58

How much will reflected signals matter in audio applications (say between an amp and a speaker, or a pre-amp and an amp)? Mostly with regards to fidelity and not power transfer.

What are the different options of matching the impedance and their pro's/cons? This can be on the output terminal, input terminal, or modifying the cable?

eruditass
  • 944
  • 1
  • 7
  • 9
  • 17
    The answers depend on whether you're a electrical engineer or audiophoole. If the latter, we can mumble on at length about oxygen free cables, exra-phancy capacitors, and lots of other expensive nonsense you must follow while waving a dead fish over your amplifier during a full moon. – Olin Lathrop Nov 23 '11 at 21:21

3 Answers3

66

Impedance matching is not used in modern audio electronics.

  • A mic output might be around 600 Ω, while mic preamp inputs are 1 kΩ or more.
  • A line output will be something like 100 Ω, while a line input is more like 10 kΩ.
  • A loudspeaker amplifier will be less than 0.5 Ω, while loudspeakers are more like 4 Ω.
  • A guitar output might be 100 kΩ, while a guitar amp input is at least 1 MΩ.

In all these cases, the load impedance is significantly larger than the source; they are not matched. This configuration maximizes fidelity.

Impedance matching was used in the telephone systems that audio systems evolved from, and was (sometimes?) used in vacuum tube amplifiers, but even then, it's a trade-off between maximum power and maximum fidelity.

Transmission line effects don't apply. With a wavelength of at least 10 km (for 20 kHz), I think the most effect you'd ever see from reflection is some comb filtering (HF roll-off) with lines a few km long? But that's totally unrealistic.

Bill Whitlock:

Audio cables are NOT transmission lines. Marketing hype for exotic cables often invokes classic transmission line theory and implies that nano-second response is somehow important. Real physics reminds us that audio cables do not begin to exhibit transmission-line effects in the engineering sense until they reach about 4,000 feet in physical length.

Maximum power theorem doesn't apply, since:

Rane Corporation:

Impedance matching went out with vacuum tubes, Edsels and beehive hairdos. Modern transistor and op-amp stages do not require impedance matching. If done, impedance matching degrades audio performance.

For why impedance matching is not necessary (and, in fact, hurtful) in pro audio applications, see William B. Snow, "Impedance -- Matched or Optimum" [written in 1957!], Sound Reinforcement: An Anthology, edited by David L. Klepper (Audio Engineering Society, NY, 1978, pp. G-9 - G-13), and the RaneNote Unity Gain and Impedance Matching: Strange Bedfellows.

Shure Brothers:

For audio circuits, is it important to match impedance?

Not any more. In the early part of the 20th century, it was important to match impedance. Bell Laboratories found that to achieve maximum power transfer in long distance telephone circuits, the impedances of different devices should be matched. Impedance matching reduced the number of vacuum tube amplifiers needed, which were expensive, bulky, and heat producing.

In 1948, Bell Laboratories invented the transistor — a cheap, small, efficient amplifier. The transistor utilizes maximum voltage transfer more efficiently than maximum power transfer. For maximum voltage transfer, the destination device (called the "load") should have an impedance of at least ten times that of the sending device (called the "source"). This is known as BRIDGING. Bridging is the most common circuit configuration when connecting audio devices. With modern audio circuits, matching impedances can actually degrade audio performance.

It's a common misconception. HyperPhysics used to show an 8 ohm amplifier output, but they've improved the page since. Electronics Design showed an 8 ohm amplifier output for a long time, but they've finally fixed it after a bunch of complaints in the comments section:

Therefore, unless you're the telephone company with mile-long cables, source and load impedances do not need to be matched ... to 600 ohms or any other impedance. --- Bill Whitlock, president & chief engineer of Jensen Transformers, Inc. and AES Life Fellow.

