|
Digital Drive Upsamplers, DACs, jitter, shakes and analogue withdrawals, this is it. |
|
|
|
Digital Drive Upsamplers, DACs, jitter, shakes and analogue withdrawals, this is it. |
|
In Reply to: RE: The 4 ways to decode digital posted by Thorsten on October 15, 2015 at 18:05:31:
Looks like you're talking about Equibit, which was originally developed by Toccata, sold to TI, used in TacT and Lyngdorf digital amplifiers.
I previously made the mistake of assuming they were generating PWM the conventional way, with a sawtooth/triangle generator and comparator. But after getting schooled, I read the patent describing the Equibit modulator and understood why they say it's pure digital. The modulator synchronously converts PCM to a uniformly sampled PWM signal. The UPWM edge transitions are clocked, so the pulse width is constrained to multiples of the bit clock period. Thus the modulator is converting from one clocked, quantized signal to another clocked, quantized signal, which is connected to the Class D output stage.
In theory, one could implement this conversion directly with just a counter but it would require a really fast clock. For example, directly converting a 16-bit pulse code to a 16-bit pulse width at 44.1 KHz sample rate would require a 2.9 GHz bit clock. I don't recall the bit clock frequencies of Equibit off the top of my head, but it's in the MHz, so the sample resolution of the pulse width will be something on the order of 6-10 bits. Obviously, noise shaping is involved.
One issue to overcome with Class D in general is that PWM conversion is non-linear. But in this digital design, the out of band noise is an additional challenge because passing it through a non-linear format conversion means it can intermodulate back into the audio band. To address the former, they developed a deterministic model for the non-linearity of PWM conversion and implemented a feed-forward correction circuit for it, and combined that with feedback to correct the non-deterministic error (noise).
Sounds like a complicated design, but since it's all implemented with digital logic it should be easy to implement in silicon. It also doesn't require analog feedback, which makes it a more elegant approach than the typical analog Class D approach of using gobs of NFB to correct for the inherently non-linear conversion. I'm surprised the TI product line is not more widely used.
This post is made possible by the generous support of people like you and our sponsors:
Follow Ups
- Guys, I think you're both right - Dave_K 09:44:18 10/19/15 (1)
- RE: Guys, I think you're both right - J. Phelan 19:34:27 10/19/15 (0)
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
