This is a review and detailed measurements of the vintage (?) Denon DA-500 DAC. It was kindly purchased by a member on the used market and drop shipped to me. Interest was to find historic products with something cool about them. In this case, it is the "alpha" processing in the DA-500 which interpolates samples and supposedly some version of it lives even today in Denon AVRs. Looking around, clean DA-500 DACs cost about $220.
Nice to see a wide, room chassis compared to cramped little DACs we see today:
Of course there was no concept of USB input at the time so the only inputs are Toslink and Coax:
Ah, the fun days of hi-fi when high-end gear could get away with an attached cheap mains cable!
The heart of the DA-500 is BurBrown (now TI) PCM1702 sign-magnitude dual-segment R2R DAC. For this reason, it seems to still be cherished today.
Denon DA-500 DAC Measurements
Let's start with our classic dashboard before we get fancy:
Interesting that it output higher than 2 volts. Likely a good number of sales were made compared to other DACs due to higher than normal volume!
SINAD which is the relative sum of noise and distortion is in "poor" bucket using today's standard, falling short of even CD's 16 bit dynamic range despite the unit sporting a 20-bit DAC. I am going to jump to linearity test now to tell us how this DAC works:
Whenever you see a zigzag like this that settles below 90 dB, it indicates that the DAC is simply throwing away extra bits in our 24 bit samples and is just using 16 bits. This type of conversion is bad as it generates steps in response that result in fair bit of distortion. Proper conversion with dither (noise addition) needs to be performed. This is important today due to existence of 24-bit content but at the time this DAC came out, the only source was 16 bits so it is partially forgiven. I say partially because the DAC was 20 bit capable so truncation should have been to 20 bits, not 16. That would have sharply lowered the impact above.
The truncation creates the busy "grass" that you see in the FFT above. If I feed the DAC 16-bit data (without dither), it cleans up that spectrum (SINAD remains the same):
The output would be the one to the right. Notice how the dominant distortion and power supply spikes remain and hence the reason SINAD does not change.
There is a dealer demo mode in this unit using secret key combinations which allows one to turn off Alpha processing which is on by default. Turning it off and using the suggested -80 dBFS sine wave gives us the display on the left. Our main tone has a skirt around it now and there are some higher order spikes close to 20 kHz. We can better see what is going on in time domain, scope display. Indeed Denon suggested dealers use a scope to show this feature!
The sine wave is compute generated and does not have dither. As a result, we see distinct steps in output level because they are rather coarse in 16 bit context. Turning on Alpha processing which interpolates between samples adaptively (switching between algorithms based on content), we get rid of steppiness. The waveform is distorted though and hence the reason the spectrum still had all the noise and harmonic distortion. Likely if an analog scope of the era was used, this would look smoother.
Let's repeat the test but now using a modern DAC (Schiit Modius). We don't have alpha processing here but what we do have is the ability to turn on dither on the source samples:
It may look messy due to noise smearing the samples as we step from one value to another. Have we done any good? Yes. Let's look at this in frequency domain:
Notice how we have achieved perfection on the right with dither. All harmonic content has disappeared like magic! We just have our main 1 kHz tone now and nothing else. There was a cost however: noise floor came up. We have taken the energy of those harmonics and randomized them throughout the spectrum. That noise may be audible but is much less annoying than not dithering. Low level signals sound gritty and harsh without dither.
In real life content is not a computer generated sine wave. Almost all production is performed using 24-bit samples. If 24-bit is then converted to 16 using dither, then we don't have to worry about steppiness and resulting harmonic distortion that is seen on the left. In this regard, the Alpha processing in Denon doesn't do anything useful. Even without dither, 24-bit content has good bit of noise (rightmost 4 bits is guaranteed to be noise) so it self-dithers by itself to some extent. So one way or the other, we have a situation far from -80 dB undithered test signals.
Personally I like to get the high-res master in 24-bit format so I don't have to worry if the producer knew to properly dither the content. This would obviate everything we have discussed so far.
For completeness, here are a few more measurements:
Conclusions
I suspect most of you don't have interest in this DAC. But hopefully found it educational with respect to mention of Alpha processing in Denon AVRs and general signal processing concept of dither.