Community
  • 1
endolith
  • 28,494
  • 23
  • 117
  • 181
  • 15
    You're missing part of the big picture. Impedance matching was done in the old days *because loads and drivers were reactive.* In the example of audio, it was common for transformers to be in the signal path, and if you didn't match impedance, it simply would not work. You can get away with not matching impedance today because most modern electronic equipment has *resistive* inputs and outputs, not reactive ones. But for equipment where reactive components are in the signal path, impedance matching is still an important consideration. – Robert Harvey Nov 18 '10 at 17:31
  • @Robert: Can you give some example circuits? Seems like that would make a good answer. – endolith Nov 18 '10 at 17:44
  • 1
    @endolith: Anything with a transformer on both sides requires impedance matching. Granted, most modern electronic equipment has resistive inputs and outputs, so the goal becomes high impedances on inputs, and low impedances on outputs. That does not always produce ideal conditions, however; if you are building a microphone input for a mixing desk, you don't want a 10 megohm input impedance, because an input that sensitive will pick up all kinds of noise. Instead, you want something more along the lines of 10K ohms. – Robert Harvey Nov 18 '10 at 23:55
  • @Robert: That doesn't sound like impedance matching, though, it sounds like using a transformer to produce an impedance bridging connection. You're still using low output impedances into high input impedances, no? – endolith Nov 19 '10 at 00:13
  • @endolith: If your mixing console has transformers, the input impedance on a balanced input is going to be about 600 ohms, because that is what the impedance is for balanced microphones having transformers. The key is *resistance* vs *reactance.* In practice, nobody thinks about this stuff anymore because inputs and outputs on modern electronic gear are resistive, not reactive. – Robert Harvey Nov 19 '10 at 17:21
  • @Robert: Transformers aren't reactive, though, other than the leakage inductance, no? A unity-gain transformer connected to a 1 kohm load will still look like a resistive 1 kohm load. How would you match to the reactance of a transformer, and why? You should really post an answer with some circuits explaining this, I'd like to understand it. :) – endolith Nov 19 '10 at 18:46
  • I thought older amplifiers tended to have very high output impedances (since vacuum tubes modulate current rather than voltage), and a failure of the speaker to absorb the current that was being sent in its direction would result in the amp outputting a higher-than-expected voltage. I'll readily admit I don't understand vacuum tubes circuits very well, so tell me if I'm wrong. – supercat Feb 23 '11 at 23:05
  • @supercat: I don't know tube amps, but if they have a high output impedance, and you're driving a low speaker impedance, you can use a transformer to make the amp see a larger load impedance, and the speaker see a smaller source impedance. I still don't think they'd be identical ("matched"); the virtual source would still be made smaller than the load. But I don't know tube amps. – endolith Feb 24 '11 at 04:51
  • @endolith: All the tube amps I've seen have had output transformers. If tubes are inherently high-impedance, one could use a transformer to adjust the output impedance of the amp to be nearer that of the speaker, but trying to lower the output impedance would also cut the gain. Consequently, with a tube amp, I don't think one would normally want to try for near-zero output impedance. I wonder if output impedance might be part of the reason tube amps are reputed to sound different from transistor amps, since output impedance would change the accoustical response of a speaker. – supercat Feb 24 '11 at 15:32
  • @supercat: From what I'm reading, they *do* try to match impedances in tube amps. Since the source impedance is high and fixed, you need to match the load to it to get maximum power. "there are two sweet spots, one for highest power, one for lowest distortion; the two spots are not the same for any known tube" So a 30:1 transformer connected to a 8 ohm speaker would show the amplifier a 30^2*8 = 7200 ohm load, which would be optimal for a tube with 7000 ohm output impedance. http://www.radioremembered.org/outimp.htm http://www.geofex.com/tubeampfaq/TUBEFAQ.htm#XformImped – endolith Feb 24 '11 at 16:32
  • 1
    @endolith: I'm surprised I haven't seen the output impedance of tube amps mentioned as a factor in "tube sound", nor have I seen any discussion of designing amplifiers with higher "effective" output impedance to mimic the sound of tube amps. One wouldn't have to use power-wasting resistors to adjust the output impedance; I would think current-sensing feedback could to a pretty good job. – supercat Feb 25 '11 at 17:36
