A New Way of Looking at Speakers and Rooms - Review of APL TDA

Jack Regula

Freshman
Jan 27, 2017
16
2
3
[h=2]Introduction[/h] TDA, from Acoustics Power Labs, provides a new way of looking at the responses of speakers in rooms. Its Intuitive 3D Display of sound level vs arrival time and frequency spotlights speaker and room anomalies.

An impulse response shows energy arrival over time; TDA shows both the amplitude and the spectrum of that arriving energy over time. It is useful in both the design and fine tuning of the speakers themselves, especially their crossovers, and in integrating those speakers into rooms. TDA is the analysis step on a ladder of tools from APL that provide increasingly more powerful equalization solutions for adapting speakers to listening spaces.

TDA uses a sine sweep to obtain an impulse response that can be viewed by use of the IR and Log IR buttons on its control panel. The calculated impulse response contains non-linear distortion information that can also be displayed. Distortion can also be seen as a series of spikes before the main peak of the IR in the log IR view. Third party tool IRs can also be imported and viewed.

The obtained IR is then processed by TDA at 126 frequencies, 12 log spaced points per octave to create its unique 3D display - a 3-dimensional map of normalized sound pressure plotted with delay and frequency along the X and Y axes and SPL encoded in both color and on the Z axis. Boundary effects, room modes, timing alignment issues with multi-way speakers, and frequency response aberrations are easily recognized in this presentation. This 3D map is a very good tool for quickly evaluating a new speaker in a familiar space or a familiar speaker in a new room or position.
TDA also provides conventional frequency, delay, and phase response graphs. TDA attempts to separate the direct response from reflections based on time of arrival and succeeds in doing so down into the modal zone to an extent limited by room modes, near reflections and the increased difficulty of precisely determining the time of arrival of low frequencies. This time selectivity enables us to see both the direct response, otherwise viewable only as an anechoic or quasi-anechoic measurement, and the effect of the room. It’s not a coincidence that the human ear has the same/similar ability to separate direct sound from reflections and the same/similar limitations.

TDA’s frequency response graph, the AFR, shows this separated direct response while its 3D display allows you to judge the extent to which it stands above room and boundary effects. Where it is reflection free, it is minimum phase and thus can guide equalization; other products from APL use multiple AFRs taken at multiple positions to do just that.
Fig 1.jpg

Figure 1 TDA 3D Display of Soundcard Loopback Response

An ideal speaker would have a response as shown in Figure 1, the looped back response of a soundcard. A straight line along the frequency response axis representing the direct sound would indicate perfect time alignment of a multi-way speaker. The sharpness of this blade would indicate the absence of diffraction and near-in reflections. Delayed reflections would show up as lighter colors to the right of the blade in the area now dark blue. These are likely to be present in a real room measurement but not in a soundcard loopback.

In TDA, one can choose which time values to display. I’ve offset the display above by 10 ms. and limited the display range to a 30 ms. range. Unless offset, time values default to being relative to the arrival of what would be the impulse peak in a conventional IR. One has the option of connecting a reference loopback via a 2[SUP]nd[/SUP] soundcard channel if there is a need to know absolute time values – perhaps to compare sound channel latency for various configurations or FIR filters.

The Z-axis defaults to a normalized scale of 0-1 as shown here but can be set into a high dynamic range mode with a logarithmic scale and a configurable maximum value. The normalization sets the amplitude of the direct response line in the 3D display at each frequency to 1. The un-normalized direct response values appear in the AFR graph.

Just past the 25 ms. mark in this loopback response, you can see a ghostly white line parallel to the frequency axis. This was determined to be sound card internal crosstalk between its microphone preamp and line level output due to its monitor mixer circuit. I missed this ghost evaluating the sound card using conventional FR and IR measurements; had I noticed it then I would likely have a different sound card. With TDA you see things easily overlooked or misunderstood in conventional speaker measurement. [h=2]A Real Speaker at 1m[/h] A close in measurement is used when the goal is to evaluate a speaker or its crossover. Some room effects are visible in the 1m measurement shown below but the speaker does dominate. This is the response of a 3 way, full range speaker with near perfect time alignment and some crossover phase shift evident. The speaker is nestled tightly into a room corner, eliminating front and sidewall reflections but leaving it susceptible to floor and ceiling reflections that are indeed visible in the display. Some floor absorption was used near the microphone to limit the disturbance to the midrange response from the floor bounce. Except for this and some relatively thin carpet, the room is untreated. Delayed reflections from the back wall of the room are visible near the 38 ms. mark.

