Home Hi-Fi Why comparing specifications is completely pointless

Why comparing specifications is completely pointless

Why comparing specifications is completely pointless

Bitrates, sampling rates, bit sizes, wattages, amplifier classes…. as an audio enthusiast, there are countless specifications to compare. But it is – virtually – all meaningless. Why? Because the specifications that matter are not reported ánd because every manufacturer measures differently. let’s explain that.

Since we started measuring ourselves, a world has opened up. Seriously. It feels a bit like we’ve suddenly found an extra light switch that allows us to better illuminate our work area. Although we have to say: we can now see all the junk around us much better. By the way, the deeper we dive into this matter now, the more questions we get…. and sometimes we also question our own “old understandings. But hey… that’s also part of it.

Back on-topic… The additional knowledge and depth gives an awful lot of insight into how devices work, which is important. Thereby we also see clearly how brands handle the (honest) specification of their products. That does make a difference.

1. Power (Wattages) means nothing!

To get straight to the point: power (wattages) means nothing. In many cases it says for example: 100 watts into 8 Ohms. And if you are lucky, there is a specification for 4 Ohms. That gives a little insight into the reserve; with a very stable amplifier the power doubles. What is often not mentioned, however, is how it is measured, and how much distortion occurs at maximum power. Is that 1%? 10%? Is it at 1 kHz? Full spectrum? In short: it is useless to compare wattages if you do not know this.

It is much more interesting to know what the behavior of the amplifier is at very low power levels. Below 1 watt for example. Below are three power measurements. Look carefully at the vertical axis … that is the distortion. The horizontal axis is the power. The Fosi amp has nearly 90% distortion at 39 uWatts. And well over 15% at 1 mW. 1 watt comes to 0.5% which is horribly high by today’s standards. In fact, most speakers are well below that. The Pass Labs X150.8 shows 0.35% at 37 uWatt, and then drops down. At 1mW it shows 0.07%. The Advance shows neat values in its price range: 1 mW comes in at 0.4%.

Why do we insist on this? Simple: the first watt is crucial. All microdetail is in the low regions of power. These values are nowhere to be found. Only maximum power, but never the distortion for 1 watt or a complete graph, when that’s what it’s all about.

Conclusion: full power specs say nothing and cannot be compared without more information. Thereby, the first watt is important.

[ppp_patron_only level=”5″ silent=”no”]

2. Sampingrates, bits and rates…?

High Res was the holy grail. The more bits and samples, the better. Basically that’s true, but here there are at least two big catches. And those are called jitter ánd the filter.

In the specifications sheets you can usually read which sampling rates a (digital) product supports: 24 bit / 192 kHz for example. Or consider 2.8 MHz DSD, or double DSD (5.6 MHz). CD quality is 16 bits per sample and 44100 samples per second. In itself enough for a nice reproduction of music. High-res has more dynamic range and more detail, provided it is recorded well.

However, what is usually not specified is what the jitter values are and how the filter works. Is it a brickwall filter (sharp)? A slow filter? And how is the slopes of the filter?

The heart of a digital system is the clock. This must do its job very stably and especially be very constant in the short term (long term says nothing, it can still oscillate (drift) in the short term; the important thing is that the clock is just stable and constant in the short term). The drift of a clock is usually expressed in ps – picoseconds. There are tests for measuring jitter. The most well-known is the JTest, but it doesn’t really measure the clock itself; the JTest measures interface jitter. This measures how immune the input to the dac is. (This measurement method is not suitable for NOS dacs, because they do not use oversampling and filters!).

Another very sound determining element is the filter that dacs use. In fact, filters make or break a dac. It’s with good reason Chord uses its own system and many enthusiasts favor NOS dacs: a NOS dac doesn’t use an output filter. Both have advantages and disadvantages. We will not go into that now.

Below are a few devices that we have measured for interface jitter. You can see that the AudioGD in NOS mode shows extremely high jitter: logical. In OS (oversampling) mode, by the way, it’s still not the best…. we see a huge “skirt” around the input. (High) jitter is audible. Lower jitter results in calmness and control. High jitter manifests itself in turmoil, blur in imaging and sharpness on T’s and S’s for example.

Below is an example of a couple of dac filters where we measured the slopes: you can clearly see the unwinding as well as the reflections in the measurements.

Filters are very decisive for the sound. It largely determines the character (signature) of the dac. A lot of brands have their own method of filtering. Think of Chord (Watts filter), Marantz has its own filters, and there are brands that allow you to choose multiple filters. And yes: this is definitely audible….

Both of these things are absolutely not included in the specifications. Partly because it’s so tricky. The point is that jitter and filters say much more about a product than bits and sampling rates…. So keep that in mind.

Conclusion: bits and sampling rates say nothing about playback. Jitter and filters say much more, but are not specified.

