Amount of audio played out of jitter buffer - audio

I am looking at a tutorial on jitter buffer. It has the following diagram:
My understanding is that in case of adaptive jitter buffer, 'd' may vary over time. My question is that since a system is asking for audio data (in case of VoIP), will the amount the data given to the system by adaptive jitter buffer vary over time ?

Related

Is interrupt jitter causing the annoying wobble in audio using the mcu's dac?

I had a assignment for college where we needed to play a precompiled wav as integer array through the PWM and DAC. Now, I wanted more of a challenge, so I went out of my way and created a audio dac over usb using the micro controller in question: The STM32F051. It basically listens to my soundcard output using a wasapi loopback recorder, changes the resolution from 16 to 12 bit (since the dac on the stm32 only has a 12 bit resolution) and sends it over using usart using 10x sample rate as baud rate (in my case 960000). All done in C#.
On the microcontroller I simply use a interrupt for usart and push the received data to the dac.
It works pretty well, much better than PWM, and at a decent sample frequency of 48kHz.
But... here it comes.. When there is some (mostly) high pitch symphonic melody it starts to sound "wobbly".
Here a video where you can hear it: https://youtu.be/xD3uTP9etuA?t=88
I read up on the internet a bit about DIY dac's and someone somewhere (don't remember where) mentioned that MCU's in general have interrupt jitter. So may basic question is: Is interrupt jitter actually causing this? If so, are there ways to limit the jitter happening?
Or is this something entirely different?
I am thinking of trying to compact the pcm data send over serial (as said before, resolution of 12 bits, but are sent in packet of 2 8bits forming 16bits, hence twice the samplerate as the baud rate, so my plan is trying to shift 12 bits to the MSB and adding four bits of the next 12 bit value to the current 16 bit variable, hence only needing 12 transfers instead of 16 per 8 samples. Might read upon more efficient ways of compacting data for transport.), put the samples in a buffer and then use another timer that triggers at 48kHz for sending the samples to the dac. Would this concept work? Or would I just waste time?
For code, here is the project: https://github.com/EldinZenderink/SoundOverSerial

What do the ALSA timestamping function return and how do the result relate to each other?

