Interfacing The Line (Part 2)
The importance of impedance matching at line levels is widely misunderstood. This article attempts to clarify the situation.
In this edition Ben Duncan casts light on equipment matching at line levels.
Perhaps you recall our investigation into microphone mating (July 84 HSR)? If so, you'll be familiar with the idea of a loading ratio. To recap, good matching means transferring a signal from the output of one black box to the input of another, without unnecessary loss or degradation. In the microphone article, we used a car gearbox as a matching analogy, but this is conceptually poor. Why? Well an engine is a prime mover, a power source, so this naturally means gearing for maximum power transfer. Otherwise we'd be wasting petrol. But for audio, it helps to think of each unit as a measuring or sensing instrument. From this, it's obvious that our aim should be to transfer maximum voltage, because this is the primary parameter, the one we use all the time to describe and compare differences in signals within the audio system.
Putting this into vivid imagery, this means our equipment interconnections shouldn't be like a car's gearbox, where the revs (=voltage) loading is substantial, for the sake of making the most of the engine's torque. Instead, it should be more like the speedometer, a measuring instrument which obtains its information about output revs (and reads them out in mph, of course) from a magnetic coupling, and therefore has a negligible effect on the very revs it is trying to measure.
In a nutshell, the rule of thumb for (voltage) matching line connections between any equipment is that the load impedance should be 5 or preferably 10 (or sometimes even better, 20) times bigger than the source impedance. In other words, we feed a LOW (source) impedance into a HIGH (load) impedance.
Source impedance is often loosely called 'output impedance'. This is bad, because it doesn't tell us whether we're talking about the actual (internal) impedance looking up the output, or instead, the (minimum) impedance the output prefers to see. It could be either, so to be sure, we call the former the source impedance (also 'generator' or 'internal' impedance), whereas the latter is quite simply the minimum load impedance. (We'll deal with this one in the next section.
At the other end of the cable, the impedance looking up the equipment input is called the load impedance, because it's the load seen by the preceding source. Load impedance is usually synonymous with input impedance, but strictly, this is what the equipment possesses, rather than what is actually seen by the source. This distinction is only important in that the two can be very different when two or more units are hooked up in parallel across a single output - as we'll see in a moment.
The idea of a minimum load impedance derives directly from the definition earlier, of optional signal transfer; it's all about avoiding signal degradation. Here we're discussing a change in the qualities of the signal (as opposed to a simple reduction in the amount of signal) - which boils down to distortion. In some instances, this facet needn't worry us, because extra distortion caused by loading is insignificant unless the load impedance is well below the norm.
The problem with the loading ratio is that it's only a ratio; it doesn't define the absolute impedances. For example, the ideal 10:1 loading ratio could work out as a 50k source impedance feeding a 500k input impedance, or a 10 ohm source feeding a 100 ohm load. These extremes leave lots of scope for confusion and error, but the minimum load impedance acts to tighten up the scope by defining the minimum input impedance of the load equipment. Normally, input impedance is set somewhat higher than the source's minimum load impedance requirement, so two or more units can be linked across the output without any hysterics. Remember: if you link two effects units across a mixer send, and each has a 10k input impedance, the overall load impedance seen by the mixing desk will be (10k/2) = 5k. It's with this thinking in mind that most UK pro-audio gear has an input impedance of around 10k to 20k, whereas the minimum load impedance at line level outputs is typically 1k or 600ohms (Figure 1). In practice, this means we can hang up to eight or more standard 10k or 20k inputs across the output - ample for almost any contingency.
At the other extreme, the maximum source impedance is defined by the sum of the cable capacitance multiplied by the cable length, or more sensibly, the assumption that up to 10 or 20 metres of cable will be hung on the output at some stage, and that this spells such-and-such a source impedance for an accurate top-end signal transfer, up to 20kHz. For a full discussion on this, see 'Matching' June 1984 HSR. For the meanwhile, suffice to say that source impedances much above 1k are a nuisance when it comes to avoiding top-end signal loss.
