Tweet
Why Use a Better DAC?
An argument for why using an external audio converter IC over one integrated on an SoC can help ensure high-quality audio performance.
By Dafydd Roche, Texas Instruments
My system-on-chip already has an onboard codec. So why should I bother with using external converters?
As an audio marketing manager, I get asked this question by many different customers. When we look at high volume consumer products like televisions, set top boxes and DVD/Blueray players, many system-on-chip (SOC) devices usually have integrated converters. On the surface of things, this looks like a great idea. Everything on one chip! However, not all is as rosy as it may look on the outside.
One of the biggest challenges semiconductor manufacturers face is balancing features and performance with the silicon process within which we make the products. Many of you may have heard of things like 90-nm processes, 45-nm processes, etc. These are all examples of the silicon processes within which semiconductor manufacturers try to make their new products. Each manufacturer has its own processes; each with its own advantages.
With the push to get more digital processing onto one small low-cost piece of silicon, the focus for all process designers is to push the digital transistor size even smaller. It does seem, however, that as we make the process smaller and tighter, the system's analog performance begins to suffer.
Making ADCs and DACs that exceed 96 dB of dynamic range is getting harder and harder in the small geometry silicon that many ASICs are made in. Consider a high-performance set top box (STB) processor. Inside a modern STB chip is a real-time video decompression processor, a microprocessor for the system, Ethernet subsystems, hard drive interfaces, and lots of other features. The majority of these blocks are digitally-based, and each benefit from the shrinking of the silicon process.
Mixed signal and analog system blocks don't always fare so well. As digital processes begin to focus on speed and size, the processor's real analog performance becomes a secondary concern. To many, digital is just overdriven analog with a very low signal-to-noise ratio (SNR)! What this means in the real world is that getting higher quality audio performance from a SOC is becoming quite difficult. Many SOCs manage around 90 dB+ dynamic range, but mostly by using a differential output to squeeze the extra 3 dB of performance.
In a cell phone system, 90 dB can be considered to be relatively good quality, however, in today's home entertainment environment 90 dB is seen as ancient technology. Modern AV systems typically provide a minimum of 105 dB of performance, with better systems specifying up to 120 dB of performance or higher.
Moving to external, higher performance audio converters does have its benefits in a home and portable environment. In a low-grade 96-dB DAC, you need each of the bits in your 16-bit world working at full scale to maximize the difference between your fixed quantization noise and your audio output.
Many systems on the market actually show users how many decibels they are attenuating the volume (just look at a typical AV receiver). For an idea of what the changes in 10-dB listening level is like, users report that a change of 10 dB is akin to doubling or halving the volume, depending on if you gain or attenuate. A gain of 20 dB is similar to quadrupling the volume. Quite a jump!
Any kind of digital volume control (attenuation) inside the DAC or DSP is going to use less bits to represent your signal, but still have the same level of noise. Suddenly, that "good enough" audio DAC has become equal to the DACs we used in the early '80s.
Figure 1 shows a regular 100-dB converter being fed a full scale CD signal. The noise level is below the dynamic range of the audio signal itself. Put another way, the quantization noise present in the audio signal is higher than the inherent noise in the converter. Now let's attenuate the signal by 20 dB in the digital domain by DSP, or digitally in the DAC. This equates to dividing the signal by 10. Now we see that the lower bits of our 16-bit, 96-dB audio will be masked by the inherent noise in our converter.
The reason I encourage designers to buy a converter that is higher performance than their source is that it provides headroom for the signal, In other words, it gives you a bit a breathing space between your signal and the minimum noise level in your converter. If you need to do 20 dB of attenuation, you can then do so without your audio suffering.
Figure 2 shows this principle. As you can see, even with 20 dB of attenuation, the full quality of the CD is still perceivable.
How does this affect the market?
Well, most home audio products that use Audio DACs currently do most of their volume control poorly in the digital domain, or use a motorized analog volume control so that it can be controlled remotely.
Changing volume in the analog domain has the advantage of decreasing the noise as well as the signal, thus, maintaining SNR. For higher performance audio systems, a range of programmable gain amplifiers is available to do just this. A simple serial word is clocked into the system, and on the next zero-crossing moment, the gain is changed.
However, for the majority of consumer systems where there is significant pressure to reduce the cost of build, keeping the signal out of the system noise floor is the key to getting better performance. A good DAC (in the 110-dB+ range) should be more than adequate to provide a good listening experience for all listeners, whilst providing the ability to attenuate a CD quality signal by 14 dB or more before any "audible" difference is heard.
That's all good for DACs, but what about ADCs?