Fig 2.jpg
Figure 2 A 3-way speaker measured at 1m on axis in an untreated room

Because of the dominance of the speaker over the room modes, TDA correctly identify the direct response well down into the bass, although the direct response line does widen at lower frequencies. It’s easier to determine the frequency ranges affected by room modes and reflections by looking down on the 3D map from above. This is the TDA “pl” display option. Fig 3.jpg Figure 3 The "TDA pl" display of the same measurement

It’s important to be able to distinguish between the properties of the speaker and properties of the room in these displays. To do that we must first understand the speakers; in particular, that they are corner speakers and rely on supportive reflections from nearby floor and walls to support the bass and to a lesser extent the lower midrange.

The wedge of red level response embracing the black vertical line of the direct response from 50 Hz up to 1000 Hz is the result of reflections from the floor at a nominal delay of 2.8 ms and from the ceiling as well where the delay is greater than 6.9 ms. For the corner woofer, this is floor support and does indeed seem to be propping up the direct response line in the 3D display. The crossover between the woofer and the multiple entry horn that sits on top of it is at 350 Hz and shows in the PL graph as an abrupt narrowing of the boundary support wedge. This narrowing continues gradually as frequency rises due to increasing directivity of the speaker and the increasing absorption from the carpet and temporary floor absorber used for this measurement. Unfortunately, what was boundary support for the woofer becomes boundary interference for the midrange resulting in nulls in the 400 -500 Hz range in unsmoothed measurements taken at the listening position.

Above the bass, we are seeing the speaker and not the room; that is abundantly clear in the 3D display. The color shading close in to the direct response line at higher frequencies is due to speaker imperfections and can be improved by minimum phase equalization. Strong room resonances appear as the red streaks at 50 and 70 Hz and a weaker one at 180 Hz. In a steady state, the standing waves of a room mode would develop but in a transient sine sweep or impulse response they can only be called resonances. The vertical white streak at 38 ms. has a timing consistent with a reflection from the back wall of the room.

The AFR is the un-normalized frequency response recorded along the black direct response line of the 3D display. Resonances, or their effects, can be seen in it. To wit, the resonances at 50 and 70 Hz result in a peak at 50 Hz followed by a shallow null at 70 Hz.

Fig 4.jpg
Figure 4 The AFR, the un-normalized frequency response graph of the direct response line of the 3D display

To the extent that the processing succeeds in separating the direct response from reflections, the effect of reflections won’t show up in the AFR. In this measurement, supportive floor reflections have elevated the bass and lower midrange but no reflection nulls are visible.
These 3 way speakers use active, DSP crossovers. They each consist of a multiple entry conical horn containing a compression driver tweeter and four 4” midranges sitting on top of a 15” sealed, slot loaded woofer nestled tightly into each of the room’s two front corners. Time alignment was achieved by setting DSP delays to align the IR peaks of individual drivers with their crossover filters and PEQs enabled. Time alignment was then confirmed with TDA. The delay bend just past the mid to CD crossover at 950 Hz is due to the mid-tweeter crossover; it increases with increasing crossover filter slope. A room resonance or reflection also affects the response there. The only hint in the 3D display of the woofer-mid crossover at 350 Hz is the narrowing of the floor support wedge there, indicating excellent time alignment. Both XOs use only 12 dB slope electrical filters, achieving 24 dB acoustical slopes. Neither crossover shows a phase wrap in its conventional frequency/phase response graph.

Fig 5.jpg
Figure 5 The Speaker Being Measured
[h=2](continued in next post)[/h]



 

Attachments

  • Fig 1.jpg
    Fig 1.jpg
    155.1 KB · Views: 8
  • Fig 1.jpg
    Fig 1.jpg
    155.1 KB · Views: 8
  • Fig 4.jpg
    Fig 4.jpg
    170.8 KB · Views: 6
  • Like
Reactions: Raimonds Skuruls
[h=2]The Speaker Measured at the LP[/h] When the mic is moved out to the primary LP, 4m from the corner speakers, the room dominates the measurement, especially in the lower registers. The TDA display isn’t as pretty a picture there but then it’s an untreated room. Nevertheless, the direct response stands out down into the modal region.

Fig 6.jpg
Figure 6 TDA 3D display of the measurement at the LP in the untreated room

To diagnose the room modes and reflections evident above, we again look at the TDA PL display and at the AFR graph to see their effect on the frequency response. The pl display in Figure 6 is configured to high dynamic range mode with a 20dB logarithmic scale.

Fig 7.jpg
Figure 7TDA PL and AFR displays of the measurement at the LP

In this room, modal effects dominate below 200 Hz and a general lack of bass damping is evident. Without TDA, seeing nulls in the 100-200 Hz region in conventional frequency response measurements and misled by SBIR and room mode calculators that assumed a rectangular room, I initially believed those frequency response nulls were due to ceiling reflections. With TDA, I at first thought it was a more distant reflection. Gradually, I came to understand that it was a longitudinal room mode and that the listening position just happened to be sitting in a null of the mode’s standing wave. Moving the microphone just 2’ forward and out of the null gave a cleaner measurement, confirming the diagnosis.