-----------
As always, questions, comments, recommendations, etc. are welcome.
Appreciate kind donation using : https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/
Nice to see a wide, room chassis compared to cramped little DACs we see today:
Of course there was no concept of USB input at the time so the only inputs are Toslink and Coax:
Ah, the fun days of hi-fi when high-end gear could get away with an attached cheap mains cable!
The heart of the DA-500 is BurBrown (now TI) PCM1702 sign-magnitude dual-segment R2R DAC. For this reason, it seems to still be cherished today.
Denon DA-500 DAC Measurements
Let's start with our classic dashboard before we get fancy:
Interesting that it output higher than 2 volts. Likely a good number of sales were made compared to other DACs due to higher than normal volume!
SINAD which is the relative sum of noise and distortion is in "poor" bucket using today's standard, falling short of even CD's 16 bit dynamic range despite the unit sporting a 20-bit DAC. I am going to jump to linearity test now to tell us how this DAC works:
Whenever you see a zigzag like this that settles below 90 dB, it indicates that the DAC is simply throwing away extra bits in our 24 bit samples and is just using 16 bits. This type of conversion is bad as it generates steps in response that result in fair bit of distortion. Proper conversion with dither (noise addition) needs to be performed. This is important today due to existence of 24-bit content but at the time this DAC came out, the only source was 16 bits so it is partially forgiven. I say partially because the DAC was 20 bit capable so truncation should have been to 20 bits, not 16. That would have sharply lowered the impact above.
The truncation creates the busy "grass" that you see in the FFT above. If I feed the DAC 16-bit data (without dither), it cleans up that spectrum (SINAD remains the same):
The output would be the one to the right. Notice how the dominant distortion and power supply spikes remain and hence the reason SINAD does not change.
There is a dealer demo mode in this unit using secret key combinations which allows one to turn off Alpha processing which is on by default. Turning it off and using the suggested -80 dBFS sine wave gives us the display on the left. Our main tone has a skirt around it now and there are some higher order spikes close to 20 kHz. We can better see what is going on in time domain, scope display. Indeed Denon suggested dealers use a scope to show this feature!
The sine wave is compute generated and does not have dither. As a result, we see distinct steps in output level because they are rather coarse in 16 bit context. Turning on Alpha processing which interpolates between samples adaptively (switching between algorithms based on content), we get rid of steppiness. The waveform is distorted though and hence the reason the spectrum still had all the noise and harmonic distortion. Likely if an analog scope of the era was used, this would look smoother.
Let's repeat the test but now using a modern DAC (Schiit Modius). We don't have alpha processing here but what we do have is the ability to turn on dither on the source samples:
It may look messy due to noise smearing the samples as we step from one value to another. Have we done any good? Yes. Let's look at this in frequency domain:
Notice how we have achieved perfection on the right with dither. All harmonic content has disappeared like magic! We just have our main 1 kHz tone now and nothing else. There was a cost however: noise floor came up. We have taken the energy of those harmonics and randomized them throughout the spectrum. That noise may be audible but is much less annoying than not dithering. Low level signals sound gritty and harsh without dither.
In real life content is not a computer generated sine wave. Almost all production is performed using 24-bit samples. If 24-bit is then converted to 16 using dither, then we don't have to worry about steppiness and resulting harmonic distortion that is seen on the left. In this regard, the Alpha processing in Denon doesn't do anything useful. Even without dither, 24-bit content has good bit of noise (rightmost 4 bits is guaranteed to be noise) so it self-dithers by itself to some extent. So one way or the other, we have a situation far from -80 dB undithered test signals.
Personally I like to get the high-res master in 24-bit format so I don't have to worry if the producer knew to properly dither the content. This would obviate everything we have discussed so far.
For completeness, here are a few more measurements:
Conclusions
I suspect most of you don't have interest in this DAC. But hopefully found it educational with respect to mention of Alpha processing in Denon AVRs and general signal processing concept of dither.
-----------
As always, questions, comments, recommendations, etc. are welcome.
Appreciate kind donation using : https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/
