This article sort of happened by accident. I have two headphone amplifiers on my workbench — the tube one that I described in my Oct. 19, 2016 article, Modding the Yuanjing 6N11 Headphone Amplifier … and a second solid-state unit. I also have a number of different pairs of headphone amplifiers. One day, I plugged my Sennheiser HD598 headphones into the tube amplifier and discovered that they sounded horrible. I was afraid that I had somehow blown them out, until I moved them to the other headphone amp, where they sounded fine again. This got me thinking: What was the difference between the two amplifiers that made the headphones sound so different?
It turns out there is a lot. First, the tube amp has an output impedance of around 30 ohms while the solid-state one is about 5 ohms. Also, the tube amp runs without negative feedback, but the SS one has about 8 dB of negative feedback, resulting in a big difference in how the audio changes when it goes through it.
I then tried different pairs of headphones with the two amps and again found that some brands sounded great with the Yuanjing (tube) amp, while others sounded their best with the solid-state amp. In particular, my Grado SR60s and Beyer DT990s sounded amazing with the tube unit and only fair with the solid-state amp, while my Sennheisers and Audio Technica ATH M50s sounded noticeably better with the solid-state amp. What was happening was a synergy between the different headphones and amplifiers.
It occurred to me that this could also apply to speakers and amplifiers, and perhaps this is what the subjective audio magazine reviewers might be hearing when they review different pieces of audio equipment. Then I remembered an old Stereophile article where Bob Carver (founder of Carver Audio — one of the preeminent American audio companies of the 1980s) claimed that he could make his newest amp sound exactly like any amplifier the editors chose — and he did it! (Check out www.stereophile.com/content/carver-challenge.)
He did so by matching the transfer functions of the two amplifiers. Transfer function is what happens to something as it passes through something else. In the case of an electronic circuit, it’s literally the difference between the output and input to the circuit.
Then my mind drifted back to my TV transmitter training. Back in the day, in UHF, the common devices to produce RF power were Klystrons and Inductive Output Tubes (IOTs). Both are nonlinear devices in that what comes out is far different than what went in… BUT — and this is an important BUT — it is easy to quantify the non-linearity and also easy to compensate for it by using a technique called predistortion. With predistortion, you deliberately alter the input signal so when it passes through the tube, its non-linearity is an exact compliment to the non-linearity of the predistortion. It is literally using two wrongs to make a right — because the two exactly complement each other.
Clearly, this is what was happening with the headphone amplifiers — the transfer function of the two amps were quite different, as were the transfer functions of the headphones themselves. Perhaps this is why things that measure identically on the bench can sound so different in the field.
In reality, we as engineers are doing this all the time — using equalizers to smooth out the audio response of a system is a form of predistortion. Understanding this concept explains a lot — for example, why do some microphones sound so different plugged into different mixers, or why do some boards sound different than others?
PUTTING THIS TO USE
So how can we use this to our advantage? For one thing, the air chain of a radio station is a bunch of equipment connected end to end. By analyzing what is going on with each piece of equipment and how each reacts with another piece of equipment in the chain, we have a powerful competitive tool.
In the “golden days” of audio, this was probably easier to do because every piece of equipment was designed for a 600-ohm input and output, so it was easy to simply connect these units together. Or was it?
Most old timers remember that when you put a 3 dB (or so) resistive pad between 600-ohm devices, it improved the sound by isolating the transformers in the units. Today, we use a matching system called “voltage distribution” with primarily op-amp inputs and outputs where the outputs are all low impedance (50–150 ohms) and the inputs all high impedance (2,200–10,000 ohms).
Can you see the potential situation here? There is a wide range in both. If we connect a microphone to a mic preamp that loads that mic with 250 ohms, is the mic going to sound different than when it’s connected to a preamp that loads it with 1,000 ohms? Absolutely. So, what’s the best solution? Spending $500 on a different mic or a couple of bucks on some resistors? I will let you decide.
I realize that you may already be thinking to yourself, “Horse Hockey! Dana is bringing subjectivity into something that we have looked at objectively for years.” But am I? I wasn’t imagining the difference in the headphones. And I also confirmed my hypothesis with several people, all of whom confirmed that they could also easily hear the difference.
By the way, this was also why the Orban Optimod 8000A was such a breakthrough product: By putting the entire audio processor and stereo generator in one box, Bob Orban was able to optimize how the different circuits within the unit “played” together. He used their synergies to make a truly breakthrough audio processor.
So how can we use this to make our radio station sound better? Let’s start with the microphone. We could make up a box with two pots that together increase and decrease the load on the mic and then listen to the difference (of course you should compensate for audio levels when doing this). You could try putting 2.7k load resistors across the inputs of your audio processor and listen for the difference (some op-amps like a bit of a load). You could even try different lengths of audio cable. This involves experimentation and listening — there is no right and wrong here.
So where do we go from here? That’s completely up to you. You can believe or not believe. You can experiment yourself. Just remember that every circuit has its own unique transfer function, which can either complement or adversely interact with the circuits both before and after it.
Dana Puopolo is chief engineer at WGLS(FM), Rowan University in Glassboro, N.J.
