St. Marks Arrays. <coeff sets 1 & 2, excel doc w/ steering values, array program, > 1. System Setup Wiring & Connection diagram...

Transcription

1 St. Marks Arrays Contents 0. Included Documents: <coeff sets 1 & 2, excel doc w/ steering values, array program, > 1. System Setup Wiring & Connection diagram Optimum Equipment Settings and Operation DSP Programming Important Notes on Programming the DSPs Current Program Connecting and Loading the Program Switching Between Presets (Live ) Modifying Settings, and Making Presets Testing Individual Speakers Other Notes FIR Coefficients Amplifier Issues & Repair Things to be done combining stereo channels to make a single mono channel (in dsps or otherwise) use balanced inputs to the DSP check phase of speakers in line array collaborate west-wall timing collaborated (equal) amplitude output of the amplifiers Prepared April 6,

2 1. System Setup 1.1 Wiring and Connection Diagram: The diagram below should help illustrate the way the system is set up and its inner workings. Keep in mind that the DSP units have eight channels in and eight channels out. Only the first two channels in should be connected on U1, and only seven output channels are connected to each individual DSP 1. Figure 1 basic wiring diagram of the system from the outs of the main mixer through the arrays. 1 There is one exception to this rule: channel 8 on U2 may in fact be connected to something feeding it an unaltered signal. But for our purposes this is irrelevant and at the time of this writing this wire is essentially a dead end wire. Refer to the large sound system manual at St. Marks for further information on this channel. Prepared April 6,

3 1.2 Optimum equipment settings and operation: DSPs Optimum settings: Settings within the DSPs cannot be altered without modifying their programming. That being said, they are currently optimized and steered with an angle of π/17 2, and should not need to be reprogrammed. However, in case of a system failure detailed instructions on how to reload or later alter the DSPs programming can be found in Section Amplifiers Optimum settings: The arrays output with maximum efficiency when the amplitudes each channel (on each of the arrays seven amplifiers) are equal. Thus, by first approximation 3 the volume knobs should all be in exact same position. It is also prudent that the knobs not be turned up all the way. This will prevent the amplifiers from clipping in normal operation, and serve as a safeguard to help protect the amplifiers from being overrun (overheating & more) and the speakers from being blown. In addition, leaving the amplifiers at a lower volume level will allow more control over the output volume from the main mixer 4. Turning the amplifiers on: When turning on or off the amplifiers one may experience a very loud pop from the speakers. As this has the potential to damage the speakers in the array, it is recommended that the mute button is active (the amplifier is muted) before turning on an amplifier. After a moment or two the amplifier can be un-muted and normal operation will proceed without the popping noise. Turning the amplifiers off should be done in the same way. First mute the amplifiers then power them down. 2. DSP Programming 2.0 Important notes on programming the DSPs: Interface cable: The cable needed to connect the DSPs to the computer is Null Modem Cable -- a Female-to-Female D9 (9 pin) serial cable with TX, RX, ground, and Handshake lines. The Figure 2 depicts how to make this cable. This cable should be connected to the backside of one of the DSPs, and to the computer s serial in. While another type of cable, a R232 cable, can also be used, one of the DSPs does not work with this cable and higher data rates are supported with the Null Modem Cable, and a Null Modem Cable is more readily available. 2 An Excel document is included in this package with the actual delay value for each speaker. The document can also be used to calculate new delay values. 3 Pending further evaluation it may be beneficial to play a sine wave through the DSPs and use a multimeter to measure the peak- to- peak amplitude of the AC voltage coming from the amplifiers (with loads) and carefully adjust the analog knob until the output voltage on each channel is exactly the same. The knob positions may be slightly off and not all pointing exactly the same way, (one may want to fix them in this position with tape). 4 The DSPs automatically attenuate the input signal by 20 db, however this is an adjustable value. Therefore the scale of volume control at the main mixer is interdependent with the DSPs attenuation as well as the volume level set at the amplifiers. Prepared April 6,

