Precision measurement and adjustment of time alignment for speaker (SP) units
Part-1: Precision pulse wave matching method
Hello friends,
Abbreviations in this post;
SW: sub-woofer, Yamaha YST-SW1000 (L & R)
WO: woofer, Yamaha NS-1000’s 30 cm JA-3058
SQ: squawker, Yamaha NS-1000’s 8.8 cm Beryllium dome JA-0801
TW: tweeter, Yamaha NS-1000’s 3 cm Beryllium dome JA-0513
ST: super-tweeter, Fostex metal horn T925A
First of all, please note that this "time alignment discussion" here is limited to pure audio-only system, and excluding audio-visual system where you need "time alignment adjustment" not only for the SPs but also for visual images/movies.
You well know that, throughout this project thread, I have been using digital music players, such as JRiver, in PC, and feeding the digital signal in digital XO/EQ "EKIO" for crossover, and then sending the divided digital signals into DAC8PRO for multi-channel multi-driver multi-amplifier stereo music listening.
In the digital signal processing, we have so many buffers or latencies; JRiver output buffer, ASIO4ALL's I/O buffers, EKIO's processing buffer, DIYINHK USB ASIO driver's buffer, and so on. Consequently, it is not straightforward to exactly measure the "absolute delay" between the JRiver's "shout" and the final air sound kick-up by SP.
I usually set all the buffers in the digital domain in rather large size, so that I should not have any latency or delay problems; in our audio setup, we have no problem at all if all the bunch of the digital and analog signal (15Hz - 30 kHz) have identical amount of delay time from the signal origin at JRiver, and this is always the case in our digital (PC based) audio system.
The
relative delay between the SP units, or
"time alignment" in multiple SPs, however, is always one of the critical issues in audio system, especially the multichannel multi-driver multi-amplifier system, as you may agree.
I have been always take my attention and care on this issue, and in my very early posts
#18 through
#21, by using REW's wavelet analysis, I briefly checked that all of my SP units, super-tweeter (ST), Be-tweeter (TW), Be-squawker (SQ) and woofer (WO) have essentially no delay with each other, while my sub-woofer (SW) has 10 - 20 ms delay against the other SP units.
Now, I became really would like to establish my own simple, reliable and reproducible precision method for "time alignment" or "relative delay" measurement, and fine adjustment(s) if needed.
For this purpose, I prepared one 6-second signal consists of multiple rectangular tone-burst (8 waves) signals of various frequencies in exact timing series with time-zero 15 kHz marker signal at 3.000 s time position;
Each of the start-up (kick-up) time positions was set exactly 200 ms intervals, except for the lowest 31. 5 Hz pulse set after 400 ms from the preceding 63 Hz pulse.
I may record the air sound of this signal by using a measurement microphone, BEHRINGER ECM-8000, and an audio interface TASCAM US-1x2HR in my second PC for the air sound recording and analyses. Again, the buffers and/or latencies of the recording system should have no problem, since the time alignment measurement would be done on relative time distance from the zero-time-marker, for relative delay assessments.
If the recorded sound of this signal has exactly the identical tone kick-up timings after the 15 kHz zero-marker, then all the SPs should have no relative delay; I can read/find the tone kick-up time position in sub-ms precision by enlarging the specific time area of the recorded sound by Adobe Audition 3.01 (or Audacity).
I should be careful enough, however, the positioning of the measurement microphone. The sound velocity in 20 degree-C temperature (my listening room now) is 344 m/s, and this means sound travels 34 cm/ms (milli-second). I fixed the microphone at 1.5 m from the surface of my SPs so that the sound traveling differences from the SPs to microphone is less than 5 cm or less than 0.2 ms, securing ms level accuracy/precision in my SP time alignment measurements.
All the data shared in this post were recorded at 1.5 m from my left SPs, and please note that the right SPs gave exactly the same results.
I first applied this method for precision measurement of sound delay with my sub-woofer, Yamaha YST-SW1000, as I already knew it has 10 ms - 20 ms delay. I played the prepared signal by JRiver, together with using the flexible "solo" buttons of digital crossover software EKIO; the highest frequency (Fq) L-panel was in solo for 15 kHz zero-time marker sound to be sung by L-super-tweeter, and the lowest Fq L-panel was in solo for 31.5 Hz and 63 Hz to be sung by L-sub-woofer.
The recorded sound track as a whole was easily time-shifted to adjust the kick-up timing of the zero-time marker at exactly 3.000 sec so that the time sequence of the recorded track would be identical to the original test track.