Frequency response

We will talk here mainly about loudspeakers. Because enthusiasts will mainly look at how deep a speaker can go and maybe the extension of the treble area. Our Focal Sopra had a range of 45 Hz – 40 kHz, according to Focal. No additional specifications. The TADs go from 30 Hz to 60 kHz. No additional specifications. So we still know nothing, since we don’t know what the deviation is on the 0-axis. Is it 45 Hz -3dB? -6dB? 30 Hz -3dB? -10dB?

Above is a measurement of the Sopra No1 and TAD E2. Both graphs have a window of 3ms, meaning everything below 200 Hz is filtered out and thus is not reliable. Keep that in mind. We do that measurement to see how the mids and treble are tuned. We often do multiple measurements to determine the ideal position of the microphone.

The Sopra has long been our reference speaker. With reason: it’s a very nice sounding speaker. Now it’s the TAD E2. They are two absolutely top-notch speakers. Both quite neutrally tuned – Focal slightly fresher – and both low distortion.

The character of the speaker is not at all apparent from the specs. The response on paper also says nothing. A measurement of the response says more, but what really says something is a combination of measurements with off-axis behavior and a measurement on the sofa. Then you get insight into the dispersion – radiation behavior – of the speaker.

What says even more is the distortion measurement. In this measurement you can see which harmonic is dominant at which frequency. It shows that the TAD has much less distortion in the lower frequency range and also shows less distortion overall (about 5dB lower). This gives an overall cleaner presentation with more calmness in the sound (we hear more layering). Unfortunately these measurements are never public (we have never seen them).

Conclusion: frequency range specifications say very little without additional information.

Nominal impedance

Very decisive for the cooperation with the amplifier, is the impedance characteristic. Now the nominal impedance is often given (8 Ohm, 6 Ohm, 4 Ohm…), but rarely the minimum impedance. Although, fortunately, we see this more and more.

We now measure the impedance (with our Sourcetronic LCR) and see that even certain affordable models show very unfavorable impedance characteristics. The result: they cannot be driven with a normal amplifier. This manifests itself in sharpness, harshness and sometimes even failure of the amplifier. If you are lucky, before the whole thing clips.

Below are the impedance and impedance phase of the Bowers 702S2, Sonus Faber Lumina II and TAD Evolution 2. The Bowers dips to 2.5 Ohms; that’s on the edge of what a “regular” amplifier still likes and can drive. There are many amplifiers that will either start clipping, or just die completely, taking the speaker with it, probably.

The TAD thinks 4.9 Ohm is enough. By the way, a peak in impedance is not very bad. Unless you play with single ended tubes for example; then it will definately add some color. Not so with transistor amplifiers. Then – also with tubes – the low impedance is a problem, because they can get out of breath. (Low impedance = more energy is needed. But if the power supply can’t handle that, things will start to clip).

As mentioned, more and more manufacturers are specifying the minimum impedance. It would be even better if the complete graph, including phase would be published, because then you can also see the diiferent slopes. Because only then you can really compare.

Conclusion: only a nominal impedance says nothing. We need at least the lowest impedance to estimate which amplifier works technically.


Cables and specifications? Yes… specifications play a role there too. After all: materials have properties. Think of conduction, resistance … and in the case of cables also capacitance and inductance, for example.

Below you can see some measurements on speaker cables. Especially the first graph is interesting, because it is measured on a real speaker. With a real amplifier. It shows that there is indeed a measurable difference in cables.

What we see in our own measurements is that the materials alone don’t say that much. What matters is the whole cable, with connectors. Because, if the connector does not conduct well, or the resistance between connector and conductor is too large, then a fancy, silver conductor does not make much sense: the connector just gets in the way.

You won’t see these things in the specifications. There you can only read how the cable is constructed: material, stranded, solid core, shielding or not and what kind of connectors are on it.

Sometimes they indicate whether it is soldered or not. However, as a ‘normal man/woman’ you will not know whether the material conducts well or not. And whether materials get in each other’s way…. Moreover, you do not yet have an understanding of impedance, reflections, inductance values, capacitance, et cetera.

These are things we are now diving deep into to see if there is a connection between these measurements and sound. It is too early to draw conclusions, but as soon as we know more, you will of course hear about it immediately.

However: specs of cables don’t say much – yet. Unfortunately. What would help is that manufacturers show basic measurements of their cables. In this way we can check whether the materials are correct. That is already more than we can see on the websites now.


Rounding up

We have touched on a few common specifications: wattages, bits & sampling rates / frequency response / speaker impedance and finally cabling. However, this is only the tip of the iceberg. There are (many) more specifications that are not or very easy, or even impossible to compare. If you have any suggestions: please let us know!

More clarity and insight is needed in the audio world. And also in other industries for that matter. In PCs, GHz’s are also not comparable without more insight about instructions per second or branch-prediction algorithms. And in the car industry, horsepower doesn’t mean much without insight into the power structure, gearbox ratios and the weight of the car.

Yes: complex matters, but relevant. And if it doesn’t interest you; you don’t have to read it: skipping is also an option. However, the purpose of this background story is to prevent you – the reader – from focusing on specifications. Above all, go listen and trust your ears: not just the specifications of a product.