What do we need to measure?
I recently watched a seminar on advanced tuning delivered by my friend and colleague, Steve Turrisi of JL Audio.
I greatly enjoyed it. Steve takes this sort of thing very seriously, which is all too rare in our field, and any technician would be well-served to watch this two-part presentation.
Plus, I think that “tuning” is “measuring the system and correcting the problems”, as I’ve heard it described (correct attribution is Andy Wehmeyer’s friend Ulrich, I believe). And it’s hard to describe Steve’s class as anything other than that.
That said, there were a few things I was unclear about, and a few things I disagreed with.
Soon after, I talked to Steve over the phone and cleared up the biggest questions I had about the presentation.
Fundamentally, I will describe Steve’s advanced process as “measure the impulse response, analyze the resulting amplitude and phase responses, then correct for amplitude and phase both”. Which sounds sensible, right? However, in the commercial car-audio environment, it’s not something I would rush into.
I also recently read Sound System Design and Optimization by Bob McCarthy. It’s all about pro sound, but it’s an interesting read for anyone interested in sound, even if around half of it isn’t applicable to what we do. Bob was a pro sound engineer who then got involved in the development of an impulse-response measurement system at MeyerSound. Steve acknowledges using this book as a resource often, and its influence on the content of his class is apparent.
How impulse measurements work
I’ve said before that fundamentally, real-time analyzers are analog AC voltmeters, and they measure the voltage at multiple frequencies at the same time. We can call this the amplitude, or the magnitude, of the content.
An RTA doesn’t measure time or phase, just amplitude.
Impulse-measurement systems measure the actual frequency response in both the amplitude and phase domains. This means that an impulse measurement system can tell you a numerical value for the amplitude and for the phase of a speaker system at any given frequency.
Phase is measured in degrees. Once you get to 360 degrees, you have traveled in a circle, and are basically back to zero degrees of phase. Engineers often refer to phase changes as “rotations” because of this aspect of phase.
We don’t hear differences in absolute phase. Play one speaker, and play with the phase. We can’t hear a quality difference. What does that mean? It means that most people can’t hear the difference, and those who could reliably hear a difference couldn’t reliably decide which one was “better”.
But play two speakers at near the same level, and if the sound from those speakers is out of phase at any frequency, there will be cancellations at that frequency – and those cancellations are usually audible.
Phase nonlinearity vs phase cancellations
When amplitude changes, phase changes. We aren’t sensitive to absolute phase, but we are sensitive to phase cancellations. Below a certain frequency, we are much more sensitive to these cancellations – that frequency seems to be around 1500-2000 cycles, but some believe it is slightly higher.
When can system-caused phase cancellations happen?
- Two drivers playing the same note – in the transition band of a crossover filter
- Two drivers playing the same note – front and rear speakers different distances
- Two drivers playing the same note – left and right speakers different distances
- Two speakers playing the same note – inadvertent reversed polarity
- Reflections, when the amplitude of the reflection is nearly as high as the direct sound, but the distances traveled are significantly different.
- When the OEM system is performing phase processing, and we didn’t notice. We assume all signals are in phase right and left at all frequencies, but that turns out not to be the case. We will consider this an OEM Integration topic, and only address the first five.
Let’s look at each of these in turn. In all five cases, the sounds must be nearly the same loudness.
- The transition band of a crossover filter: This is where two speakers overlap. The problem can be caused by crossover filter selection, but if the two speakers are different distances from the listener, the problem is magnified. The first step in preventing these cancellations is thoughtful crossover-filter choice. When choosing active crossover filters, I use 4th-order Linkwitz-Riley filters exclusively. The steeper filter slope means the transition band is very narrow, they are phase-neutral once levels and arrival times are corrected for, and they sum flat (whereas 4th-order Butterworth filters don’t sum flat, because they are literally a mistake). Steve advocated an approach which I will call what I’ve called it for years – “crossover whispering”. There is a lot of time being spent in this process deciding what crossover filters to use based on the interior of the car, and I’m not convinced there is any commensurate benefit.
2. Front and rear speakers playing the same notes, but from different distances: This is probably one of the reasons I’ve hated rear speakers for many years (my formative years were before I knew how to use delay). However, rear speakers are the best way to make a system louder, and that’s important to a lot of us and a lot of our clients. Just make sure the rears are time-corrected, or make sure they aren’t loud enough to ruin your stage. Many upmixer systems use rear door speakers primarily for stereo sound for rear-seat passengers, and rear-deck speakers or D-pillar speakers for ambiance channels.
3. Left and right speakers playing the same notes, intended for the center of the stereo stage, but different distances from the listener: These cancellations only affect sounds intended to appear in the center of the stage, which is a lot of sounds – including the drum kit in most recordings. Using delay to correct for these differing arrival times became a very popular tuning method around ten years ago, but it is only correcting the problems for one unique listening location. The lower you go in frequency, the bigger this problem gets, and the higher you go, the less pronounced this problem gets.
