Post by Michael Lawrence on Jul 9, 2018 12:44:22 GMT
This is a follow-up to this thread that examines the data from an acoustic treatment installation in a project studio space.
The studio owners subsequently purchased a system processor (the Ashly Protea 4.8SP) and I prepared a technical memo for them documenting the tuning process. The memo is more technical than I would usually get with a client because they are engineers. It is not a full summary, as showing every step and every graph would be very lengthy indeed, but it shows some of the workflow. I have simply copied and pasted the memo below. Comments welcome. Enjoy.
This document describes the system optimization work that was conducted on the monitor system at [studio address removed]. In the existing setup, a Presonus Central Station monitor controller switches the stereo DAW output between two sets of nearfield monitors, the Yamaha HS series and the JBL LSR series, both with subs. The existing configuration used the loudspeakers’ built-in frequency divider (“crossover”) circuits and thus could not be optimally configured to meet the needs of the space. Generally speaking, steeper (4th order or better) crossover filter slopes can reduce the frequency overlap range between physically separated sources and reduce lobing. In addition, there existed no capability of applying equalization or alignment delay to either system components or the system as a whole.
In the original configuration, two stereo feeds (Monitors A and B) left the Presonus unit and were patched directly to the subwoofers, which created high-passed outputs for each set of mains while mono-summing the input signals to derive the subwoofer input.
In the updated configuration, an Ashly Protea 4.8SP system processor was installed in the signal chain between the monitor controller and the loudspeakers. The Ashly unit accommodates 4 inputs and 8 outputs. The matrix allows each loudspeaker to be directly driven from any combination of inputs.
Table 1 shows the routing configuration.
Thus, each loudspeaker is still being routed the correct signal, but we have gained the ability to adjust EQ, delay, and gain on every input, along with EQ, gain, polarity, delay, and crossover filters on every output.
Fig 1 shows the routing configuration in the Protea NE control software, with outputs 4 and 7 for the subwoofers deriving their input signals from both stereo channels.
Dual-channel FFT analyzer software Smaart was used to capture a transfer function of the system in both frequency and phase. For brevity, only a small number of steps are shown.
We can begin looking at the frequency response of the left Yamaha loudspeaker. It is important to focus on large-scale trends, which are audible and treatable with equalization, rather than small details
in the response. Fig 2 (top) shows the unequalized response (black trace) while the bottom adds the EQ filters applied (green trace) and the resulting response (red), which is much closer to flat.
Below, we see the right channel, presented in the same way.
At the reference mic (engineer’s head position), the left channel arrival is earlier by about 0.17 ms, due to slightly asymmetrical loudspeaker placement, on the order of 2.3 inches inches. Adding 0.17 ms delay to Protea output 2 aligns the arrivals from both sides to the highest frequencies, as seen in the matched phase responses (below, bottom pane).
30° of phase offset at 20 kHz, as shown, means that signals from L and R arrive at the listener within approximately 4 microseconds (µs) of each other, equivalent to achieving a placement accuracy of better than 1⁄16”.
The JBL left speaker was then equalized:
And the right:
Here we can see the delay alignment effects in action. With no delay, the left channel arrives first (blue trace, shallower phase slope in bottom pane of plot), by approximately 90° at 1 kHz, or 0.25 ms.
Sure enough, adding 0.25 ms delay to output 2 aligns the two arrivals perfectly:
Next, the total response was viewed for one top loudspeaker plus the relevant subwoofer. Here, we see JBL right & sub. Any EQ changes near the overlap (crossover) region need to be done to the input rather than the output in order to preserve phase alignment between top and sub. Below, we see a 16o Hz cut being applied to the Input 2 EQ rather than having to do it at both outputs for the right loudspeaker and the subwoofer. This makes sense, as we’re now compensating for the combined effects of two sources, so we use an EQ that affects both.
Finally we see the final amplitude traces for both systems.
Quite flat and audibly improved. The 63 Hz – 160 Hz region is highly dependent on room acoustics, and was recently improved by adding acoustic treatment. For more information, see this post.