Incase if SW sound has no delay against the zero-marker, then the 63 Hz burst should start at 4.800 sec, and the 31.5 Hz burst at 5.200 sec. The precisely measured time points, however, were at 4.815 sec and 5.216 sec showing the SW sound delays in 15 - 16 ms;
As I use SW for 15 Hz - 55 Hz Fq zone, now I could precisely measure and confirm that the SW sound delays in 16 ms.
Then, as shown in above diagram, I delayed ST sound (and TW, SQ WO) in 16 ms by EKIO's numeric group delay function to be fully matched with the SW sound. As you can see, after the 16 ms delay of ST, and hence 16 ms delay of the zero-marker, the SW sounds appeared at exact 4.800 sec and 5.200 sec, means perfect time alignment was achieved for SW.
Having this 16 ms delay set in ST+TW+SQ+WO sound against SW, I measured the 6 sec test signal with all the SPs singing together, and again recorded the air sound for precise analysis. It was quite easy precisely identifying each of the kick-up timing of the tone bust signals in ms (millisecond) accuracy using time-enlarged view of Adobe Audition 3.01;
As shown in above diagram, the kick-up positions of the tone bursts were exactly identical to those of the input test signal means there is no relative delay at all in millisecond accuracy between the five SP units, ST, TW, SQ, WO and SW; perfect time alignment for all the SPs was established.
Furthermore, since now I have the 16 ms delay adjustment between WO and SW, I am very much curious about the tone burst wave shapes before and after the delay adjustment at the Fq area where WO and SW are singing together, which is given by the 63 Hz tone burst signal.
I carefully recorded the WO-only sound, SW-only sound, and WO+SW sound of the 63 Hz 8-wave tone burst signal, after and before the 16 ms delay adjustment for WO;
We should note that the
"63 Hz, -10 dB gain, 8-wave excitation" for WO is a rather strong one, and as seen in the above diagram, WO gives one additional "inertia aftershock" followed by the proper 8-wave response.
On the other hand, the transient response of SW for 63 Hz tone burst signal is rather nice with 8 proper wave peaks as in the input signal, but followed by slower low-gain sound of inner air movements which can be easily understood in consideration of the unique "Helmholtz resonance" mechanism for bass boost with the heavy (48 kg) and rigid Yamaha YST-SW1000.
I will further share and discuss the "measured" transient characteristics of my WO and SW in my separate post(s) coming hopefully within a few weeks.
Here, it is just an accidental coincidence that the delay of 16 ms given to WO sound is identical to peak-to-peak interval of the input 8-wave 63 Hz excitation which is also 16 ms (8 waves in 128 ms period).
We can see and understand the shapes of WO+SW sound after and before the 16 ms delay in WO sound in the above diagram. In this case, the
"before delay adjustment sound" has considerably longer singing time than the
"after delay adjustment sound". In real music listening circumstances, much more complicated intermodulation may happen if there is significant delay (asynchronization) between WO sound and SW sound in overlapped Fq zone.
In any way, it should be better to have as perfect time alignment as possible throughout all the SP drivers. I decided, therefore, to have 16 ms delay settings in EKIO's output panels for WO, SQ, and TW+ST.
I am again very much impressed by Yamaha's original design and physical alignment of SP drivers in NS-1000's rigid cabinet with almost no relative delay between the Yamaha TW, SQ and WO.
Then, what would be the subjective difference in music listening before and after the fine "time alignment" tuning?
I should say the difference is minimal in my usual classical music listening, at least a few days after the adjustment, but today I felt tighter and rigid kick-up of large bass drum sound (ca. 33 Hz, ref. my post #650 for the spectrum) in the orchestra music, in Rachmaninoff Piano Concerto No.2 at the beginning and later-on in the third movement (after 23:47 of the video clip), even though it may be a placebo effect;
I feel the positive effects are more clearly heard in jazz music; especially the sharp kick-drums and low bass string sound which contains wide range of Fq spectrum including the finger and/or nail touching on the lowest (thickest) string;
and,
In my next post in a few days, I will share and discuss;
Precision measurement and adjustment of time alignment for speaker (SP) units
Part-2: Energy peak matching method
I don't think the process of disassembling is very difficult. I was worried more about not damaging the rubber ring more than denting/cracking Be driver. At one point I had to brak super glue joint I made with acetone, it wasn't an easy task but it worked. I am not sure if compressed air will get rid of the dust, but something I can try with NS1000M, other speakers are just too heavy to move down/up by myself.
