Announcement

Collapse
No announcement yet.

Some background information on Class D amplifiers

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Some background information on Class D amplifiers

    The discussion in the previous thread on Class D amplifiers raised a number of questions about Class D technology and specific chips, but the discussion was a bit thin on answers. If you are interested in knowing more about how Class D works and what are some of the challenges, there is a good overview at the Analog Devices web site: http://www.analog.com/library/analog...06/class_d.pdf

    I've tried to fill in some details about specific chips in the discussion below, keyed to the 3 modulation techniques referenced in the article. Hope this helps, and at least partially addresses why some Class D amps may sound harsh and why there are differences. Some of this discussion includes my personal opinions, so please read this with some sympathy .

    The first modulation technique described in the article (in the "Modulation Technique" section) is PWM. The TI chips, including the TAS5630, use this approach. TI has been making chips used in highly regarded Class D amps since they acquired Tact back in 1999. The PWM designs use a fixed clock. The fixed clock results in strong harmonics that need to be filtered from the audio and requires special attention to EMI (RF interference). However, the fixed clock also simplifies multi-channel design (this is a good approach for that 7.1 amp). There are many PWM chips to choose from: NXP (TDA8920 series), ST (TDA7498), Maxim, TI, etc., and several have some potent integrated DSP capability. Class D chips with good DSP include TI's TAS57XX and TAS55XX series, ST’s STA308a and the STA32X series and the Cirrus CS4225. My personal interest in Class D amps has been spurred on by wanting to control that embedded DSP capability, but I’ve always been put off by having to use a PWM-based design to leverage this capability.

    The PWM-based chips can have very low distortion when properly implemented. The conversion from analog or digital to PWM involves several steps, including resampling and using a noise shaping filter that moves quantization noise out of the audio band. The distortion is typically below .02%, however, a lot of the PWM implementations tend to have a THD peak of around .1% to .2% in the region between 5KHz-10KHz. This elevated distortion possibly contributes to some of the negative impressions of certain Class D amps. I believe this distortion in some of the PWM chips is primarily due to the implementation of the noise shaping filter. The "high-end" TI devices such as the TAS55XX chips use a 5th-order noise filter, and the distortion is very low at all frequencies. However, most of their medium-power devices use a 4th-order filter, and they show that increased distortion in the 5KHz-10KHz range.

    The PWM chips usually use feedback to provide good power supply rejection, although some of the earlier chips from TI and ST don't use any feedback. On the chips that use feedback, the amp doesn't put the output filter in the feedback loop, and as a result this type of amp is "sensitive" to the load impedance interaction with the filter. That is, you may see peaking or roll-off of the frequency response if the loudspeaker is not the specified impedance. However, if you are careful about matching the speaker to the amp, these amps are flat in the audio band, with no peaking or roll-off.

    A point worth noting is that these chips are commonly used in budget designs, and sometimes the vendor may "cut corners" in implementing the output filter. Trevor Marshall has a good article on how the output filter was implemented poorly in a TDA7498 board, where the resulting distortion was an order of magnitude higher than shown in the manufacturer's data sheet. Anyone interested in modifying these amps should read his article.

    The next type of modulation discussed in the Analog Devices article is the self-oscillating amplifier. There are many variations on the self-oscillation amplifiers, but most designs are now based on the "UCD" approach developed by Bruno Putzeys. You can Google his name to get more information about UCD amps, but you will also get a lot of negative verbiage about other types of Class D amps :rolleyes:. The self-oscillating amps are Class-D, but they are entirely an analog circuit—there is no digital filtering like there is in the other Class D modulator techniques. Hypex and IcePower were the first commercial UCD products, but recently International Rectifier developed the IRS2092, which is "a UCD amp on a chip". The IRS2092 is now used in many low-cost amp modules, and the price/performance ratio is very impressive. You can also get more info on the self-oscillating amp at the International Rectifier website (irf.com)--they have some nice briefings on Class D design that explain the concepts and provide implementation details.

    All of the self-oscillating amplifiers, including the low cost IRS2092 designs, provide very good audio quality, with very low distortion (less than .01% for all frequencies) and good power supply rejection. The IRS2092 amp uses feedback from the chip output, not the filter output, so the same comments on the load sensitivity of the PWM amps apply. However, there are newer implementations of the self-oscillating amps from Hypex and others that use feedback from the output of the filter. These newer UCD implementations are not as sensitive to the load.

    One limitation of the self-oscillating amp is that it is difficult to synchronize multiple amps without compromising performance, so you need to be careful about making that 7.1 amp with self-oscillating Class D amps. It’s not impossible to make a 7.1 amp with self-oscillating Class D amps (IcePower has one that is used in the Pioneer Elite AVR), but it requires care…

    The last type of modulation described in the article is sigma-delta (or delta-sigma--both terms are used in the literature). This type of modulation is used in the excellent but low-power SSM3302 chip from Analog Devices, and it is used in some of the smaller Class D amps targeted for battery operation. Delta-Sigma is also used in the Zetex DDFA chips (now CSR) that are used in the latest NAD equipment. Sigma-Delta modulation was also used in the Tripath chips. Delta-Sigma amps do not use a fixed clock like the PWM modulators, and they are actually a form of pulse density modulation, or PDM.

    A very good discussion of this modulation technique is described in an article about the AD1994 (the AD1994 is a precursor to the SSM3302). As noted in the AD1994 article, the distortion and noise is primarily dependent on the order of the noise shaping filter, and by using a 7th order filter the AD1994 was able to reduce distortion to levels below .005%. Zetex claims similar distortion levels, which actually exceed the levels achievable with the self-oscillating designs.