  • "This variable loading affects the amplifier's performance both because the amplifier has finite output impedance (it cannot keep its output voltage perfectly constant when the speaker load varies) ... The influence of the speaker impedance is different between tube amplifiers and transistor amplifiers, principally because tube amplifiers normally use output transformers and they tend to use low feedback." - [Wikipedia - "Tube Sound"](http://en.wikipedia.org/wiki/Tube_sound#Harmonic_content_and_distortion) Also: http://www.diyaudio.com/forums/solid-state/100799-current-mode-amp-speakers.html – endolith Feb 25 '11 at 20:26
  • 2
    Does any of this apply to Class D amplifiers (they have reactive outputs?) – finnw Jun 06 '11 at 21:36
  • @finnw: Hmm... yes, the basic principle applies. The inductive output of the Class D amp gives it a low impedance at audio frequencies (milliohms), so the energy all goes into the load, and a higher impedance (10s of ohms) at switching frequencies. The output impedance of an amplifier is very low if there's negative feedback, but some class D amps are operated without it. I think the feedback is usually taken before the LC, too. – endolith Jun 07 '11 at 14:40
  • Current feedback in solid state guitar amplifiers is old hat. For example, the pre-1980's Roland JC120 (Jazz Chorus) amplifiers used negative current feedback. From early 1980-something, they took out that circuit. A more recent example is the Marshall 8008. Schematics for all these can be found online. – Kaz Oct 02 '12 at 21:44
  • I modified an Alesis RA-100 "reference amplifier" for current feedback. I added knobs on both channels which can be used to dial in increasing amounts of current feedback while reducing voltage feedback. This works great for guitar. I cannot play that amp anymore when it's in voltage drive mode. Subjectively, effect is similar to putting stuffing in the cabinet: the tone becomes stiff, losing dynamics, sparkle and life. This is just as bad for clean tones as for distorted. Under current drive, there is a more "acoustic" quality to the cleans, and distortions have more dynamics and depth. – Kaz Oct 02 '12 at 21:46
  • @Kaz: Yes, you're intentionally decreasing fidelity. :) – endolith Oct 03 '12 at 02:06
  • Yes, that is true. I mean, all smileys aside, a resonating box also has bad fidelity. The sounds emanating from the different parts of a violin are not a high fidelity reproduction of the scratching of the bow across the string (thank goodness). – Kaz Oct 03 '12 at 02:14
  • 1
    @Kaz: Yes, but a violin is an instrument, not a speaker enclosure. The goal of instrument design is to produce a nice sound out of nothing. The goal of speaker enclosure design is to reproduce the nice sound that was originally recorded, without any alteration. (Unless you're designing guitar amps. Those are more like instruments.) – endolith May 02 '13 at 14:54
  • @endolith But guitar is what all my previous comments were about. Still, I don't agree that current feedback necessarily decreases fidelity. If we have a perfect setup: a voltage amplifier with speakers that are matched to it such that they have a flat frequency response, then messing with the feedback structure by adding c.f. will change the response. But the speakers could be designed around current drive. In a way the drivers already are because the field in the voice coil of a dynamic speaker is a function of current. – Kaz May 02 '13 at 15:06
  • +1 for use of the wayback machine to guarantee future visitors the same content. – BAR Jul 31 '15 at 02:35
  • @BAR No I used it because the content changed from when I originally saw it, but I remembered it being wrong. :) – endolith Jul 31 '15 at 03:17
  • Impedance matching matters when you are transferring power rather than small signals. You have to match the Ra's of a valve output stage to the primary of the output transformer, and the secondary to the speaker impedance, to get peak power transfer. You have to provide any calibrated source with its required target impedance in order to get correctly calibrated output. The BBC used to ship audio between London studios and a transmitter at Brighton at 600 ohms at high voltage to minimize noise, and they had to correctly terminate both end to avoid reflections. – user207421 Jun 21 '19 at 00:59
  • @user207421 1. Power tubes were addressed in previous comments. 2. London to Brighton is 90 km, far longer than the distances mentioned in the answer. – endolith Jun 21 '19 at 23:10
  • That does it, I'm switching all my speaker cables to 10BASE-T. – Robotbugs Mar 10 '22 at 20:25
  • @Robotbugs Are you going to bundle all the wires together per terminal? Speaker wires need to be much lower impedance than the speaker's impedance for fidelity. – endolith Mar 10 '22 at 20:47
13