There are several "hi-res" timestamping functions in ALSA:
snd_pcm_status_get_trigger_htstamp
snd_pcm_status_get_audio_htstamp
snd_pcm_status_get_driver_htstamp
snd_pcm_status_get_htstamp
I would like to understand what points in time the resulting functions represent.
My current understanding is that trigger_htstamp represents the time when stream was started/stopped/paused. snd_pcm_status_get_trigger_htstamp returns a constant value and when I add audio_htstamp to that value the result is very close to the current system time.
audio_htstamp seems to start from zero on my system and it is incremented by a value that is equal to the period size I use. Hence on my system it is a simple frame counter. If I understand ALSA correctly audio_htstamp can also work in different more accurate way depending on the system capabilities.
driver_htstamp I guess by the name is a timestamp generated by the audio driver.
Question 1: When is the timestamp driver_htstamp usually generated?
With htstamp I am really unsure where and when it is generated. I have a hunch that it may be related to DMA.
Question 2: Where is htstamp generated?
Question 3: When is htstamp generated?
Question 4: Is the assumption audio_htstamp < htstamp < driver_htstamp generally correct?
It seems like this with a little test program I wrote, but I want to verify my assumption.
I can not find this information in the ALSA documentation.
I just dug through the code for this stuff for my own purposes, so I figured I would share what I found.
The purpose of these timestamps is to allow you to determine subtle differences in the rate of different clocks; most importantly in this case the main system clock that Linux uses for general timekeeping compared with the different clock that determines the rate at which samples move in and out of the sound device. This can be very important for applications that need to keep audio from different hardware devices in sync, since the rates of different physical clocks are never exactly the same.
The technique used is sometimes called "cross-timestamping"; you capture timestamps from the clocks you want to compare as close to simultaneously as possible, and repeat this at regular intervals. There is usually some measurement error introduced, but some relatively simple filtering can get you a good characterization of the difference in the rate at which the clocks count.
The core PCM driver arranges to take a system clock timestamp as closely as possible to when an audio stream starts, and then it does a cross-timestamp between the system clock and audio clock (which can be measured in different ways) whenever it is asked to check the state of the hardware pointers for the DMA engine that moves samples around.
The default method of measuring the audio clock is via DMA hardware pointer comparsion. This isn't terribly precise, but over longer periods of time you can still get a good measure of the rate difference. At the start of snd_pcm_update_hw_ptr0, a system timestamp is captured; this will end up being htstamp. The DMA pointers are then checked, and if it's determined that they've moved since the last check, audio_htstamp is calculated based on the number of frames DMA has copied and the nominal frequency of the audio clock. Then, once all the DMA pointer update is done and right before snd_pcm_update_hw_ptr0 returns, another system timestamp is captured in driver_htstamp. This isn't meant to be used when you're using the DMA hw_ptr method of calculating the audio_htstamp though.
If you happen to have an audio device using the HDAudio driver, you can use an alternate and much more precise method of measuring the audio clock. It supplies an extra operation callback called get_time_info that is used instead of the default method of capturing the system and audio timestamps. It the HDAudio case, it takes a system timestamp for htstamp as close to possible to when it reads an interal counter driven by the same clock source as the audio clock; this forms the audio_htstamp. Afterwards, the same DMA hw_ptr bookkeeping is done, but the code that translates the pointer movement into time is skipped. The driver_htstamp is still taken right before the routine ends, though; this is "to let apps detect if the reference tstamp read by low-level hardware was provided with a delay" as the comment says in the code. This is because there's no guarantee that the get_time_info callback is going to take a new system timestamp; it may have previously recorded an audio timestamp along with a system timestamp as part of an interrupt handler. In this case, the timestamps you get might not match with the available frames and delay frames counts calculated by hw_ptr bookkeeping, but the driver_htstamp will let you know the closest system time to when those calculations were made.
In any case, the code is designed in both cases to capture htstamp and audio_htstamp as closely together as possible, and for htstamp - trigger_htstamp to represent the amount of system time that passed during the period measured by audio_htstamp of the audio clock. You mostly shouldn't need to use driver_htstamp, but I guess it might be used with the USB Audio driver, as I think it and HDAudio are the only ones that do anything special with these interfaces right now.
The documentation for this, although it doesn't contain all the details you might want to know, is part of the kernel documentation: http://lxr.free-electrons.com/source/Documentation/sound/alsa/timestamping.txt?v=4.9

gnuradio phase drift of AM demodulation

I am beginning a project using GNUradio and an inexpensive SDR.
http://www.amazon.com/gp/product/B00SXZDUAQ?psc=1&redirect=true&ref_=oh_aui_search_detailpage
One portion of the project requires me to generate a reference audio tone and compare the phase of that tone to demodulated audio.
To simulate this portion of the system, I have generated a simple GNUradio flowchart:
I had some issues with the source and demodulated audio in that they would drift relative to each other. This occurred on the scope sync on the original flowgraph. To aid in troubleshooting I sent the demodulated audio out thru the soundcard’s second channel and monitored both audio streams in addition to the modulated RF on an external oscilloscope:
Initially all seems well but, the demodulated audio drifts in relation to the original source and RF:
My question is: am I doing something wrong in the flowgraph or am I expecting too much performance out of an inexpensive SDR?
Thanks in advance for any insights
You cannot expect to see zero phase drift in anything short of a fully digital simulation, or a fully analog circuit with exactly one oscillator, because no two (physical) oscillators have identical frequencies.
In your case, there are two relevant oscillators involved:
The sample clock in the RTL-SDR unit.
The sample clock in your sound card output.
Within an GNU Radio flowgraph, there is no time reference per se and everything depends on the sources and sinks which are connected to hardware.
The relevant source in your flowgraph is the RTL-SDR hardware; insofar as its oscillator is different from its nominal value (28.8 MHz, as it happens), everything it produces will be off-frequency in an absolute sense (both RF carrier frequencies and audio frequencies of demodulated output).
But you don't actually have an absolute frequency reference; you have the tone produced by your sound card. The sound card has its own oscillator, which determines the rate at which samples are converted to analog signals, and therefore the rate at which samples are consumed from the flowgraph.
Therefore, your reference signal will drift relative to your received and demodulated signal, at a rate determined by the difference in frequency error between the two oscillators.
Additionally, since your sound card will be accepting samples from the flowgraph at a slightly different real-time rate than the RTL-SDR is producing them, you will notice periodic glitches in the audio as the error accumulates and must be dealt with; they will start occurring either immediately (if the source is slower than the sink, requiring the sound card to play silence instead) or after a delay for buffers to hit their maximum size (if the source is faster than the sink, requiring the RTL-SDR to drop some samples).

