Maxim Engineering Journal #59; Class-D audio

I just got the latest Maxim Engineering Journal, volume 59 (pdf). You can sign up for a paper copy delivered to your door by a federal employee right here. It has a good article about class-D audio amplifiers. I am pretty good with the basics of class-D but when it comes to doing digital feedback of a PWM signal to a digital PCM pulse train the way Freescale does, well I still need to study up on that. Same thing for dithering—purposely adding noise that supposedly improves signal to noise. High-end D to A converters that take SPDIF and puts out analog can get better sound by doing this. Go figure.

Maxim talks about the same thing here, as well as using spread-spectrum modulation in order to reduce EMI. They have a patent on it. Since Linear Tech and few other companies also have a patent on this I can assume that no one really has a valid patent. Maybe this little missive will get the legal and PR departments at the four or five companies that claim to have a patent on EMI reduction using spread-spectrum who will all send in the lawyers to prove why their patents are valid and do not infringe on any of the other patents. Disaffected cynic that I am, I figure all this is obvious to one skilled in the arts. And I should point out that using a spread-spectrum clock on your gizmo really does not reduce EMI, it just smears the energy around so as the FCC spectrum analyzer goes by the EMI is moving around the instruments resolution bandwidth so that it is easier to pass the test. It is almost cheating if you ask me. How about doing some good shielding like Henry Ott talks about and then you don’t need to spread anything. But I digress.

Anyway, the Maxim Journal had those nice little diagrams done by some artist showing how a sine wave and triangle wave can interact to make a PWM signal. It bothered me because the figure (#2) was single-ended and really did not show a better representation, which would be a sine and triangle wave that centered around zero volts and then went both positive and negative. All this single-ended battery powered stuff gets me down. What ever happened to +/- 15-volt rails? So anyway, I wondered what more accurate PWM signals would look like so I spent an hour in PSPICE and did a little circuit with a sine source, and a triangle source (a pulse source with really long rise and fall times). A little look on Maxim’s comparator selector and I found a part that Orcad Capture uses in the standard PSPICE library. A few minutes playing with the inductor and cap and voila, a nice SPICE plot. Notice the small ratcheting in the sine wave output and more importantly notice the pole or lag that the output sine wave has due to the RC filter on the output of the PWM stage. This makes doing analog feedback pretty tricky, hence the love for digital feedback. Here is a zip file of the Orcad files, open the opj project file if you want to have the SPICE profile that I used.

After the class-D article the Maxim Journal talks about a Dallas chip that has encryption that can be used for DRM. Don’t get me started. I am still mad at them for making those chips that authenticate your battery pack so you cannot use cheap Chinese battery packs in your consumer equipment, you have to buy high-zoot OEM battery pack with real Sony batteries that catch fire anyway. I’ll take my chances with the Chinese, thank you very much. Anyway, the good part about this article is that they reprint the IMB taxonomy for secure systems. This was done 10 years ago and is pretty interesting, like most anything IBM does.

.
.
.
Click images for bigger size.