Impedance matching isn't really a concern for audio frequencies and in your examples its not really preferred. However you do need to pay attention to your input and output impedances.

You generally impedance match for 2 reasons:

  1. Minimize Reflections - Reflections become an issue when the length of the transmission line gets into the same order as the wavelength of the signal. There are varying rules of thumb here. Some say worry when the length of the wire is 1/4 the wavelength, some say 1/6, 1/10 etc. It depends on the signal and reactance of the transmission line. In this case it really doesn't matter because the electrical wavelength of a 20khz signal is ~49,000ft. In other words reflections are a non issue for the application your asking about.

  2. Maximum power transfer - Matching the output impedance of a driver to the input impedance of a load allows max power transfer. At first this sounds important for driving a speaker but there are more important considerations (see below).

Amp example:

With a modern amp design (active power stage, no output transformer) your actual goal is the highest dampening factor possible, among other things. When you drive a speaker, the speaker itself actually generates current as its being driven, this should make sense considering that your driving a devices to move a coil inside a magnetic field. In the ideal case, this wouldn't matter since the cone/coil would react instantly to the incoming signal. In reality there is delay and overshoot of the cone due to the mechanical nature of the speaker. As a result the speaker produces currents that are sent back at the amplifier.

To put this in simpler, more applicable terms. A high dampening factor allows the amplifier to have better control over the speaker cone. This is especially important near the resonance point of the speaker. The dampening factor is (speaker resistance) / (amp output resistance) and some correction for wire resistance. So in this case your goal is the smallest possible output resistance in the amplifier.

line level between devices (pre amp):

Again impedance matching isn't the goal. You generally want the lowest output impedance and highest input impedance possible. This minimizes current draw and as a result voltage drop. This is the lowest distortion configuration and allows maximum voltage transfer.

Mark
  • 11,627
  • 1
  • 31
  • 38
  • 3
    "Matching the output impedance of a driver to the input impedance of a load allows max power transfer." Not quite. Matching the load to a *fixed* source impedance maximizes power transfer, but if you have control over the output impedance, you want it as small as possible to increase the power in the load. – endolith Nov 16 '10 at 20:36
  • 2
    I disagree, the value of the output impedance being lower doesn't increase power transfer unless you can lower the load resistance to match, it would just mean you would need a source resistor added to the mix to match the load resistance. Assuming the load is mostly resistive. Maximizing voltage on the load and maximizing power are two different things. – Mark Nov 16 '10 at 20:52
  • 7
    By that logic, you'd just make everything 0 ohms and get infinite power. :) Maximum power theorem only applies when the source impedance is fixed. In that condition, you make the load equal to the source to get the most power out of it. http://en.wikipedia.org/wiki/Maximum_power_theorem But if you have a fixed load (loudspeaker) and you can change the source's output impedance, you want to make it as small as possible. A 0 ohm source drives the entire supply into a 4 ohm load, while a 4 ohm source into a 4 ohm load would just waste half the available power. – endolith Nov 16 '10 at 21:39
  • 2
    One thing I've wondered about is how the output impedance of loudspeakers varies with frequency, and how the impedance-versus-frequency response tracks with sound-output-versus-frequency (at uniform voltage) response, and whether an amplifier could make use of such variations in trying to produce a flat frequency response for a speaker in a room. Any idea if that's been looked into? – supercat Feb 23 '11 at 23:08
  • 3
    @supercat Well, the impedance to SPL level is a complicated function and is definitely impacted by the speaker enclosure, driver mechanic features, and t-s parameters among other things. An average '8ohm' speaker will have impedance swings from ~3ohm to 50+ohms yet can achieve very close to flat frequency response. You wouldn't generally adapt a speaker to fit an acoustical environment as all rooms are different (especially at low frequencies). Room correction is done with EQ. Google Audyssey for the most popular automatic room correction system. – Mark Feb 23 '11 at 23:19
3

The seminal article on speaker cables was written by Bob Pease of National Semiconductors in 1990, titled "What's All This Splicing Stuff, Anyhow?". Read and enjoy - then get on with your life while safely ignoring the snake oil salesmen!

Nick Alexeev
  • 37,739
  • 17
  • 97
  • 230
uɐɪ
  • 3,538
  • 21
  • 27
  • I don't see anything in that article about impedance matching – endolith Nov 18 '10 at 16:14
  • The article is in reference to the last part of the question "...or modifying the cable". It also addresses the effects of connectors, reflections from impedance mismatch points etc. – uɐɪ Nov 19 '10 at 08:39
  • It would be better to include the relevant parts of the link in your answer, otherwise this is a bit of a link-only answer, and won't be much use if the link dies... – Greenonline Apr 23 '23 at 21:09