At ElectroneX, an Electronics Design and Assembly Expo held at Rosehill Gardens Event Centre (Sydney, Australia) from September 5-6, 2018, Mr. Dave Jones interviewed Mr. Bart Schroder of Cleverscope on the CS448 isolated channel oscilloscope project. (If you haven’t seen the video, see it here:
There were a lot of important and interesting topics covered in the video. Most of them were pragmatic and in my opinion deserve deeper insights in this article. 

What is all this H-bridge and Isolation Stuff, Anyhow? 

The lay viewer may be immediately distraught upon hearing the terms “H-Bridge” and “Isolation”. Ergo, we dedicate this section to provide a comprehensive introduction to both. Motors are typically driven by a switching signal to induce motion through a mechanism depending on the type of motor used. The switching signal is produced by MOSFET pairs that form a letter “H” (hence the term “H-bridge”, or it can signify the term “half”, whichever floats your boat) that turn on and off at intervals separated by a “dead time” (which prevents both higher and lower side FETs from turning on simultaneously causing a short to ground). The speed of the motor can be controlled by regulating the duty cycle of the switching waveform. In the video, pulse width modulation (PWM) was employed to do this. The FETs here are usually called the motor’s drivers, and sometimes may even have pre-drivers (i.e. another set of high/low side FETs increasing the number of components used in the circuit). 

The term isolation was used here in an electrical sense, i.e. the probe’s ground is not connected to the ground of the oscilloscope (which is usually at earth ground potential – necessitating other measurement instruments to be disconnected from earth to avoid ground loops). In this scenario, a differential probe is preferred, however as argued by Mr. Bart in the video, they have poor common mode rejection ratio (CMRR) at higher frequencies.

Miller Capacitances Drive FETs to the Wrong Regions of Operation 

In the zoom-in oscilloscope plot of the H-Bridge demo, Mr. Bart pointed to an anomaly at the rising edge of the switching waveform, which lined up neatly to another spike in the other (synchronized) switching waveform. The most common problem in the event of such spikes at both the input and output is mis-triggering of the protection circuits or biasing that significantly affects the circuit’s region of operation. Since the drivers are FETs, the regions of operation of these FETs are affected. The gate-source and drain-source potentials have ranges (interdependent and not unique) that determine whether a FET is in cut-off, triode or saturation. Therefore, spikes or disturbances caused by wanton phenomena can drive the gate-source and drain-source potentials out of the necessitated range, effectively turning off (cut-off) or on (triode) the FETs at the wrong time. Even worse, such peculiarities may extend past the dead time and short the supply to ground, causing destructive currents on the high and low side FETs. 

How to Measure Idiosyncratic Signals 

To probe these kinds of signals, we would want to put our probes on the FETs, which was on the gate terminal in the demo. But there is a precaution when monitoring the high side FETs. The reference or ground of the probe would not be at 0V. They would be at the potential of the source terminal of the high side FET. Oscilloscopes are typically fixed at earth ground (0V) and connecting the two would cause an early (or late) 4th of July. 

So what tools do we have to work around this limitation? Differential probes operate without a reference by checking the rate of change of the signal. Unfortunately, their CMRR is alright at a few Hz, but gets worse at higher frequencies as previously mentioned. In the video, this equates to 10V riding on the signal, rendering detection of glitches and aberrations futile. 

The CS448 Isolated Channel Oscilloscope 

An elegant solution introduced by Mr. Bart is the CS448 isolated channel oscilloscope. The specs of the product, as well as its formative stages have been well-documented in this article (Please see: Before going any further, it’s worth noting that Mr. Bart turned off the module before opening the case. This is a general rule of thumb not just for high voltage equipment but for any electronics or electrical apparatus (because, even though you won’t feel the shock of Thor’s hammer from a measly 12V supply, you may be the one to inflict damage on the equipment - unless you are fully aware of the vulnerable regions). 

Opening the instrument, quickly noticeable were the FPGA, the heat sinks for the 4 channels, and the pulse generator. The blue and green fiber optics were sparse and operated at half duplex. Mr. Bart mentioned something interesting when explaining these fiber optics and how it differed from the way the Tektronix guys did things. I’m not sure if I got this right but the Tek guys sent an already encoded signal down the fibers, in effect risking higher quantization noise (remember digital signal processing?) after decoding. With the CS448, the 14-bit ADC came after sending the raw light-modulated signal down the fibers. The lesson to be learned here is that simply changing the location of the ADC stage drastically affects performance or the end-product of the overall system. 

Power Supplies are Really Tricky 

“Power supplies are really tricky.” -is an understatement. At times, it may even start to lean on the science of trial-and-error(t.a.e.). I’m guessing getting the >100µV CMRR also had to resort to t.a.e. somewhere down the design stage. Though, the comment that really caught my attention was the one on bootstrap capacitors. I was previously under the impression that they served the purpose of pushing the voltage to the desired point from the supply (remember at t=0 caps are short – for more information on bootstrap capacitors, please see this application note: However, it was mentioned that they can be used to route noise. Mr. Bart said they weren’t needed because of the custom power supplies they designed to meet the design criteria (a grueling process that took an entire year). Having designed a power converter for USB type-C, I can attest a lot of things can go wrong when fine-tuning a power supply, especially when the specifications are demanding. T.a.e., perhaps? 

Synchronization and Probing Problems 

Every product design has its impediments, and the CS448 is no exception. Gain and phase measurements rely on probing 2 signals simultaneously. If 2 channels cannot be synchronized, then the measurement will have aberrations caused by the instrument itself. Another issue arises from standard passive probes. They do not interface well with the CS448, as they exhibit nonlinearity at higher frequencies. I believe passive probes do not have exceptional high frequency characteristics, as FET probes become preferable in this case – at the cost of a lower operating voltage. Mr. Bart and his team solved this dilemma by modifying the probes (a good application of when customizing something for your own needs becomes beneficial), necessitating a change in the product’s front-end design as well. They adjusted the poles of the circuit, probably as far away from the region of operation as possible. Resistive components are a flat line on a bode plot, so there is no use in changing them. Capacitors have a rising slope on the bode plot (and inductors vice versa), with steepness depending on the value of capacitance and inductance, so they're the elements most likely varied. 

Importance of Verification in Product Design 

The last take-away to address in the video (though there are copious more to be discussed) is the actual front-end design. Mr. Bart stated that the front-end design was full of hassles as components were not functioning according to plan. He cited an op-amp example, where the input was dragged to -5V when it was turned off. This last tidbit emphasizes the importance of verification in developing a product (it is the stage I am most involved in as well in my career), as anything unexpected can turn up out of the blue (remember Murphy). Personally, I too have a heap of experience where things go wrong when they shouldn’t (and vice versa), but that is for another article. 

Overall, the sound – in spite of the mic dropping near the beginning of the video, was still satisfactory. Both Mr. Bart and Mr. Dave’s voices were audible, though not euphonic due to some perennial white background noise. The visuals were commendable as the plots and product board were lucidly visible. We certainly have a lot to look forward to in the next interviews.