive current whenever its LED is turned on, the IsoLoop devices use only a short pulse of magnetism whenever the input signal changes state (even though a small but steady current is required for the detection and latching circuitry in the chip).

The IL717 that I used requires only a 2.5-mA power supply current on its input side, and 6 mA on its output side. This difference arises from the fact that the device has three channels in one direction and only one in the other.

In my design, I did not have to give any more thought to the SPI timing on the floating channels than I did to the channel that wasn’t isolated. Basically, what goes into the IL717 is what comes out the other side!

There are only two cautionary notes that I would add regarding these devices. First, IsoLoop devices transmit their signal across the isolation barrier only on signal transitions. The recovered signal on the other side of the barrier is then electrically latched. Practically, this means that the output of the devices is indeterminate until input transitions occur. For some applications, this means that an initialization routine must be performed to ensure that the device’s outputs are in a known state after power-up.

The second cautionary note is just as important. Because the devices rely on sending a short magnetic pulse at each input transition, it is important to place at least a 47-nF ceramic decoupling capacitor between VDD and ground on both input and output ports of the device. The capacitors should be placed close to the actual device pins.

I tried to share one capacitor between two IsoLoop devices on the common MCU port side of the two devices. This didn’t work. There were random output errors on the device farthest away from the sole capacitor that disappeared completely when I followed directions!