4 Figure 2 Null Modem cable wiring diagram. Network delay: The U1 DSP has an inherent delay when sending the input signal to the second DSP. This delay must be accounted for if the arrays are to work properly. To correct the phase difference all output channels from U1 are given an additional ms delay. On FIR Filtering: If FIR filtering is to be used with the arrays, it is important to note that each channel will be subject to a unique 5 time delay due to the filter in place. These delays must be accounted for before implementing steering delays. 6 Also, a side note: no presets are loaded with coefficients so all channels must have coefficients loaded (see sections below for how to do this) or output will be absent. 2.1 Current Program: The DSP program itself has many possible setup configurations. Many more, in fact, then can be supported all at once. For this reason, while a few settings can be altered as the DSPs are running (these are presets), many must be adjusted in the program and loaded 7 to be used. In the current program there are two primary configurations: one for pure array steering, Default, and another for when using FIR filtering, FIR. Every other setup is just a variation on these two base configurations. The configurations are best described by the wiring diagrams in the figures below. For the Default configuration the program for U1 is shown in Figure 3 and U2 is depicted in Figure 4. The FIR configuration is depicted in Figure 5 for U1, and Figure 6 for U2. 5 The length of this delay should be proportional to the number of FIR coefficients in use, by (1/48000) th a second per coefficient. 6 Ideally one should set all steering delays to zero and check with an oscilloscope that every output line is in phase. Once this is adjusted steering delays can be returned to normal values 7 To load the DSPs means to take them offline so that program code can be injected. During this time the DSPs will not function. Prepared April 6,

5 Figure 3 - U1 s Default configuration diagram. Figure 4 -- U2 s Default configuration diagram. Prepared April 6,

6 Figure 5 U1 s FIR configuration diagram. Figure 6 - U2 s FIR configuration diagram. Prepared April 6,

7 When loading the DSPs with either configuration, it is very important that both DSP units (U1 and U2) have matching configurations. If they are not both the same, the arrays will not work properly directivity will be lost. The program has many presets available for both configurations. Each preset corresponds to a different steering angle for the arrays. The presets available for steering angle are: (π/14)(π/15) (π/16) (π/17) (π/18). Loading presets 8 will automatically select which configuration is used, for both DSPs. An important note on the FIR configuration and presets: delay values due to coefficients are NOT loaded with presets, nor are the coefficients themselves! For FIR filtering to work both must be set! 2.2 Connecting and Loading the Program Connecting and loading the program can be quite easy if the steps below are followed: 1) Open up the SoundWeb program 2) From the File menu choose File Open 3) Navigate to the location containing St. Marks Adaptation SDF. 4) Highlight St. Marks Adaptation SDF by selecting it, and press Open 5) Now we will make sure the correct configurations are loaded. In the File menu, choose Window Network Diagram. This will bring up the window shown in Figure 7. Double click the rectangular box with title U DSP. In the upper left corner, of the resulting window, a drop down menu will allow you to select the configuration you wish to set. 6) Repeat step 5 but in place of U DSP double click on the rectangular box with the title U DSP, and double check that both U1 and U2 have the same configuration selected! 7) Plug the Null Modem Cable into the back of one of the DSPs 9, and into the serial port of the computer with SoundWeb installed. Make sure the DSPs are on and Ready. 8) In the SoundWeb program, select View Network Window in the File menu. 9) A connection window will open up, click the Go Online button to connect. 10) Once the connection has been established, click the Load All button, and in a few moments, the DSPs will be reprogrammed. 8 See section on 2.3 on how to load presets. 9 It doesn t matter which DSP both will work without consequence. Prepared April 6,

8 Figure 7 The Network Diagram window. 2.3 Switching Between Presets: Live While SoundWeb is connected to the DSPs it is possible to switch between various settings of each configuration. To do this select View Presets, highlight the configuration you wish to be applied, and then press the Recall button. Recall: the notes on the FIR configuration and presets mentioned above! 2.4 Modifying Settings, and Making Presets Changing settings is done in the DSP configuration window. To get to the DSP configuration window, select Window Network Diagram from the file menu, and then double click on the DSP you wish to edit the settings for. A helpful reminder: make sure both DSPs are using the same configuration before loading the program or saving a preset. Changing Delay Times: In the DSP configuration window, which should look something like what is shown in Figure 3 thru Figure 6, double click on any of the delay boxes to bring up a dialog used to set the delay time. Delay boxes are allowed a maximum delay length; by default all are set to 80ms, to change this right-click on a box and select Properties. Note: when adjusting the steering, see the included Excel document to calculate the delay times by inputting a steering angle. Changing Gain: Just like changing the delay times, open U1 s Configuration Window, and double click on the gain filter to bring up a dialog that can change the delay. This setting can be changed while live or online. Changing FIR Coefficients: To change FIR Coefficients (or make sure they are loaded properly) right-click on a FIR Filter box 10 and select Import coefficients, navigate to the proper 10 In a DSP Configuration Window. Prepared April 6,

