Understanding and Calibrating Your System: Convertors
Welcome back my friends for another round of system calibration! So far in the series we have covered everything from general setup, to basic calibration, to speaker designs, to bass management, to amplifier design. So what then is left? Our good old friends the converters of course!
This next part in the series is dedicated to analyzing just what exactly goes into these converters, why they have such an impact on the sound, and ultimately what we need to look for in them as audio engineers. And if you are worried that this tutorial will get too technical since it involves electronics, fear not. Everything will be kept clear, simple, but still detailed so that everyone can follow along!
So with that in mind, prepare for conversion!
Also available in this series:
- Understanding and Calibrating Your System: An Understanding
- Understanding and Calibrating Your System: Monitor Speakers
- Understanding and Calibrating Your System: Amplifiers
- Understanding and Calibrating Your System: Convertors
How Do Converters Work Anyway?
While engineers will talk all day about how this speaker sounds compared to that one because this one is ported, or how this tube amp is cleaner than that solid state amp because they are using higher voltages, you often do not hear engineers talking about converters in overly technical terms. You will always hear, "Well this one just sounded better!" And while that is arguably the most important aspect, most engineers often cannot explain why it sounds better or let alone what is going on inside the converter.
At its core a converter is going to take an analog signal (voltage) and generate a proportional digital number relative to the magnitude of the input voltage (in our case the amplitude of say our microphone signal). The process can also work in reverse in which we take our digital numbers and generate a proportional analog voltage; this in turn is what comes out of converters and goes to the amplifier. A converter that goes from analog to digital is known to as an ADC and a digital to analog is known as a DAC while a converter that can do both is a AD/DA.
But some of you may be wondering, how does it know the amplitude of my signal if it is constantly changing? The way we handle the constant changing signal is to quantize the analog signal into discrete individual samples so that we can generate our digital numbers over time. The problem with quantizing is that by turning our continuous analog signal into discrete samples we generate error because our signal will no longer be continuous but in individual steps; this is known as quantinization error.
However if we sample fast enough, our steps will be so close the original continuous sample that the error is minimized to the point of being negligible; this of course refers to sample rate. Also keep in mind that to properly record a sine wave at a given frequency, our sampling rate needs to be at least twice as fast as that given frequency; hence why we record at 44.1 kHz to give a frequency range up 22.5 kHz.
How Do Converters Work Anyways? - In Depth
While the above information may seem common knowledge to some readers, there are other aspects to the conversion process that are not so commonly known and are vital to understanding conversion. Keep in mind that this section will be kept very simple as the math that can arise is mind numbing!
First and foremost you will find that if you look into a converter we do not actually directly convert into PCM information. Instead the better chipsets (which thankfully is most of them now and days) start with what is known delta-sigma modulation. This form of conversion to be put very simply (and as a gross understatement), guesses as to what the next amplitude change will be relative to the previous input and how close its previous guess was. However it does it so fast (in MHz territory!) and by such small amounts that we end up with an extremely accurate representation of our signal.
So for example our input was 0.5 our guess was 0.6 and with that being said we were pretty close. So logically we are going guess somewhere around there for the next amplitude change. However our next amplitude shift was not 0.6 or 0.8 but instead 3.0 and we guessed 0.7! So to compensate our next guess is going to be around 3.0. While that error might seem high, remember we are sampling the signal many times faster even 192 kHz so that error is so quick it might as well be insignificant.
After we generate this insanely fast conversion we need to create a PCM data stream that our computers can more easily understand as doing processing with a direct delta sigma stream is very difficult and most software and hardware will not handle it. This is done through a decimation filter which turns our signal into 44.1 kHz, 96 kHz, etc.
Keep in mind too that we also need to employ a great deal of filtering to minimize and errors that do occur before conversion and after decimation. First and foremost in order to accurately digitize a signal without inducing aliasing is place an anti-aliasing filter before the delta-sigma; more commonly this is a low pass filter with a very high and steep cut off point. In addition, a high pass filter is ideally placed after decimation filter as the decimation filter can induce a dc offset that needs to be corrected.
A Case of the Jitters
If you were to crack open most converters you would see they are all running the same few converter chips from a select few companies (usually Cirrus Logic, Asahi Kasei, and Texas Instruments (who acquired Burr Brown). How is it then that different audio converters can sound so different in clarity? It usually comes down to the jitter.
Jitter is the tendency for a converter to deviate from its periodic signal, or more simply, it is error in the time domain. In order for us to accurately sample our incoming signal over time, we need make sure that one second is always one second, or more specifically, one sample is always the length of one sample. In order to keep things organized and regulated we need a clock to help ensure accuracy.