Several reflections and resonances in Figure 6 were annotated and traced back to their sources by iteratively moving absorber panels and re-measuring. Marker 1 shows the impact of the longitudinal room mode just discussed. In general, the bass is elevated in part due to the modes and floor support but also as part of a house curve. That voicing needs to be redone based on a set of measurements taken over the listening window. Marker 2 points at reflections from objects on the front wall between the speakers, the flat panel TV and audio equipment rack. Marker 3 points at reflections from the unterminated conical horn mouth itself. Marker 4 shows multiple reflections or modes between 400 and 500 Hz, where floor bounce nulls occur.

Eventually, I placed bass traps in the corners above the speakers, which reduced the room modes dramatically. Using the AFR as a guide, I attenuated the modal peaks at 80 Hz and 180 Hz and obtained the vastly improved measurement below in which the direct response stands out well down into the bass.

Fig 8.jpg
Figure 8 Measurement with bass traps in place and attenuation of remaining modal peaks

Prior to TDA, I found even windowed measurement data taken at the LP in room overwhelming – too many frequency response nulls and impulse response peaks to make sense of easily. TDA helped me distinguish between room modes and reflections, to locate the sources of reflections, and immediately showed me the effectiveness of treatments I applied. With the speaker’s direct response now prominent in the measurements taken at the listening positions, I’m confident the room equalization process will result in a more than satisfactory listening experience.

(continued in next post)
 
  • Like
Reactions: Raimonds Skuruls
[h=2]Controlling and Configuring TDA[/h] The discussion so far has been focused on the unique2D and 3D maps of APL_TDA. Let’s look at its controls and see how else it can present measurement data.

Fig 9.jpg
Figure 9 TDA's main screen showing an IR log display

In the upper left hand corner of the main screen, TDA has shown me all the audio devices it has found on my system. Below it, I tell TDA which ones to use for input and output. Then I just click in the green “RUN MEASUREMENT” box and, after a delay for computation, get a measurement and a result display. Below the run button is a column of 4 grey buttons that allow the saving and subsequent importing of recorded measurement data and the impulse response computed from it. TDA automatically saves measurement impulse responses in a “HISTORY” directory below its executable.

Results appear in the large window to the right. Display options, as shown in the column of buttons immediately to the left of the main display window are: recorded measurement data, IR normalized, IR Log, the 3D and pl displays we’ve seen, AFR – the frequency response, DFR – delay vs frequency and non-linear distortion (NLDA). If the “SW” box is checked then pressing any of these buttons shows the curve in a new window from which the graph can be saved to memory in a variety of formats.
[h=2]Impulse Response Display[/h] Figure 9shows the IR Log display. The height of the peak above the noise floor shows the SNR ratio of the measurement. Reverberation time can be determined from the slope of the response envelope following the main IR peak. Zoom in/out controls are available. A linear IR graph is also available.
[h=2]Visualization of Reverberation Time[/h] Another way to visualize reverberation decay is with the PL display zoomed out to 300 ms. and configured for 60 dB dynamic range.

Fig 10.jpg
Figure 10 TDA pl Display Over 300 ms. for Visualization of Reverberation Time

(continued in next post)
 

Attachments

  • Fig 9.jpg
    Fig 9.jpg
    399.2 KB · Views: 4
[h=2]FFT-Q Window Analysis[/h] Frequency, group delay, and phase delay response curves can be viewed with varied FFT-Q window settings to remove reflection interference from the frequency, phase, and group delay response graphs.

The controls for FFT-Q window analysis are below the lower left corner of the main display screen. A low Q setting represents a narrow time window, better able to keep out reflections. A high Q keeps the window open longer making more detail of the measured response visible. The “FFT-Q” is roughly equivalent in effect to a frequency dependent window in other measurement tools.

With Q set to 8, the widest the window can open, we get this detailed picture of the AFR of a 1m measurement:

Fig 11.jpg
Figure 11 AFR of the 1m Measurement with Q = 8

Reducing Q to 1.8, the narrowest opening, results in this AFR from the same measurement:

Fig 12.jpg
Figure 12 AFR of the 1m Measurement with Q = 1.8

FFT-Q windowing also affects group delay, GDR, and phase delay, PDR, response curves. With Q set to 8, the curve is affected by the room to a high degree, despite the close in measurement, no doubt because of its use of boundary support in the bass and low midrange.

Fig 13.jpg
Figure 13 GDR, Group Delay Response of the 1m Measurement with Q = 8

Narrowing the window by reducing Q to 1.8, we see a curve more representative of the group delay of the woofer and its cross over and less of the room:

Fig 14.jpg
Figure 14 GDR, Group Delay Response of the 1m Measurement with Q = 1.8

The subtract minimum phase option has a further effect on the FFT-Q windowed response. Viewing GDR or PDR with a narrow window and minimum phase subtracted reveals the residual non-minimum behavior of the response, most often due to the crossover and/or equalizer.