The whole concept of using higher performance ADCs rather than the ones on your SOC are just as compelling, if not stronger than the reason for DACs. In today's home entertainment market, systems typically use 2 VRMS (5.6 Vpp) to transmit audio around the system. However, many systems still use +4 dBu (1.23 VRMS) and -10 dBv (0.316 VRMS). Such a span between 0.316 VRMS and 2 VRMS is around 22 dB! Assuming that you calibrate your ADC's signal chain to accept a 2-VRMS input. Any older -10-dBv signal that you bring into your system will be perceived as four times quieter (or around -20 dB below full scale). Typically, what you could do in this case is run a digital automatic gain control algorithm and increase the gain digitally by 20 dB or so. However, the downside of such an algorithm is that you'd most likely increase the ADCs noise floor by 20 dB, as well as the signal.
There are two ways around this problem, just like the DAC side of things. One is to use a better ADC (something at around 107 dB or more). The other is to use a programmable gain amplifier in the front end, to adjust the gain as required. A higher performance ADC will provide a low enough noise level, even when 10 or 15 dB of gain is applied.
Many ADCs on the market today include a multiplexer and a programmable gain amplifier. For example, the PCM185xA range of ADCs from Texas Instruments have a PGA with a span of 22 dB, with steps of 0.5 dB. That gain can be gradually applied when the system detects that a -10-dBv peak signal is present.
Where to from here?
The next step in the process is to look at the difference in pricing. A typical 105-dB DAC (PCM1754) would add an additional $1.30 in low volume (around 100 units). However, that would provide the designer with a clean single-ended output that could be driven directly to a line driver and headphone amplifier without an additional differential to single-ended amplifier. That alone saves the designer $0.41 in low volume (RC4580).
A better system for all?
Having a higher quality output for a net difference of $0.80 (in 100 unit lots) can really provide the opportunity to differentiate in a market that is looking for a product to stand out. Releasing products that are "just" good enough will put you in the same firing line as your competitors. At which point, your only differentiation feature is price.
So, in an attempt to slow down the race to the bottom... please... put some thought and care into the signal chain you are designing in. Your customers' ears, your sales teams' bonus and most of all, your company's bottom line will benefit greatly from the added revenue you can bring with a better product than your competition.
An argument for why using an external audio converter IC over one integrated on an SoC can help ensure high-quality audio performance.
By Dafydd Roche, Texas Instruments
My system-on-chip already has an onboard codec. So why should I bother with using external converters?
As an audio marketing manager, I get asked this question by many different customers. When we look at high volume consumer products like televisions, set top boxes and DVD/Blueray players, many system-on-chip (SOC) devices usually have integrated converters. On the surface of things, this looks like a great idea. Everything on one chip! However, not all is as rosy as it may look on the outside.
One of the biggest challenges semiconductor manufacturers face is balancing features and performance with the silicon process within which we make the products. Many of you may have heard of things like 90-nm processes, 45-nm processes, etc. These are all examples of the silicon processes within which semiconductor manufacturers try to make their new products. Each manufacturer has its own processes; each with its own advantages.
With the push to get more digital processing onto one small low-cost piece of silicon, the focus for all process designers is to push the digital transistor size even smaller. It does seem, however, that as we make the process smaller and tighter, the system's analog performance begins to suffer.
Making ADCs and DACs that exceed 96 dB of dynamic range is getting harder and harder in the small geometry silicon that many ASICs are made in. Consider a high-performance set top box (STB) processor. Inside a modern STB chip is a real-time video decompression processor, a microprocessor for the system, Ethernet subsystems, hard drive interfaces, and lots of other features. The majority of these blocks are digitally-based, and each benefit from the shrinking of the silicon process.
Mixed signal and analog system blocks don't always fare so well. As digital processes begin to focus on speed and size, the processor's real analog performance becomes a secondary concern. To many, digital is just overdriven analog with a very low signal-to-noise ratio (SNR)! What this means in the real world is that getting higher quality audio performance from a SOC is becoming quite difficult. Many SOCs manage around 90 dB+ dynamic range, but mostly by using a differential output to squeeze the extra 3 dB of performance.
In a cell phone system, 90 dB can be considered to be relatively good quality, however, in today's home entertainment environment 90 dB is seen as ancient technology. Modern AV systems typically provide a minimum of 105 dB of performance, with better systems specifying up to 120 dB of performance or higher.
Moving to external, higher performance audio converters does have its benefits in a home and portable environment. In a low-grade 96-dB DAC, you need each of the bits in your 16-bit world working at full scale to maximize the difference between your fixed quantization noise and your audio output.
