Multitest plug in filters – ten plug in filters – measurements and blind test


Do those ‘plug-in’ filters really do anything? That is the question we want to answer in this big test of ten ‘filters’. You know them; those plug-in plugs…. plug and play ‘junk cleaners’…. Right? We have samples from Furutech, Ansuz, Akiko, Kemp, Isotek, IFI and NordOst. A fine cross-section of the market, we think. And we have several variations as well. We have measured and listened blind(!). In short: a complete story with – we promise – a surprising result!

Let’s start by noting that these are extremely complex tests if you want to do them correctly. Überhaupt, testing hi-fi equipment is far from easy; there are countless variables. Always. But power conditioning – in this case filters – is even more complex we have concluded.

There are many reasons for that. First of all, filters are load-dependent. Simply explained, the degree of filtering depends on the load on the filter. This is evident in the measurements below where we simply adjusted the impedance. That makes it so that a filter can sound different in different setups and with different devices. And so it does, in our experience. The effect of the Isotek Titan is different on the Pass Labs than on the Bryston, for example.

A second element that turns out to be particularly complex with filters is the measurement side. What should we measure? How do we quantify quality? What ís quality in filters? Is more filtering better? What kind of filter slope is correct? Honestly: we don’t know (yet). We can only show that a filter ís doing something in a certain bandwidth. We were only able to measure a few elements in this test: the – possible – filter slope, the degree of filtering with white noise and the degree of filtering in a ‘real’ situation. So with noise from the mains. But that is only the beginning, because there are many other elements that play a role. Consider the influence of other devices, the speed of the filter (impulse behavior) and any output impedance. But we’ll certainly get to all that in a sequel.

By the way, measuring the effect on real noise on the mains is quite difficult, because the noise patterns are not stationary. So not every measurement is exactly the same. We try by means to compensate for this somewhat, but know that.

For the “synthetic” tests – i.e. a sweep or white noise – we loaded with 50 Ohms if necessary. We have both a variable resistor and a dummy load. Both are compensated by first normalizing on the Spectrum Analyzer. Then the baseline is straight and we can then measure the filters. Between each round we normalize again. We also make sure that the measuring equipment is warmed up for 30 minutes. Both the analyzer and the LCR meter are (very) sensitive to this, we have noticed.

We measure within a bandwidth of 0 – 96 kHz on the Prism dScope, 10,000 Hz – 1 MHz and 10,000 Hz – 35 MHz on the Spectrum Analyzer. This is to determine what the filter does over a wider bandwidth. It is quite interesting to see to what band the filter throws the ‘crap’. After all, it has to go somewhere. That can be seen in the measurements.

To do the measurements, we soldered a socket between the measurement cables. To properly load (terminate) the filters, we connected a 50 Ohm load via a T-piece. This was definitely visible in the measurements. The filtering shifts per load. Just look at the measurements below.

Filter response on different load

You can clearly see that the degree of filtering shifts per load.

Measurement setup

We use practically all equipment in this test. We try to map the characteristics of the filters with the LCR meter (capacitance, inductance, resistance) and map the degree of filtering and effectiveness with the Rigol Spectrum Analyzer and Prism dScope. While the Prism cannot measure a very broad frequency-band; it can measure much more accurately in the audio range. And that is certainly interesting. In addition, the Prism allows us to overlay graphs more nicely. The Rigol can only superimpose 3 traces. That is not enough in this test; we have ten filters (we auditioned nine of them; the Old Kemp is not interesting in that sense).

Not measurable

You will see in this test that some products show nothing. The question is: do they do anything in practice? We’re going to find that out in the listening tests. That was done blind on Sept. 17, 2022. You can read the findings for each product. In it we also have the measurements with explanations. Please know that some plugin filters have an active system that may operate at 110 / 230 volts. In our setup, we operate at 10 volts or less for some measurements to measure the filter characteristics and not blow up our equipment.

The test system

The test was done – live and blind – in Alpha Audio’s listening room. The system consists of:

The Filters were tested on the Sonnet Pasithea. All (brief) descriptions you read are relative to the baseline which consists of no filter in the socket to which the Sonnet Pasithea is connected. In short: it is about the (subtle) differences between nothing and the device under test (DUT).


Leave a Reply