If all studio equipment met UK pro-audio impedance standards, there'd be few matching hassles. Trouble is, most budget recording gear and effects units operate at higher impedances, principally to cut costs. The fact that the input impedances of budget gear are frequently higher than 10k doesn't matter so much, but the high source impedances - typically 1k to 20k of oriental electronics - are a royal pain. To demonstrate this, Figure 2 outlines the effects of hooking up a wide range of source impedances into a professional-standard 10k. As you can see, high source impedances result in a signal loss, not to mention an increase in distortion. Meanwhile, the tinted area outlines the source impedances most suited to a 10k input impedance - these are usually limited to professional equipment.
Figure 3 highlights matching results, from the angle of different load impedances, expressed against a 2k source impedance. This impedance is typical of many expensive, but nevertheless oriental cassette machines. This time, the tinted area highlights the much higher load impedances that budget equipment prefers to see.
At this juncture, let's not forget that many inputs exhibit double their rated impedance when operated in the balanced mode, ie. 10k unbalanced becomes 20k when wired for balanced operation. Given that it's OK to drive an unbalanced signal up a balanced input (See HSR Oct 84 pp.52/33), this is a good opportunity to ease the loading on oriental equipment. The results may include cleaner top-end, an increase in signal level and a small yet useful improvement in signal-to-noise ratio. Any small increase in signal level is all the more important when we remember that budget gear prefers to operate at -10dBu levels, so that make-up gain is already required to reach up to the higher line levels (eg. +4dBu) anticipated by pro-equipment.
You can now envisage the chronic effects of feeding an oriental tape machine into a 600 ohm load: the 2k to 600ohm (mis)match cuts the signal level by 13dB (Figure 3), so the total gain make-up from -10 to +4dBu is (13 + 14) = 27dB. Of course, this circumstance is extreme, and the loss can doubtless be made up, but unless you're careful, any extra gain will raise the noise level by the same amount if applied after the load. Alternatively, if the make-up gain is wholly prior to the bad matching, there's a dire risk of overloading.
One answer is to carefully split the make-up gain between the preceding and succeeding equipment until you're satisfied that the twin evils, noise and overloading, are balanced out. Another is to avoid a 27dB loss by being aware that the two items of equipment are at worst incompatible and at best, some active impedance matching is called for. But before going for this, perhaps ask yourself whether there's an alternative output available, with a lower source impedance. Here, I'm referring to oriental tape machines which commonly sport both phono and DIN outputs. More often than not the DIN output will exhibit a horrendously high source impedance, perhaps 47k or worse. The lesson is to stick to the phono outputs, because their source impedance is usually much lower, at around 2k to 5k (See HSR Feb 84, page 43).
When equipment standards don't add up, effective impedance matching means active electronics, more cost, more leads, connectors, and more batteries or mains plugs. This need not be an obstacle, provided it doesn't happen very often. If a black box helps to make a particular interconnection work, all is well, but if a lot of them are required, something has obviously gone badly wrong.
The equipment for active matching comes under several headings:
i) Active DI boxes
ii) Line amplifiers
iii) Buffer amplifiers
The exact purpose of the box doesn't matter, provided it meets three elementary criteria. First, the input impedance should be high. 220k is a good, general purpose figure which virtually any line source - however weedy - can drive without signal losses. But the active DI boxes made by Technical Projects for example, boast a 10M input impedance. That's almost 50 times bigger than 22k, but this needn't worry us, provided (i) it's tied down to a lower source impedance, and (ii), we keep our cables short and make sure they're fully screened.
When hooking up to very high impedances, always remember that the effective circuit impedance falls to the line feed's source impedance. Contrary to popular opinion, this means that correctly hooked-up high input impedances don't directly aggravate susceptibility to hums and buzzes. Aggravation is strictly the fault of excessively high source impedances, which fail to bring the line impedance down to a respectable level, ideally 1k or much less.
Second, the 'active box' needs to exhibit a low source impedance, ideally 100 ohms or less. Third, the equipment should be capable of driving +20dBu (8 volts) or more into a 1k load, or less. In other words the minimum load impedance should be 1k or less.
