• Posted: 14/03/26

AD / DA Converters, Overlook them at your Peril!

The process of recording in the technology infused environment of today revolves around a melding of the analog and the digital domain. Computers are everywhere as they have been for years, and in most music production we use computers in various stages of our recording chain. The most integral part of our use of computers and the recording of music or audio is the process of making those two worlds work in relative harmony. This process requires analog-digital to digital-analog conversion and a slightly better understanding of what this means to our signal chain is the subject of the seventh installment of the SPS series.

What is Analog to Digital and Digital to Analog Conversion?

Really simply, analog signals consist of voltage information received through a microphone or other source. This voltage is then passed through the analog circuit and then passed to the A/D converter that “reads” this voltage and converts this voltage information into Binary code (0’s and 1’s). This Binary code can then be read by a digital device (i.e. your computer or digital audio interface) and then the computer outputs that digital data to the D/A converter to convert that Binary code back to voltage information. That voltage information is then output to your monitors for audio reproduction. That’s basically it….Simple eh? The movement of air by your analog source (voice, instrument et al.) is converted into voltage information, that voltage information is then converted into binary code, that binary code is converted back to voltage information and then at the end of the chain converted back into moving air….which is what we hear. smiley

What determines one A/D-D/A converter from another?

Now that we’ve identified a simple explanation for A/D-D/A conversion, let’s identify the factors that we should be concerned with when choosing a converter. If you take the basic process of conversion that we talked about in the previous section, and break it down it’s all about the accuracy of the “translation”. When we are looking for a converter we are looking for the unit that can most reliably and accurately translate our initial audio/movement of air to the end result, which is the movement of air by the speaker cone in our monitors. Look for the device that can keep the signal closest to the exact information that it received throughout the four step process of: audio to voltage, voltage to binary code, binary code to voltage, and voltage to audio. No converter is exact in this process because of analog signal sampling. The figure below helps to illustrate how this process works.

Analog Signal being digitally sampled - A to D / D to A Converters - Overlook them at your Peril

Looking at the figure you’ll see that the digital device samples the analog signal at different times. If you were to recreate the signal from the digital samples you’ll notice that there are parts of the analog signal that are lost. The digital samples don’t reproduce the entire signal, just the interpolations that get measured. You may be thinking…”Well if there are more digital samples then the analog signal will be more faithfully reproduced”. You would be RIGHT! So the converter with the higher sampling frequency capability has a higher propensity for more realistic conversion. This, of course, isn’t the only thing that sets apart converters, but this is definitely one of the reasons you’ll see that most converters today sample up to 192Khz. Sampling frequencies of this magnitude lend to more closely reproduced digital representations of the analog input.

Another factor closely related to the sampling frequency brings us to the relationship between the audio signal we are trying to reproduce and the non-linearity’s inherent in the conversion process. To try to be as brief as possible let’s delve into sample rates really quickly. CD quality audio or audio available for Digital download are rendered at 44.1/16 bit, but are often recorded at sampling levels approaching 96 kHz and resolution levels of 24 bit. Why do we record at these frequencies when the human ear only hears from 20 Hz-20 KHz? It’s all because of the Nyquist theorem. The theorem basically states that near-perfect reconstruction of the signal can be maintained only if the band limit of the sample is less than half the sampling rate i.e. ½ of 44.1 would be about 22 khz, 48 kHz would be 24 kHz, and so on. As you can see the higher the sampling rate, the higher the ½ band limit is. Theoretically this would lead us to be able to faithfully reproduce up to 24 kHz of bandwidth if we are sampling at 48 kHz.  But, we can only hear up to 20 kHz right? Um…yea. So what’s the point of the extra 4 kHz? That’s up for discussion. Some feel that an original recording able to faithfully reproduce upwards of 24, 44, and 48 kHz will sound better on final conversion down to 44.1 then an original recording at 44.1. Others disagree but this is a reason why you may want to be sure that your converter has the ability to record up to 96 kHz.

Edification side note: The other half of the bandwidth that is above our hearing range is where aliasing takes place, this is where the non-linear quantization (converting into numbers) error is taken. If you only sampled the audible bandwidth, that bandwidth would encompass the quantization error, and that error would be audible. This obviously would affect the sound of your signal. With the sampling frequency double the audible frequency this allows for the quantization error to be stretched across the entire frequency range minimizing the quantization error actually in the audible range. The anti-aliasing filters that are used are never completely perfect in their ability to take aliasing out of the audible range for perfect signal reconstruction after the conversion process, but it’s this tradeoff between over-sampling and the quality of the filters that play a large role in the outcome.

I/O – This is very basic if your converter is a dual unit with a digital audio interface, but something you definitely want to consider, the I/O of the unit. Be sure to get the unit that works for your studio equipment setup. I don’t do a huge bunch of recording, so I/O capacity isn’t a major deal for me, but if you are mixing as well as recording you want to be sure the unit has the phantom power units you need, and the I/O capable to record the amount of inputs as well.

THD+N levels are something you will notice on many spec sheets of most converters. The lower the number the better. This is the Total Harmonic distortion plus Noise number. This number relates the ratio of total harmonic content to harmonic distortion plus noise. If you see a number such as -113dB, this would tell you that the harmonic content you are recording should measure 113dB louder than the loudest distortion plus noise factor.

Jitter..Jitter..Jitter. This is another area to be concerned with when choosing a converter, and is created by converters having different clock systems close together. Take this image:

Audio Converter Clock Setup - A to D / D to A Converters - Overlook them at Your Peril

Noise on interface clocks don’t add jitter by themselves, but it’s important that the manufacturer keeps the support system clocks away from the conversion clocks to protect each from noise bleed. The conversion clock must be isolated to protect the integrity of conversion and to minimize jitter. The overall quality of conversion is measured in terms of E.N.O.B (the effective number of bits) which basically measures the quality of your signal minus the quantization error (we discussed earlier how this can be mitigated by oversampling) and that number converted into bits. That ENOB number is largely dependent on many of the factors we discussed earlier in the article and the quality of filters along the analog circuit, conversion clock and data I/O clock. Filters can introduce many types of distortion (pre-echo, ringing, ripple, truncation, and phase distortion) so the filter circuits should be of high quality.

Analog signal input and output circuitry is critical as well, look for a converter with good topology that has balanced inputs and outputs with low impedance values.

Summary

We introduced this article with a brief definition of what really happens during the conversion process. We hope that this, in addition to explaining the process, helped to illustrate why converters are so essential to the sound of the audio we deal with. Obviously, if we aren’t hearing an extremely accurate representation of what was recorded then we are hindered in our ability to translate anything consequential to our listening audience. I often think of converters as the unsung heroes in the signal chain. We all know it’s a big deal to choose the right mic for the source and to get relatively transparent monitors, but it’s often over looked how important the process of getting the mic signal to the computer, and ultimately to the monitors. Sampling rates, bit resolution, THD+N figures, analog filters, topology regarding conversion clock isolation and even I/O specs are all some of the most important factors when comparing converters with one another and choosing the best converter for your situation. Specifically on the sampling frequency conversation we also delved into why these numbers are important and how you may decide to use them in your recording situation. There are always other figures to consider, and even other factors that can contribute to the degradation of your signal, but we hope this take on the converters discussion has at least convinced you that it’s an interesting area that’s worth learning more about. It can have a meaningful impact on the clarity of your mix and the way it translates to your audience.

(By: Ivan-the-engineer)

 

**If you have any comments on this or any other article, catch us in the forum and let’s discuss! **

#yroftheengineer

left arrow                                                                                                                           right arrow