Note that the depth of the null around 110 Hz was decreased using the combined effects of treatment and equalization.
The studio owners subsequently purchased a system processor (the Ashly Protea 4.8SP) and I prepared a technical memo for them documenting the tuning process. The memo is more technical than I would usually get with a client because they are engineers. It is not a full summary, as showing every step and every graph would be very lengthy indeed, but it shows some of the workflow. I have simply copied and pasted the memo below. Comments welcome. Enjoy.
This document describes the system optimization work that was conducted on the monitor system at [studio address removed]. In the existing setup, a Presonus Central Station monitor controller switches the stereo DAW output between two sets of nearfield monitors, the Yamaha HS series and the JBL LSR series, both with subs. The existing configuration used the loudspeakers’ built-in frequency divider (“crossover”) circuits and thus could not be optimally configured to meet the needs of the space. Generally speaking, steeper (4th order or better) crossover filter slopes can reduce the frequency overlap range between physically separated sources and reduce lobing. In addition, there existed no capability of applying equalization or alignment delay to either system components or the system as a whole.
In the original configuration, two stereo feeds (Monitors A and B) left the Presonus unit and were patched directly to the subwoofers, which created high-passed outputs for each set of mains while mono-summing the input signals to derive the subwoofer input.
In the updated configuration, an Ashly Protea 4.8SP system processor was installed in the signal chain between the monitor controller and the loudspeakers. The Ashly unit accommodates 4 inputs and 8 outputs. The matrix allows each loudspeaker to be directly driven from any combination of inputs.
Table 1 shows the routing configuration.
Thus, each loudspeaker is still being routed the correct signal, but we have gained the ability to adjust EQ, delay, and gain on every input, along with EQ, gain, polarity, delay, and crossover filters on every output.
Fig 1 shows the routing configuration in the Protea NE control software, with outputs 4 and 7 for the subwoofers deriving their input signals from both stereo channels.
Dual-channel FFT analyzer software Smaart was used to capture a transfer function of the system in both frequency and phase. For brevity, only a small number of steps are shown.
We can begin looking at the frequency response of the left Yamaha loudspeaker. It is important to focus on large-scale trends, which are audible and treatable with equalization, rather than small details
in the response. Fig 2 (top) shows the unequalized response (black trace) while the bottom adds the EQ filters applied (green trace) and the resulting response (red), which is much closer to flat.
Below, we see the right channel, presented in the same way.
At the reference mic (engineer’s head position), the left channel arrival is earlier by about 0.17 ms, due to slightly asymmetrical loudspeaker placement, on the order of 2.3 inches inches. Adding 0.17 ms delay to Protea output 2 aligns the arrivals from both sides to the highest frequencies, as seen in the matched phase responses (below, bottom pane).
30° of phase offset at 20 kHz, as shown, means that signals from L and R arrive at the listener within approximately 4 microseconds (µs) of each other, equivalent to achieving a placement accuracy of better than 1⁄16”.
The JBL left speaker was then equalized:
And the right:
Here we can see the delay alignment effects in action. With no delay, the left channel arrives first (blue trace, shallower phase slope in bottom pane of plot), by approximately 90° at 1 kHz, or 0.25 ms.
Sure enough, adding 0.25 ms delay to output 2 aligns the two arrivals perfectly:
Next, the total response was viewed for one top loudspeaker plus the relevant subwoofer. Here, we see JBL right & sub. Any EQ changes near the overlap (crossover) region need to be done to the input rather than the output in order to preserve phase alignment between top and sub. Below, we see a 16o Hz cut being applied to the Input 2 EQ rather than having to do it at both outputs for the right loudspeaker and the subwoofer. This makes sense, as we’re now compensating for the combined effects of two sources, so we use an EQ that affects both.
Finally we see the final amplitude traces for both systems.
Quite flat and audibly improved. The 63 Hz – 160 Hz region is highly dependent on room acoustics, and was recently improved by adding acoustic treatment. For more information, see this post.
Note that the depth of the null around 110 Hz was decreased using the combined effects of treatment and equalization.