The measuring rig — building it and the first measurements

 Now for the detail. Since the rig was only needed for comparative measurements against the competition (yes, yes, as if I could compete with anyone) and for relative figures, so as to see plainly how the drivers change as the cup design changes, no very high demands were made of it.

 For the measuring elements I picked WM-61A102CW microphone capsules — you know which marketplace from. In the reviews people swore they were genuine; in reality they are a good copy. They add no noise and the sensitivity is fine as well. At the same time I ordered a measurement microphone built on the same capsule; the reviews were positive. I bought it mostly for the future, to measure drivers separately from cups. And, to be honest, I liked its body, into which something of my own can be fitted if need be.

 The microphone preamplifier circuit is a bulldog crossed with a rhinoceros, put together from the common schematics found online in articles titled "building a measurement microphone". The main part is built around the simple and widespread NE5532 op-amp, and that part is, so to speak, standard. What I did not like was powering the capsule with the voltage set by the drop across a green LED, so I decided to use a TL431 to hold the voltage at 2.5 V. Also, for some reason in almost every schematic the output coupling capacitor of the preamplifier is fairly large (10 µF, 4.7 µF, 2.2 µF) and polarised on top of that, and in places there is a filter cutting frequencies as well. But with that much capacitance at the output the gain comes out weak, so by trial and error a non-polarised 0.68 µF capacitor was picked, which let me pull the line-in gain slider on the sound card down to zero. The NE5532 wants 15 V by the datasheet, but in practice 12 V is enough. The finished preamplifier board came out simple and compact.

 The rig itself is a box designed in an evening, with seats for the silicone ears and two channels for mounting the microphone capsules. On the front there are places for three-position switches that control the left ear, the right ear and the measurement microphone mounted separately. The whole thing sits on a stand, and on the lid there is a support arm the headband of the headphones rests on.

 The rig turned out rather well and adds almost no noise to the measuring path, which cannot be said of the sound cards themselves. I have a USB sound card on the not-so-great CM108 chip which, by the graphs, is far quieter than the one built into the motherboard, but by ear there is a clearly audible hum from the USB +5 V rail. So it had to go. The built-in card shows much higher noise on the graph, but by ear everything is better, and the SPL meter gave a perfectly acceptable result for my purposes.

 In the end the rig was connected to the line input of the built-in sound card and calibrated with Room EQ Wizard. Standing two metres from the PC, the rig easily picks up the noise of the quiet fan in the power supply.

 The first measurement

 The measurement was taken on the alpha version of the QymBuds Air earphones running over Bluetooth with the AAC codec.

 When I got that graph I froze for a few seconds and thought it had all been for nothing. The frequency response looks nothing like the ideal line the experts draw while insisting that this is what reference sound is. But once the first shock passed, a thought came: if the graph is so odd, why do the earphones sound so pleasant to me? And it is not only my own opinion — everyone who tested the earphones, wired and wireless alike, said the sound was very decent. It cannot be that all of them simply wanted to please me and not give offence.

 At that point I decided to turn to the internet and look at the frequency response of other brands (yes, that is not fair, because everyone has a different rig). And what did I see there? (I will not publish other makers' curves, to stay out of trouble, but you can easily find them yourself.) The response of other headphones is just as non-linear or worse, with noticeable dips at various frequencies, mostly between 1 kHz and 10 kHz. A full roll-off on the graph past 10 kHz. But taking the averages, my response is no worse and in places better than that of some famous and expensive models. And in general, a number of articles made it clear that a flat response in headphones is not compulsory (well, it is needed, but not by everyone — we are not sound engineers), and it all comes back to how it feels to the person listening. As the saying goes: do you want the taxi to look smart, or do you want to get there? It is the same with headphones — do you want a flat response, or do you want it to be a pleasure to listen to?

 Conclusions

 Draw them yourself, because nothing written in this note is absolute. You certainly cannot compare headphones this way, but I think it shows that QymBuds are turning out to be worthy competitors, at least in their price bracket. And that while they are made in limited batches.

 What next?

 Next come tasks no less interesting:

1. Measure headphones in the same price segment on my own rig, to get more objective figures on how the sound compares.
2. Build or buy an external sound card with its own power supply, ideally a transformer one with good filtering.
3. Take new measurements, play with the internal volume of the cups, improve the rig, and much more.