It is not really the measurment accuracy that is the problem, it is the measurment method Mike. I could make an MFT with a 24 bit A to D converter, which might have an inherent accuracy of about 22 bits. This is by no means the limit, digital meters with 10 or more decimal digits are made and used (around 36 binary bits) but they are frightningly expensive. However Andy is right in that the method of measuring the EFLI is fundamentally inaccurate, and that is why I carefully described the proper method. Adding a tiny load to an unstable supply and trying to measure millivolt or microvolt differences is a fools paradise, but the MFTs are believed to be accurate. I rather doubt (from experience) that an MFT ELFI of less than 0.1 Ohms is anything like 5% accurate, the metering method might be, the practical result is not. Calibration is not carried out with a mains derived supply, it is a very carefully controlled oscillator and an exact resistance. In other words variations in the mains supply are eliminated. That may give 5% plus a few counts, I am sure the mains will not. This is not a critism of MFTs, but there are significant difficulties with the results when we get onto more than 100A supplies, with much higher PSCCs, much lower Ze etc. At the final circuit DB you are on much firmer ground. There are a lot of other types of A to D converter inside chips, offering much faster conversion, greater accuracy etc than the old discrete types, just try a dual slope type with an audio signal! A to D of at least 6GHz is easily available today, and the number of bits probably unlimited given enough money and effort (and a completely noise free signal to measure, which of course doesn't exist). I am sure MFT design will improve with time, and the modern bells and whistles have not really made the measurments much better, just quicker and easier to use. It is still necessary to understand exactly what is really being measured, and why it might be useful. I would really like my MFT to take 50A for perhaps 10ms right at the waveform peak, but the resulting power loss can be a problem, and then compute the ELFI. I do have one meter which does something like this but it only allows about one measurement per minute to allow for cooling! It is usually better than the Megger in its estimates compared with calculation.

It is not really the measurment accuracy that is the problem, it is the measurment method Mike. I could make an MFT with a 24 bit A to D converter, which might have an inherent accuracy of about 22 bits. This is by no means the limit, digital meters with 10 or more decimal digits are made and used (around 36 binary bits) but they are frightningly expensive. However Andy is right in that the method of measuring the EFLI is fundamentally inaccurate, and that is why I carefully described the proper method. Adding a tiny load to an unstable supply and trying to measure millivolt or microvolt differences is a fools paradise, but the MFTs are believed to be accurate. I rather doubt (from experience) that an MFT ELFI of less than 0.1 Ohms is anything like 5% accurate, the metering method might be, the practical result is not. Calibration is not carried out with a mains derived supply, it is a very carefully controlled oscillator and an exact resistance. In other words variations in the mains supply are eliminated. That may give 5% plus a few counts, I am sure the mains will not. This is not a critism of MFTs, but there are significant difficulties with the results when we get onto more than 100A supplies, with much higher PSCCs, much lower Ze etc. At the final circuit DB you are on much firmer ground. There are a lot of other types of A to D converter inside chips, offering much faster conversion, greater accuracy etc than the old discrete types, just try a dual slope type with an audio signal! A to D of at least 6GHz is easily available today, and the number of bits probably unlimited given enough money and effort (and a completely noise free signal to measure, which of course doesn't exist). I am sure MFT design will improve with time, and the modern bells and whistles have not really made the measurments much better, just quicker and easier to use. It is still necessary to understand exactly what is really being measured, and why it might be useful. I would really like my MFT to take 50A for perhaps 10ms right at the waveform peak, but the resulting power loss can be a problem, and then compute the ELFI. I do have one meter which does something like this but it only allows about one measurement per minute to allow for cooling! It is usually better than the Megger in its estimates compared with calculation.