4. Inadvertent reversed polarity: It’s not as apparent with other speakers as it is with dual-subwoofer systems. Test your speakers with a polarity tester before tuning. It may seem like a rookie move, but you’d be amazed at how often this is the problem. Apply the science of quality control management to connecting speakers, and the statistics say that the odds are not in our favor.
5. Reflections: These are often the result of speaker locations, but it’s worth remembering that in many cases reflected sounds which are nearly as loud as direct sounds also travel a very similar distance as direct sounds, so the comb filtering effect is not as bad. When the distance traveled by a reflection is significantly greater, some attenuation also results. That said, most installations use factory speaker locations and accept whatever reflections are present as a result.
Delay is how we often address 2 and 3, and how we partially address 1.
Tape measures – do they work to set delay?
Steve stated that the best way to set delay is to measure phase, and set delay to correct it, rather than to use a tape measure. In our conversation, I double-checked that he wasn’t saying that tape measures don’t work for predicting delay. I will elaborate in a moment, but please understand that “tape measures don’t work” isn’t Steve’s position.
Steve’s position, as I understand it, is that you can correct for phase problems best after measuring the phase response of each driver, and comparing the various amplitude and phase responses.
This sounds reasonable. However, I find this method for car stereo to be difficult to perform, very difficult to teach, and inefficient compared to other approaches which start by measuring distance.
The big problem with delay as a phase cancellation correction method is that the delays you need are all interrelated – left and right drivers, front and rear drivers, and adjacent drivers (subs and midbass, for example, or mids and tweeters).
If you start setting delay on a basis other than distance, you can easily end up chasing your tail, as each delay change you make to address one problem causes other problems (either with adjacent drivers, or rear drivers, or drivers on the other side). This is not a problem if you initially set your delays based on distance. Now the relationships in the transition bands are correct, and the relationships in the off-center listening position are correct.
Far better, in my mind, to accept distance as the primary influence on 2 and 3, (front/rear arrivals and left/right arrivals) and as a major component of 1. (Transition-band cancellations are made worse in most instances by different arrival times). Otherwise, you end up juggling an awful lot of interrelated problems.
Measure phase, or just amplitude?
I’ve messed around a bit with measuring time and phase, microphonically and also electrically. I rarely do it. The biggest reason is that when I do this, I’m in a production environment, and I don’t have the time.
In answer to a question, Steve indicated that this can be done in 2 hours once you have some practice. I think he’s being very optimistic there. I might be wrong, but I’m pretty confident that it would take a LONG time practicing to get an impulse-response measurement-based tune down to two hours. I think it would take at least a full day for quite a while, and that’s a tough chunk of time to monetize. The number of guys in this industry who could do this consistently in 2 hours, I bet I could count on the fingers of one hand (I wouldn’t be on the list).
This is a commercial activity, not a contest. At the other end of the spectrum, we’ve got professionals suggesting tuning methods which don’t even use an RTA – which use an SPL meter and a bunch of test tones – because “many shops have loud compressors and impacts going and can’t use an RTA with any degree of accuracy”. That is not my position. We need to measure sound, even if we have to wait for our window.
My approach is this – if there’s a phase cancellation we can hear, there’s a phase cancellation we can SEE when we measure the system’s amplitude. If you can’t see it in the amplitude response, it’s not much of a cancellation. I’ve watched Steve explain phase cancellations twice now in trainings, and each time he shows the summed amplitude measurement, the cancellation is plain as day on the amplitude measurement graph (the same measurement you would see on an RTA). It seems really apparent to me that you can still address that cancellation and confirm the result in a quantified manner without the actual phase quantification.
Play Channel A. Play Channel B. Play both at the same time. If there’s a dip with both, but not with either, you have a phase cancellation problem.
Now, if I were designing a speaker system for an OEM project, or a home speaker, I would TOTALLY use the full-power impulse-response approach. Systune, SMAART, and REW are all pieces of software that can do it. They are not simple, and I speak as someone who spend a lot of unbilled hours over the last 15 years figuring out how to tune cars with test gear. I don’t think I want to start trying to convince people to sort this out on their own time.
We have a long road of adoption just to get one-channel real-time analyzer systems into most shops and get technicians comfortable with them. That’s my goal.
Now perhaps I’m wrong. As Bill Murray said, “If I’m wrong, I’ll go to jail – peacefully, quietly, I’ll enjoy it”. But for now, my belief is, learn how to get good results with one channel, with amplitude measurements, with a defined process, and in a minimal amount of time, FIRST. When you feel you’ve reached the limits of that method, then start looking at where you want to go.