Dear Don,
Thank you very much for this detailed reply. I appreciate it!
Would group delay be more of a concern in a bi-amped system in which the line stage preamplifier is feeding a full-range signal to solid-state amps driving woofers and a second preamp output is feeding the same full-range signal to tube amplifiers driving midrange/tweeter panels?
What about phase shift? Can the phase shift of interconnects be measured?
How important is it for interconnects to have similar if not identical phase shift characteristics?
How much more of a concern is group delay and phase shift when solid-state amplifiers are driving woofers and tube amplifiers are driving midrange/tweeter panels?
Hey Ron,
GMKF answered for you (and me, thanks!)
Group delay is usually in the mud as far as audio is concerned and we are talking about wires. Where delay is a problem is usually larger venues when all the speakers around a concert hall, church, or stadium need to be delayed appropriately so everyone hears a cohesive sound. Signals in the wires travel very fast, from maybe 60% to 80% the speed of light, so that delay is not usually a problem. If the speakers are positioned at different distances, then the speed of sound in air comes into play, at ~1127 feet/second instead of maybe 93,000 miles/second or better (491,040,000 feet/second).
The other place group delay (latency, in this case) is an issue is when processing the signal. When you read about xxx thousand-tap filters, that means the signal going in is delayed by that amount times the sample period. At 44 kS/s, each sample is 1/44,000 = 22.7 us so a 1000-tap filter means a 22.7 ms delay -- roughly the same as moving the speaker about 25 feet! As long as all speakers are delayed the same amount, you will not know, but part of the magic of AVRs is getting the delays electrical and physical to even out so sounds from all speakers arrive at your ears at the same time. Avoids sonic chaos. At least from that source...
Group delay is related to phase shift; in fact, it is the negative of the derivative (slope) of phase with frequency. Constant group delay (just a single number, implying a fixed slope in phase over frequency) is desirable for pulse integrity. How much it matters in audio is debatable (of course) but I've always felt it a good goal. If group delay is not constant over frequency, then some frequencies will arrive at different times, smearing the signal (and potentially the image etc.)
If you have two amplifiers with significantly different delay (or phase shift) then yes you may need to compensate. I did that using analog circuits when I had a tube mid/upper amp and SS lower amp driving my Maggies. One other thing to watch is polarity inversion -- in my case, one amp inverted the signal with respect to the input, and the other did not. I found that out when I was trying to set everything up, and worked up a pretty nice all-pass circuit that would provide a full 360 degrees of phase shift without changing amplitude of frequency response. One of those engineer things; it was a fun challenge. I had it done and was testing it out when I realized I could have just flipped the speaker cables.
As it turns out, I needed it anyway, because the bass and tweeter inputs had a common ground and one of the amplifiers had a floating output that would not have worked if I flipped the connections and shorted the "wrong" side to ground through the other amp. At least I did realize that before I smoked an amp (maybe two, and a speaker or two).
Crossovers also introduce significant phase shift and is why it is most important to align say subs and mains at the crossover point so the signals from them add instead of cancel (subtract). As you move away from the crossover, one speaker contributes less and the problem diminishes. But, with typical AVR crossovers, there is still a lot of signal (about half-loudness in terms of SPL) an octave either side of the crossover, one reason I prefer to keep my sub's crossover about an octave above the -3 dB point of my main speakers. Phase shift typically gets larger below the -3 dB point, especially for ported designs.
Yes, phase shift is (and many other things are) very easy to measure in an interconnect. You can use an oscilloscope as mentioned above, but I use a vector network analyzer (VNA) that plots amplitude and phase for you. Most modern VNAs will also generate group delay and many will do an inverse FFT to give you the time-domain response (pulse or step) as well. A time-domain reflectometer (TDR) can also be used to measure the impedance along a cable and provide the delay by looking at when a pulse sent down the cable returns.
Most cables are well-matched unless something is very wrong, at least as far as audio is concerned. If the delay is different it means the length, resistance, capacitance, inductance, conductance, and/or all of the above are different. But again in the real world you'd have to have a very large difference in cables to hear anything. At work I have to match delay (skew) among cables to a picosecond or less, and sometimes well below, but that is multi-GHz RF stuff and not audio.
I am a skeptic when it comes to interconnect and power cables affecting the sound but there are cases where it happens, and usually the reason is rational and measurable. Shielding, or lack of it, can affect things. How ground is handled is significant. Some cables include passive networks and those can also affect phase as well as the effective cable resistance. Those intentionally modify the signal so are outside the scope of my discussion (babbling). I've always been curious why Teflon-type dielectrics are sometimes touted by the same companies that advocate lower sensitivity to charge traps and sensitivity to motion, both of which tend to be worse with PTFE (Teflon). Etc.
Lot to digest, I'll stop there for now.
HTH - Don