Not forgetting that most equipment needing this treatment operates at budget line level (-10dBu), whilst the equipment it's feeding is probably operating at 0dBu or +4dBu Zero level, some gain can also come in handy. Figure 4 outlines a suitable circuit. The maximum gain is +18dB; that leaves us with 4dB spare (-10 up to +4). A gain pot provides a continuously variable swing of 12dB (ie. - 6dB to +6dB), whilst the switch SW1 allows us to relinquish 12dB of gain, just in case the output levels are already high enough. The overall gain range is therefore -6dB to +18dB, which enables 99% of practical line levels to be accommodated.
In the early years of this century, Bell Telephone Labs in the USA laid down a matching technique that would persist for 50 years or more. In effect, they saddled audio with the dogma of 600 ohm lines and matching power transfer. Fact: a very long pair of wires, such as a 10 mile long telephone line, displays a characteristic impedance of around 600 ohms. Under these conditions, terminating the end of the line with a 600 ohm load, and feeding from a 600 ohm source resistance is perfectly acceptable, because it results in the least reflections. Without this power matching arrangement, the line's frequency would be poor and there'd be an excessive signal loss.
But the pioneers at Bell Labs made the curious mistake of assuming that even short distance interconnections between equipment exhibit a characteristic impedance, and therefore need the 600 ohm treatment (ugh!). This is not so. We only need to know about characteristic impedance when cable lengths approach a significant fraction (ie. 10%) of the signals' natural wavelength. For example, a signal at 20kHz has over ten times BBC Radio 4's 1500 metre wavelength; that's 15000m - or 8 miles. We can, therefore, safely dispense with the concept of characteristic impedance for all normal studio work, where even a 10 metre cable run represents only 0.06% of the wavelength at the highest operating frequency.
Nevertheless, the upshot of this is that 600 ohms source and load impedances became firmly established early on in audio interconnections. Mitigating fact no.2 : a transformer likes to see well defined impedances at its input and output, for a smooth frequency response and the absence of ringing (an overhang effect at high frequencies, prevalent on transient, percussive sounds). The transformer designer can choose to make the transformer match any impedance - so in the 1930s, what could be more natural than to specify 600 ohms for both source and load terminations? This was the sad consequence of letting telephone companies define audio standards!
Much later in the 1940s, perceptive sound engineers realised that power matching (ie. 600 ohms source feeds a 600ohm load) was pretty stupid, and led to grief whenever two or more inputs had to be hooked across one output. Moreover, 600 ohm matching made the design of passive filter equalisers and alternators unnecessarily tedious. No pocket calculators in 1949 remember! But rather than sweep away the whole practice outright, they chose to retain the 600 ohm source impedance. It is, after all, tolerably low. But the 600 ohm termination at the other end was abandoned, and replaced with a 10k bridging load. 'Bridging' is meant to imply a pair of wires that bridge across 600 ohm lines without loading down the 600 ohm source. Thus arrived 'voltage matching'. Amongst other things, this removed the 6dB loss at every interconnection, massively reducing the amount of make-up amplification required...
Today, 600 ohm lines are for the most part restricted to obsolete equipment, seen only on the secondhand market. Nevertheless, a large body of more modern pro-audio gear has 600 ohm transformer-coupled line outputs. By and large these should be terminated with a 10k load impedance. If the input impedance of the succeeding equipment is actually 5k or 20k (or similarly near the mark) all should be well. Significantly higher or lower termination impedances will upset the performance of the transformer, typically killing the low bass, or peaking up to the top.
The same malaise may be brought on by using long cables. Ditto, equipment with transformer coupled line inputs prefers to be terminated with a 600 (or sometimes 200) ohm impedance even though its own load impedance looking up the input is 10k. This is the original reason why so much pro-audio gear exhibits a 600 ohm source impedance, even though much lower source impedances have been possible at no extra cost since the early 1970s; the circuit's natural output source impedance, typically 0.5 ohms or less, is padded out with a series 600 ohm build-out resistor.
Given that modern outputs are short circuit protected, this is perfectly functionless, except when feeding a transformer-coupled input. So, this condition apart, you can usefully replace this resistor with a 100ohm component giving improved drive capabilities with no loss of protection.
Feature by Ben Duncan