The problem is that the lab measurements have very little to do with real-life applications1. This article will focus on measurements of EMI levels in that actual use, more specifically, conducted EMI.
Semantics: while EMI is often defined as a process of interference with, or disruption of, operation caused by high-frequency signals, in this article we will bend this definition a bit for brevity sake and use it as a convenient abbreviation for high-frequency signals that cause such problems.
- Equipment downtime, i.e. interruptions in operation of production equipment;
- Errors in measurements, often leading to altered process parameters in automated production;
- Errors in data communication; and
- Exposure of sensitive devices to electrical overstress (EOS)2 which lowers yield and increases probability of product failure in the field (latent damage).
Most EMI signatures in real-life applications are transients, i.e., short pulses, periodic or not. Our instrumentation and methodology will be focused on this type of signal.
One of the key sources of EMI-caused problems resides not on live and neutral wires of AC mains but on ground. This includes both facility ground and grounding inside the equipment. At high frequencies there can be voltage differences between different grounded points inside equipment even if the resistance at DC or impedance at the mains’ frequencies (50/60 Hz) is very low. This paper3 shows high-frequency signals with peak amplitude exceeding 1V between grounded points with impedance of just 0.2 Ohms between them. At high frequencies impedance is quite a bit higher than at 50/60Hz due to parasitic inductance of wires and skin effect. High impedance at high frequencies inevitably leads to voltage difference causing problems with reference points for electric circuits and electrical overstress exposure for sensitive devices assumed to be safe when contacting grounded surfaces. Another conducted EMI problem – high-frequency voltages and currents on data lines – is often caused by induced signals on runs of data cables.
Measurements of EMI on facility power lines reflect total EMI “dumped” on the shared facility power by a combination of all equipment in the facility. Measuring EMI presents several challenges.
- Galvanic separation from power line
- Balanced input
- 50 Ohms output
- Flat frequency response within the required range
- Overload limiter to protect other instruments
We already know that there can be significant high frequency voltage between two otherwise well-grounded points. Why is this bad? In short, problematic reference voltage for electronic circuits and a possibility of electrical overstress. This is well covered in literature2,8. In this paper we will focus on measurements. Measurements of voltage between different grounded points using an AC-powered oscilloscope are worthless because an oscilloscope’s own ground connected via its AC cable to the mains’ ground adds yet another variable to this equation. A battery-powered oscilloscope or any oscilloscope with the EMI adapter described above is a better way of measuring EMI between different grounds. User should be careful when using just the battery-powered oscilloscope alone because sometimes the presumably grounded object isn’t really connected to ground and has rather high AC voltage – it pays to check before connecting your oscilloscope. If you are using an EMI adapter this is not a problem because it galvanically separates mains’ voltage from its output.
AC Mains
Please review notes about measurements on EMI on mains earlier in this article for safe and accurate results.
EMI on power lines can be differential (between Live and Neutral) and common mode (between Live or Neutral and Ground). The nature and the pattern of such emission are different – both need to be measured. You may find situations with plenty of differential EMI and very little common mode one, or just the opposite. A good portion of EMI on mains (except occasional commutation events – On/Off) are synchronized with the waveform of the mains’ voltage.
Make sure that the equipment is fully operational when performing the measurements – equipment in “Off” state produces no EMI.
As an illustration, Figure 2 shows EMI at the output of uninterruptable power supply (UPS)9. In short, this is EMI from operation of a switched mode power supply (SMPS). Note how it is synchronized with 60Hz mains.
EMI on Ground
Some factories have separate facility ground – either a separate ground cable or ground bars. One of the reasons for such separate ground is potential reduction of EMI. This assumption sometimes fails spectacularly. Figure 3 shows EMI voltage between such “special” ground and the mains’ ground. A piece of equipment or a workbench connected to both grounds may have a hard time with reference voltages for its electronics and with exposure of devices to EOS.
In manufacturing, different tools often are conjoined for performing a task. It helps if there is no EMI voltage between grounds of these tools. AC powered oscilloscope by itself, of course, is not the right tool for such measurements.
Measurements of EMI on ground or between grounds of different tools can also be done using a current probe, but only after verifying that both grounds are actual grounds.
Automated Handling Process
Figure 4 shows points of measurements of EMI between the robotic arm and the chassis of an IC handler (same applies to other tools handling semiconductor devices such as SMT pick-and-place machines, wire bonders, die attach, etc.). The assumption that a device is “safe” when in contact with ground and nothing else fails a basic equipotentiality test – the voltage difference between different grounded points at high frequencies may not be zero10.
Manual Soldering
From an electrical overstress aspect, soldering is one of the worst processes one could imagine – a metal tip of the soldering iron makes a galvanic connection to the pins of the devices which are connected to a potentially different voltage point. The soldering iron tip is grounded via AC outlet; PCB is coupled to the bench ground which is often connected to so-called “ESD” ground. When the tip of the iron touches the component, resulting current can easily subject device to electrical overstress. Whether the PCB is galvanically connected to ground or capacitively coupled to it, this can still happen because capacitive coupling offers very low impedance to high frequency signals11.