Many systems on the market actually show users how many decibels they are attenuating the volume (just look at a typical AV receiver). For an idea of what the changes in 10-dB listening level is like, users report that a change of 10 dB is akin to doubling or halving the volume, depending on if you gain or attenuate. A gain of 20 dB is similar to quadrupling the volume. Quite a jump!
Any kind of digital volume control (attenuation) inside the DAC or DSP is going to use less bits to represent your signal, but still have the same level of noise. Suddenly, that "good enough" audio DAC has become equal to the DACs we used in the early '80s.
Figure 1. Digital volume control (attenuation) in a 100dB DAC.
Figure 1 shows a regular 100-dB converter being fed a full scale CD signal. The noise level is below the dynamic range of the audio signal itself. Put another way, the quantization noise present in the audio signal is higher than the inherent noise in the converter. Now let's attenuate the signal by 20 dB in the digital domain by DSP, or digitally in the DAC. This equates to dividing the signal by 10. Now we see that the lower bits of our 16-bit, 96-dB audio will be masked by the inherent noise in our converter.
The reason I encourage designers to buy a converter that is higher performance than their source is that it provides headroom for the signal, In other words, it gives you a bit a breathing space between your signal and the minimum noise level in your converter. If you need to do 20 dB of attenuation, you can then do so without your audio suffering.
Figure 2 shows this principle. As you can see, even with 20 dB of attenuation, the full quality of the CD is still perceivable.
Figure 2. Digital volume control (attenuation) in an 118-dB DAC.
How does this affect the market?
Well, most home audio products that use Audio DACs currently do most of their volume control poorly in the digital domain, or use a motorized analog volume control so that it can be controlled remotely.
Changing volume in the analog domain has the advantage of decreasing the noise as well as the signal, thus, maintaining SNR. For higher performance audio systems, a range of programmable gain amplifiers is available to do just this. A simple serial word is clocked into the system, and on the next zero-crossing moment, the gain is changed.
However, for the majority of consumer systems where there is significant pressure to reduce the cost of build, keeping the signal out of the system noise floor is the key to getting better performance. A good DAC (in the 110-dB+ range) should be more than adequate to provide a good listening experience for all listeners, whilst providing the ability to attenuate a CD quality signal by 14 dB or more before any "audible" difference is heard.
That's all good for DACs, but what about ADCs?
The whole concept of using higher performance ADCs rather than the ones on your SOC are just as compelling, if not stronger than the reason for DACs. In today's home entertainment market, systems typically use 2 VRMS (5.6 Vpp) to transmit audio around the system. However, many systems still use +4 dBu (1.23 VRMS) and -10 dBv (0.316 VRMS). Such a span between 0.316 VRMS and 2 VRMS is around 22 dB! Assuming that you calibrate your ADC's signal chain to accept a 2-VRMS input. Any older -10-dBv signal that you bring into your system will be perceived as four times quieter (or around -20 dB below full scale). Typically, what you could do in this case is run a digital automatic gain control algorithm and increase the gain digitally by 20 dB or so. However, the downside of such an algorithm is that you'd most likely increase the ADCs noise floor by 20 dB, as well as the signal.
There are two ways around this problem, just like the DAC side of things. One is to use a better ADC (something at around 107 dB or more). The other is to use a programmable gain amplifier in the front end, to adjust the gain as required. A higher performance ADC will provide a low enough noise level, even when 10 or 15 dB of gain is applied.
Many ADCs on the market today include a multiplexer and a programmable gain amplifier. For example, the PCM185xA range of ADCs from Texas Instruments have a PGA with a span of 22 dB, with steps of 0.5 dB. That gain can be gradually applied when the system detects that a -10-dBv peak signal is present.
Where to from here?
The next step in the process is to look at the difference in pricing. A typical 105-dB DAC (PCM1754) would add an additional $1.30 in low volume (around 100 units). However, that would provide the designer with a clean single-ended output that could be driven directly to a line driver and headphone amplifier without an additional differential to single-ended amplifier. That alone saves the designer $0.41 in low volume (RC4580).
A better system for all?
Having a higher quality output for a net difference of $0.80 (in 100 unit lots) can really provide the opportunity to differentiate in a market that is looking for a product to stand out. Releasing products that are "just" good enough will put you in the same firing line as your competitors. At which point, your only differentiation feature is price.
So, in an attempt to slow down the race to the bottom... please... put some thought and care into the signal chain you are designing in. Your customers' ears, your sales teams' bonus and most of all, your company's bottom line will benefit greatly from the added revenue you can bring with a better product than your competition.