9 directory, highlight the correct file, and click open. Coefficients should be imported automatically. Notice: FIR Filters have a number of coefficients associated with them. You will want to use the smallest possible number of coefficients that will fit a particular set. Change this in the FIR Filter properties box by right-clicking the FIR Filter and selecting Properties. The FIR coefficients included in this document set are numbered 1-14 and should be loaded to their corresponding speaker. Also pay attention to the number of coefficients in use, as each coefficient will add a (1/48000) second delay to the channel. For 100 coefficients the delay time of that channel will be (1/480) second. To correct for the delay you must take the maximum delay due to FIR filtering 11 and subtract the current channel s delay from the maximum delay. The time resulting is the necessary delay so that all channels are in-phase. After this, one should still use an oscilloscope to check that all outs of the DSP are in phase 12! Saving Setup as a Preset: Once you have configured all the settings the way you want them (delays, FIR coefficients, etc), you can save a preset by opening the Presets window ( View Presets from the File menu). In the Presets window select New, give the new preset a title, highlight the new preset, and then press Store. Your preset has now been saved. 2.5 Testing Individual Speakers The Default configurations for each DSP have an additional gain on each channel (see Figure 3 and Figure 4). This gain is set to zero for each channel but is in place for testing each speakers. While connected (running live) to the DSPs, open up the configuration diagrams and double click on the gain icons to open a dialog that is used to adjust the gain. It is recommended that each channel s gain dialog is opened and arranged in a logical order as in Figure 8 below. Once arranged, it is easy to mute all but one channel and quickly switch between them to run various tests. It is also noteworthy that one can use the gain boxes to invert a speaker s polarity by checking the Pol button. One last note: 11 The maximum delay time of all channels. If there are separate channels using 40 coefficients, 60 coefficients, 100 coefficients, and 120 coefficients, the max delay is given by the 120 coefficients filter. 12 Do this by setting steering delays to zero, and generating an input sine wave, while comparing two channels at a time. If there is a slight deviation, correct by increasing or decreasing the delays in each channel until they all match phase. Afterwards the steering delays can be reinstated. Prepared April 6,

10 Figure 8 Setup to test individual speakers. Each Gain control corresponds to an output channel. The top row all correspond to DSP U1, the bottom row corresponds to DSP U2 and the channels one to seven are read left to right. In this particular example only U2, channel 4 is switched on. 3. Other Notes 3.1 FIR Coefficients [ FIR coefficients sets are symmetrical; there are really only seven distinct sets for the arrays, each of which reflects about the center of the array. If testing FIR coefficients it is recommended to double check that each set has in fact been placed symmetrically. The goal of Prepared April 6,

11 FIR coefficients is to disperse the sound disk more evenly across the crowd with change in frequency 13. Included in this package are two groups of coefficients. One group is the original set of coefficients, the other a set derived for preferable/ideal disk dispersion. The frequency responses of each set of coefficients is displayed below; fig and fig show the theoretical and recorded 14 (respectively) responses, fig shows the theoretical 15 of the ideal set of coefficients. The theoretical set has a significantly greater attenuation relation and much smoother frequency falloff with each speaker the result of which should be a considerably more uniform wave dispersion compared to the original set. Unfortunately in order for the DSP to produce the desired frequency response a very large number of coefficients per channel must be employed too many coefficients for this set of DSPs to use, as they do not have enough processing power. The coefficients are still included in this package in case something changes in the future (perhaps upgraded DSPs), or for further exploration. ] Amplifier Issues & Repair At the time of this write up the 420 amplifiers have had ¼watt filter resistors 16 replaced with 1watt resistors to filter out/reduce some high frequency distortion. The previous ¼watt resistors had all been burned and decayed over time, which introduced this distortion. According to an almost identical amplifier schematic using the same 25watt transistors, nothing less than 1watt resistors should have been used there. Both full schematics for this design and the 420 amplifier can be found in the acoustics lab. 13 As frequency increases the sound disk narrows and becomes pointed and directed. This can be noticed at St. Marks when listening to an infinite loop of speech; even with steering one can notice a difference in sound between standing up and sitting down in certain parts of the church. 14 At the time the DSPs were measured there were additional gains in place on each channel by the previous regime. These gains scale the frequency response and thus differ from the theoretical response without gains. 15 It should be noted that this theoretical set of coefficients were calculated to inherently include proper gains, unlike the previous set, which should minimize deviation from ideal in practice. 16 One resister per channel. The resistance value is unknown at this time. Prepared April 6,