How should we test for EMI-caused EOS in a soldering process? Figure 6 provides some suggestions – voltage (a) and current (b and c) measurements. A user is advised to put together a kind of a fixture as shown to avoid burning fingers or melting test probes. The resistor in Figure 6c is 10 Ohms (any value between 10 and 50 Ohms will do).
The probes shown belong to an EMI adapter that resolves ground loop issues. An oscilloscope without it would work only if it is battery-powered, otherwise the measurements would be completely meaningless.
What if the EMI is caused by a soldering iron itself? Absolute majority of professional grade soldering irons do not introduce high-frequency artifacts by themselves. The problem lies in EMI on the facility’s AC power and ground.
Figure 7 demonstrates this cause-and-effect. The current spike from the tip of the iron is well-synchronized with the corresponding spike on the mains (as measured with an EMI adapter so that the oscilloscope is well-insulated from the mains).
Is what we see in Figure 7 safe for the devices? This will be discussed further in this article.
Figure 5 shows how to measure current between robotic arm and the tool’s chassis. A special current probe should be used for this purpose. The bandwidth of such a probe should start at low kHz and extend to at least 30MHz. The probe’s sensitivity to 50/60 Hz is a detriment as it masks the signals of interest.
A metrology purist at this point would note that all these wires will adversely affect the measurements and will be highly susceptible to radiated emission which is always plenty in a manufacturing environment, especially inside the equipment. That purist would be correct. But the only alternative would be not doing the measurements, so we will accept inherent unevenness in frequency response and signal ringing as a “fact of life.”
A digital oscilloscope has finite resolution – both its A/D converter and its screen, with typical resolution of 8-bit, or 256 steps. When the measured signal is in the lowest few bits, the oscilloscope adds to it its own noise and A/D dither, inflating the signal value. Unless the vertical scale is pegged up to maximum sensitivity already, the data in Figure 8a offers little value. In the ideal case the signal should occupy ~2/3 of the screen as shown in Figure 8b. The signal, the very same signal as in Figure 8a, but is now clearly visible and with more believable values. Watch for overload – if the signal is clipped on the screen, the front stage of the oscilloscope may be distorting the waveform and the measurements are no longer accurate.
Note that signals A and B have similar maximum amplitude but different polarity. These peaks can be both positive and negative, therefore it is important to repeat capture with different trigger polarity.
To the author’s knowledge, there is no IEC-level standard addressing broad industry needs for low levels of EMI. There are, however, some industry-specific documents specifying acceptable levels of EMI.
According to this table, the voltage between the robotic arm and the chassis shown in Figure 9 (400mV) exceeds even the most generous level of conducted emission – 0.3V. If you are reading this article, most likely your devices have smaller geometry and are consequently more sensitive and demanding lower levels of EMI.
Even if you are not overly concerned with EMI today, tomorrow you may be – sensitivity of components and of electronic circuits to EMI is continually increasing. You may be at a place where resolving EMI issues presents an urgent need rather than academic curiosity. The next step is mitigation of EMI in your process, but this is outside of the scope of this article.
- V. Kraz, “Electromagnetic Compliance – a View from the Field,” In Compliance Magazine, October 2017.
- TR23.0-01-20. ESD Association Technical Report for the Protection of EOS/ESD Susceptible Items – Electrical Overstress in Manufacturing and Test, EOS/ESD Association 2020.
- J. Salisbury, et. al. Reducing EOS Current in Hot Bar Process in Manufacturing of Fiberoptics Components, EOS/ESD Symposium, 2016.
- Bogdan Adamczyk, “Conducted Emissions Measurements: Voltage Method,” In Compliance Magazine, August 2017.
- K. Armstong, “Guide to Testing Conducted Emissions (Based on the Methods in EN 55022 and EN 55011),” In Compliance Magazine, July 2011.
- IPC-A-610G: Acceptability of Electronic Assemblies, http://www.ipc.org
- OnFILTER, Power Line EMI Adapters, https://www.onfilter.com/emi-measurements.
- R. Dorf, The Engineering Handbook, CRC Press, 2018.
- App. Note: CleanSweep® AC EMI Filters and UPS, OnFILTER.
- V. Kraz, S. Meyer, R. Fitzpatrick, “The Implementation of SEMI E176,” In Compliance Magazine, September 2019.
- T. Iben, et.al., EMI-Generated EOS in Wire Bonder, EOS/ESD Symposium, 2017.
- SEMI, http://www.semi.org.
- SEMI E176-1017 Guide to Assessing and Minimizing Electromagnetic Interference in a Semiconductor Manufacturing Environment.
- V. Kraz, “Dealing with EMI in Semiconductor Manufacturing: Part II,” In Compliance Magazine, May 2018.
- SI/ESD STM13.1 Electrical Soldering/Desoldering Hand Tools.