Capturing sound on Linux with low latency

I want to capture audio on Linux with low latency in a program I'm writing.
I've run some experiments using the ALSA API, using snd_pcm_readi() to
capture sound, then immediately using snd_pcm_writei() to play it back.
I've tried playing with the number of frames captured, and the buffer size,
but I don't seem to be able to get the latency down to less than a second
or so.
Am I better off using PulseAudio or JACK? Can those be used to play the
captured audio?
To reduce capture latency, reduce the period size of the capture device.
To reduce playback latency, reduce the buffer size of the playback device.
Jack can play the captured audio (just connect the input ports to the output ports), but you still have to configure its periods/buffers.
Also see Relation between period size of speaker and mic and Recording from ALSA - understanding memory mapping.
I've doing some work on low latency audio programming,
My experience is, first, your capture buffer should be small, like 10ms period buffer. (let's assuming you're using 512 frame buffer, and 48000 sample rate).
Then, you should config your Output device start_threshold to at least 2 * frame size ( 1 * frame size if your don't have much process of recorded data).
For record device, like CL. said, use a relative small period size is better, but not too small to avoid too much irq.
Also, you can change your process schedule to FIFO schedule.
Then, hopefully, you will get about 20ms total latency.
I believe you should at first ensure that you are running a Linux kernel which actually allows you to achieve low typical latency.
There are several kernel compile-time configuration options which you might look into:
CONFIG_HZ_1000
CONFIG_IRQ_FORCED_THREADING
CONFIG_PREEMPT
CONFIG_PREEMPT_RT_FULL (available only with RT patch)
Apart from that, there are more things you can do in order to optimize your audio latency in Linux. Some starting reference points can be found there:
http://wiki.linuxaudio.org/wiki/real_time_info

Real-time Audio processing - latency feasibility check

I have an application concept that required real-time audio signal processing that can be broadly described as: a) sampling incoming audio (from microphone), b) performing signal processing functions (such as filtering, fourier transform, filtering and manipulation, inverse fourier transform) c) play-out (via speaker jack)
I believe that the "end to end" round trip timing (a) to (c) would need be in the order of 2 to 5 ms for the application to work in the real-world.
So, my question is this possible on today's generation of iphones and android phones?
On iOS, it is possible, but not guaranteed. I have managed to get ~6ms (22050 sampling rate, 128 samples buffer size) in my iOS app which does real-time processing of speech input. Take a look at Novocaine (https://github.com/alexbw/novocaine) - which provides a nice encapsulation of Audio Units and makes programming easier.
However, keep in mind that even if you request a particular buffer size, at run time iOS may decide to send larger buffers at longer intervals (=> higher latency) based on resource constraints. For example, if you have requested a buffer size of 128 (~6ms), you MAY end up getting 256 size buffers at 12ms instead. Your app has to take this into account and handle the buffer accordingly.
On Android, unfortunately, low-latency round-trip audio is a much bigger problem. This is because latency is driven by a host of device/manufacturer driven factors like hardware/driver level buffers and these vary from device to device. You can find a discussion of this long-standing Android handicap here: https://code.google.com/p/android/issues/detail?id=3434
My suggestion would be to ignore Android for now, and implement/validate your signal processing algorithms on an iOS device. Later, you can consider porting them to Android.

Resources