However this clock can drift overtime and when it does jitter errors are introduced in our signal as we start sampling the wrong part of the signal at that given moment. In order to regulate the clock either a crystal or a PLL (phase locked loop) is used to ensure stability and minimize jitter. The more jitter present in the clock, the more likely we are to induce amplitude problems and effectively start to reduce the bit depth of our signal and at the same time causes very subtle phase shifts which creates a more blurry stereo image.
Usually we will not notice this phase shift until we compare a converter against a better converter and then all of a sudden it becomes obvious (or in some cases not so obvious still!). This shifting around of the clock is also a primary contender for why some converters have better stereo imaging and perceived depth as they have severely minimized the jitter. Of course other parts of the signal chain play into this as well, but having a rock stable clock to minimize the jitter is key.
If all of this seems a little confusing think of jitter like this. If you try to take a picture with a camera in your hand, you need to really try and make sure your hand is still or you will get motion blur. However over time your hand may get tired and you may start to get more motion blur. Sure a picture every now and again will be nice and clear but the larger chunk of your pictures will get blurrier.
However if you were to use a tripod from the beginning then you would almost guarantee clearer pictures from the start. For audio, our converter is the camera, the jitter is the motion blur and the tripod is the ideal clock. Simple no?
Clocking on the Outside
So if the clock is so important to us, can we buy some ultra precise clock to control our converters? Yes you can! But should you? That depends.
There are many types of clocking found in the audio realm but more than likely we have all seen a BNC word clock input on the back of interfaces and converters. These are used to link together two pieces of gear and have them operate in sync with one being a master and one being a slave.
You can of course daisy chain them together but you start to lose fidelity when you do this. Instead, we can use a complete external clock with multiple outputs to clock together all of our digital gear. You will usually see these in post houses with video production equipment and various digital mixers, etc. that all need to operate on the same time scale. In this case the external clock is amazing as it will keep everything in sync and stable; sweet!
So why would we not want it? Because no matter how good that external clock is, it is not internal!
Internal clocks that are even just moderately good are far better than an external clock because it is difficult for a piece of gear to sync to external clock. Sure we can do it, but it will not sound as good as the internal unless the internal is very poorly designed. So unless you need to connect two or more devices, stay away from external clocks!
What to Look For
As with most all things audio related, being able to compare two or more units side by side is always your best bet. At the end of the day your ears and what they hear is the most important criteria for selecting a standalone converter. However most of us will not get the luxury of being able to do these comparisons in person. So what then should we look for in a converter?
Having the highest possible bit depth possible is paramount in a converter. You really should have nothing below 24 bits as with 24-bit converters we are able to push the noise floor to an extremely low level that should pose no problem for us during mix down. However keep this in mind, the theoretical signal-to-noise ratio (SNR) limit for 24 bit converters is -144 dB however the best chips in reality can only reach -120 dB! Now you see why bit depth is so important?
Multi-stage PLL and Noise Shaping
Another very handy feature set to find a converter is the use of multistage PLL and noise shaping. While using a single stage PLL is very useful, we are limited to certain bandwidths of jitter reduction depending on the design of the PLL. By incorporating multistage PLLs we can reduce jitter at different bandwidths and ensure a clearer conversion. In addition, another technique used by a few high end converters is the use of noise shaping. Essentially the jitter noise is modulated to a much higher frequency far beyond the audible spectrum and is then easily filtered out with a basic low pass filter.
While we discussed why it is both good and bad to have an external clock, having the option is always nice just in case. However if you really do not think you will have no need for an external then do not worry about it. However if you are setting up a post house or maybe a live rig with a lot of digital connectivity (which is becoming rather common now and days) then by all means make sure you have an external clock input
Determining what types of inputs and outputs you need can help you narrow your focus when trying to pick a converter. If you need a direct connection to your computer then obviously you will need USB, Firewire, or Thunderbolt. If however you have an internal PCI-e card then you can look at AES, ADAT, etc as additional options.
For those of you using a digital console like the Presonus Live series or the Tascam DM series, you could potentially use the digital inputs such as AES, ADAT, etc. to go directly from the board to the converter without having to go from digital to analog to digital again to back to analog. Pointless conversions should always be avoided!
So What is the Point?
Because conversion is arguably the weakest point in our signal chain; it is also the trickiest to hear how it affects of resulting signal. When we digitize a photograph with a scanner we technically are always loosing quality, but with a good scanner this degradation should be imperceptible.
Audio conversion is essentially the same idea but we use our ears and not our eyes. In addition, if we were to print our scan and then scan it again and repeat this process over and over we would begin to see the degradation more clearly with each rescan. With audio, the more we convert the signal to digital from analog and back the more we add noise and jitter to our signal and begin to cloud our stereo imaging and reduce dynamic range.
So make sure you get the best converters you can, and make sure you minimize the amount of conversions!
Until next time!