WilmaFAQ.txt v2.1.0 by SawTooth http://members.nbci.com/toothsaw Summary : --------- 1) The decoded sound is not fluid. I hear cuts. It's choppy. 2) I hear distortion in the decoded sound, or pops and clicks. 3) How do I change Wilma's AutoDetect Sensitivity ? 4) AutoDetection sometimes briefly switches ON or OFF when it shouldn't. 5) The Input Volume slider next to the VU-Meter does not affect the volume (nor the VU-Meter). The Mute button doesn't mute the sound. 6) Lip sync is not perfect. The words come out slightly after the actors have spoken. 7) Can the lip sync delay be measured ? 8) How does the Buffer Timing Form work ? 9) Can I reduce Wilma's CPU utilization ? 10) What is the use of the Priority setting? I don't see any difference when I change it. 11) How does Wilma handle the Mixer ? 12) I manually edited the Wilma.ini file, and now I think I'm having problems. 13) I have a WinTV NICAM card and can't get any sound with some TV viewers? 14) Volume of Decoded Sound is too low. What can be done ? 15) I'm using Win95 and I can't see any of the images on Wilma's toolbar. 16) In Wide mode, I get low volume sound. I have the sound card output connected to an external mono device using a stereo-to-mono cable or adapter. 17) How does the 'Tansmit AutoDetect' feature work ? *** FAQ: ---------- 1) The decoded sound is not fluid. I hear cuts. It's choppy. A) You need to increase the number of recording buffers. This number depends on the buffer size, sound card (ISA or PCI), CPU speed, and also on whether any 16-bit applications are running. Just slowly increase the number of buffers until the sound becomes continuous, then add 1 more buffer for security. Do not add many more buffers than is needed to have a continuous sound stream. Watch the number of 'Out Buffers Missed', and make sure it stays at zero for a minute or so, if not, add one more buffer, etc. ----- 2) I hear distortion in the decoded sound, or pops and clicks. A) The volume is too high, somewhere along the line. You can try to: 1) Turn on the Filter 2) Lower the recording volume (next to the VU meter) or press the Auto-Desaturate button. 3) Lower the WaveOut volume 4) Lower the Master Volume on your mixer However, if your TV sound output voltage or impedance is not matched to the sound card input you have chosen, you may still get distortion. Avoid using the microphone input for TV sound, if possible. If you do, you should turn off the microphone Auto-Gain from your mixer. ----- 3) How do I change Wilma's AutoDetect Sensitivity ? A) Two factors affect Wilma's AutoDetect sensitivity, or how fast Wilma will detect a change in the signal coding: The input sound's volume, and the amount of low frequency noise in the signal. Wilma's AutoDetection requires a minimum level of input sound volume. If a change of sound encoding is not being detected quickly enough, make sure that the VU meter is going at least above 20%-25%, preferrably much higher. However, if the volume is too high, and the VU-Meter is flashing Red all the time, this can also result in incorrect detections. Therefore, to get reliable autodetection, make sure the VU-Meter is showing a volume above 20% but below around 80%. As for signal noise, the default 'Reliable' setting is tuned to tolerate a large amount of low frequency noise in encoded sounds. If you wish to try and increase the sensivity, you should watch the Frequency Spectrum of the Input while listening to encoded sound. If the low frequency peak (to the extreme left of the graph) reaches or goes above -40dB, you cannot hope to increase the sensitivity without causing misdetections. If the low frequency peak is below -60dB you can try to increase the sensitivity. The other factor affecting AutoDetection speed is the number of Extra Confirmations. Every time a change in encoding is detected, an FFT detection is required to confirm the change. This is the case whether you choose the normal method of AutoDetection, or the FFT method. By default, Extra Confirmations are set to zero, meaning that only 1 confirmation is required. You could, if you feel like experimenting, increase the sensitivity, and if you get incorrect detections, try to increase the number of confirmations. This could result in an AutoDetection that is as reliable, and a bit faster. But it still depends on the quality and volume of the input sound. ----- 4) AutoDetection sometimes briefly switches ON or OFF when it shouldn't. A) If this happens when the incoming sound is not encoded, raise the input volume. If it happens when the incoming sound is encoded, lower the input volume. ----- 5) The Input Volume slider next to the VU-Meter does not affect the volume (nor the VU-Meter). The Mute button doesn't mute the sound. A) Make sure the sound card and input you selected correspond to the input on which your TV sound is connected. The Mute button should mute the original, encoded, sound, not the decoded sound. To stop the decoded sound, use the Decode or Run buttons. If you have 2 or more sound cards connected in a cascade (output of 1 connected to input of 2, etc..) make sure you select the last card's input (the card to which your speakers are connected). ----- 6) Lip sync is not perfect. The words come out slightly after the actors have spoken. A) Bad lip sync can be due to the program or movie you are watching, especially if it is translated (dubbed). However, there will always be some delay due to the recording-decoding-playback and multitasking process. Using an excessively large buffer size can produce large delays, but Wilma does not allow you to select such sizes. However, by choosing a much bigger number of buffers than is necessary to just obtain continuous sound, you will also produce longer delays. In general, try to use buffer sizes between 500 and 2000 bytes. The small sizes will result in shorter delays, and shorter interruptions when a buffer is lost every now and then, but will consume a bit more CPU time. If you select much more buffers than is required, the delay will grow over time. If you notice that effect, reduce the number of buffers to the minimum number needed to have a continuous stream of sound. If you're interested in the technical details, here is an example using 882-byte buffers (the Small setting) with an SBLive card : The decoder receives the sound buffers obviously after they have been recorded (it takes exactly 10ms to record one 882 byte mono buffer of 16-bit samples at 44100Hz). It processes these buffers very fast (around 0.2ms per buffer), then immediately sends them out for playback. Assuming the sent buffer is immediately played, this brings the total delay to around 10.2 ms, which is almost not noticeable. The above reasoning makes several assumptions: 1) That every 10ms recording buffer will be returned to us immediately after the sound it contains has been recorded. 2) That every 10ms playback buffer will be immediately played back. As you can guess, the above assumptions do not hold. Windows sound drivers use fixed-size DMA recording and playback buffers. On the recording side, they wait until their DMA buffer is filled, then use the data to fill our small 10ms buffers and return them to us. Hence we only receive recording buffers when the recording DMA buffer is full, at inervals corresponding to the capacity of that buffer. For example, on the SBLive, the DMA recording buffer is 5120 bytes (58 ms), so we receive 5 or 6 10ms buffers every 58ms. Similarly, on the playback side, the sound driver uses a fixed-size DMA playback buffer. On the SBlive, the size of this buffer is 11024 bytes (62.5 ms stereo). So the driver will not play the 10ms buffers we send to it, until it has received 7 such buffers. When we first start to record, we receive 5 buffers only, because the driver only has 58ms of sound, it can only fill and return 5 buffers. We then process and send out 5 playback buffers, but the driver needs 7 buffers in order to start playback. This means that output sound will not play back until we have received the next batch of recording buffers, 58 ms later. This brings the total initial delay to something of the order of 2*58 = 116ms. If this situation is maintained, our lip sync delay would be constant and equal to 2 times the capacity of the DMA recording buffer (for the case of the SBlive, whose playback buffer is bigger than its recording buffer). But this situation is not maintained over a long period of time, for the simple reason that the size of the input and out DMA buffers is not the same! It is rather complex to explain, but intuitively, you can guess that the delay will move over time. The more slack there is, the longer the delay can become. So to keep the delay within reasonable bounds, you should keep it on a tight leash. This means: 1) Avoid...
sumsara