Fig 15.jpg
Figure 15 GDR, Group Delay Response of the 1m Measurement with FFT-Q = 1.8 and minimum phase subtracted

The rising group delay in the bass in Figure 15 is in good agreement with the woofer simulation shown below in Figure 16 over at least the 20 to 100 Hz range. The small bump just past 1 KHz is due to the mid-CD crossover and might be eliminated with further development of the crossover. Using TDA’s GDR and PDR curves with minimum phase subtracted allows one to zero in on crossover phase and group delay using indoor measurements for the final stage of crossover tuning.

Fig 16.jpg
Figure 16 Simulated Woofer Group Delay

(continued in next post)
 
[h=2]Non-linear Distortion Analysis[/h] TDA can show harmonic distortion relative to the fundamental or as a percentage. Examples are shown below.

Fig 17 nlda.jpg
Figure 17 NLDA Graphs [h=2]Conclusion[/h] APL_TDA has become my favorite tool for evaluating my DIY speaker efforts in room, the listening position itself, and the effectiveness of various room treatments I’ve been experimenting with. It has helped me see issues that I missed or misinterpreted with conventional tools. I still use those programs for simulation and crossover development but TDA is where I turn to see (as opposed to hear) how well I’ve done in speaker and system design and implementation and integrating the speaker into my room.
 
[h=2]Probing the Limits of TDA’s Ability to Discriminate Reflections from the Direct Response[/h] To be useful for evaluating crossover timing using indoor measurements and in guiding equalization, TDA must be able to discriminate against reflections. What are the limitations on TDA’s ability to do this?

To explore this topic, we will look at a speaker both alone and in the presence of a strong reflection, delayed by 5 ms. and down only 3 db. For most small rooms, a 5 ms. delayed reflection would have traveled more than twice as far as the direct sound and would be down by more than 6 db. Thus, it is a worse than worst case situation and has been created artificially rather than by measurement.

The speaker has 4 ways with crossovers at 200 Hz, 1.6 kHz, and 5 kHz that use 4[SUP]th[/SUP] order LR filters. Measured close up and free from room effects, TDA shows a clean pl graph, with group delay increasing towards low frequencies and an impressively flat frequency response.

Fig 18.jpg
Figure 18 Reflection-free TDA pl

Fig 19.jpg
Figure 19 Reflection free FFT-Q AFR

Fig 20.jpg
Figure 20 Reflection free GDR, minimum phase not subtracted

The AFR and GDR above were taken with FFT-Q = 8. The GDR below is plotted with minimum phase subtracted, leaving the crossover group delay. The slight group delay ripple at 50 Hz correlates with the European AC mains power and is an equipment artifact, not part of the speaker’s response.

Fig 21.jpg
Figure 21 Crossover GDR obtained by subtracting minimum phase


Crossover frequencies are apparent in the PFR, phase frequency response:

Fig 22.jpg
Figure 22 Reflection free phase response

(continued in next post)
 
Next, we will look at the same waveforms in the presence of a reflection delayed by 5 ms. and down only 3 dB. The TDA 3D display shows twin blades at high frequencies, with the direct response blade bending towards the reflection as the mid-bass is approached.

Fig 31.jpg
Figure 31 Undistorted PFR obtained by subtracting minimum phase

The GDR and PFR obtained by subtracting minimum phase in the presence of the strong reflection match those obtained in the absence of a reflection. This provess that with TDA, one can evaluate crossover timing using indoor measurements in a reverberant environment.


***the end***
 

Attachments

  • Fig 23.jpg
    Fig 23.jpg
    169.2 KB · Views: 4
  • fig 24.jpg
    fig 24.jpg
    131.8 KB · Views: 3
  • Fig 25.jpg
    Fig 25.jpg
    175.3 KB · Views: 3
  • Fig 25.jpg
    Fig 25.jpg
    175.3 KB · Views: 3
  • Fig 25.jpg
    Fig 25.jpg
    175.3 KB · Views: 3
  • Fig 25.jpg
    Fig 25.jpg
    175.3 KB · Views: 2
  • Fig 26.jpg
    Fig 26.jpg
    164.7 KB · Views: 2
  • Fig 26.jpg
    Fig 26.jpg
    164.7 KB · Views: 2
  • Fig 27.jpg
    Fig 27.jpg
    134.1 KB · Views: 2
  • Fig 28.jpg
    Fig 28.jpg
    121.7 KB · Views: 2
  • Fig 29.jpg
    Fig 29.jpg
    163.2 KB · Views: 2
  • Fig 30.jpg
    Fig 30.jpg
    113.4 KB · Views: 2