    The Zetex design uses feedback from both the chip and the output filter to compensate for filter non-linearity and interaction with the driver. However, earlier Delta-Sigma chips like the Tripath devices just use feedback taken at the chip output, before the output filter, so they are subject to the same issue noted earlier for matching the load to the output filter.

    It’s probably worth noting that the Tripath chips didn’t have the same level of sophistication in the noise shaping filter as the AD1994, and they were limited by the early analog/digital IC fab processes, which resulted in relatively noisy input stages. But it’s interesting to see Tripath’s approach get vindicated in more modern IC’s.

    So what’s going to be the “go-to” Class D technology in the next few years? I think all three of these approaches are going to continue to be used, even for the audiophile market. PWM allows using digital inputs and it simplifies multichannel design, and both of these are important for the high-end market. The next generation of self-oscillating amp IC's will incorporate feedback from the filter output and provide a good cost-effective solution for high-end stereo. And I suspect that there will be a few more high-power delta-sigma chips to choose from in the near future that will set the bar for noise and distortion even lower.
    Last edited by neildavis; 01-09-2014, 09:32 PM.
    Free Passive Speaker Designer Lite (PSD-Lite) -- http://www.audiodevelopers.com/Softw...Lite/setup.exe

  • #2
    Re: Some background information on Class D amplifiers

    Thanks! Now my head hurts :D

    Comment


    • #3
      Re: Some background information on Class D amplifiers

      Excellent explanation of the different classes of D. ;)

      I just want to add that in everything I've read, the output inductor is a very crucial component in any class-D circuit. The type of core used and component size is very strict to get the best performance out of your amp.
      "I just use off the shelf textbook filters designed for a resistor of 8 ohms with
      exactly a Fc 3K for both drivers, anybody can do it." -Xmax

      Comment


      • #4
        Re: Some background information on Class D amplifiers

        http://www.irf.com/technical-info/appnotes/an-1071.pdf
        http://www.ti.com/lit/an/sloa031/sloa031.pdf
        "Not a Speaker Designer - Not even on the Internet"
        “Pride is your greatest enemy, humility is your greatest friend.”
        "If the freedom of speech is taken away, then dumb and silent we may be led, like sheep to the slaughter."

        Comment


        • #5
          Re: Some background information on Class D amplifiers

          Hypex White Papers are a good read as well, especially the original UcD publication:

          http://hypex.nl/downloads/white-papers.html
          "I just use off the shelf textbook filters designed for a resistor of 8 ohms with
          exactly a Fc 3K for both drivers, anybody can do it." -Xmax

          Comment


          • #6
            Re: Some background information on Class D amplifiers

            Originally posted by dcibel View Post
            Excellent explanation of the different classes of D. ;)

            I just want to add that in everything I've read, the output inductor is a very crucial component in any class-D circuit. The type of core used and component size is very strict to get the best performance out of your amp.
            If anyone is interested in digging deeper into class-D modulation theory, I have a number of technical articles, mostly pulled from IEEE journals. I can't post any because they were purchased, but I could work with a private request.

            The article that I found the most useful for understanding class-D output inductors is the Ferroxcube Application Note for class-D Amplifiers.

            The ST briefings are also a good source of information on class-D amp design--I have a couple of them that I could make available. And I've got all of the earlier Apogee class-D amp app notes, which cover a wide range of topics on their PWM designs. One of the ST app notes ended up as a series in EETimes on output filter design that is a good introduction to the topic.
            Free Passive Speaker Designer Lite (PSD-Lite) -- http://www.audiodevelopers.com/Softw...Lite/setup.exe

            Comment


            • #7
              Re: Some background information on Class D amplifiers

              Thank you...VERY nice explanation of the different types of Class D amps.

              So at this point, there isn't no one technology that's preferred over the others for a 2-channel amp.?

              I noticed NAD is using Hypex modules in their new D 3020 integrated. Though not a powerhouse, it's been getting a lot of positive reviews:

              http://www.whathifi.com/review/nad-d-3020

              Comment


              • #8
                Re: Some background information on Class D amplifiers

                Originally posted by foxfire3 View Post
                ...So at this point, there isn't no one technology that's preferred over the others for a 2-channel amp.?
                Nope :D
                * But i guess that depends on the criteria placed upon "preferred" ;)
                "Not a Speaker Designer - Not even on the Internet"
                “Pride is your greatest enemy, humility is your greatest friend.”
                "If the freedom of speech is taken away, then dumb and silent we may be led, like sheep to the slaughter."

                Comment


                • #9
                  Re: Some background information on Class D amplifiers

                  if you look at most of what is consider the high end market the PWM is the design most people use.
                  craigk

                  " Voicing is often the term used for band aids to cover for initial design/planning errors " - Pallas

                  Comment


                  • #10
                    Re: Some background information on Class D amplifiers

                    Originally posted by foxfire3 View Post
                    Thank you...VERY nice explanation of the different types of Class D amps.

                    So at this point, there isn't no one technology that's preferred over the others for a 2-channel amp.?***
                    I'd say that any design that does not produce a constant frequency response regardless of load (equivalent to a good solid state amp) for is not preferred for a 2-channel amp intended to drive a multiway/passive crossover.

                    For active systems, where the filtering is based on measurements of the amp+drive unit in situ, other options may have their own advantages.
                    --
                    "Based on my library and laboratory research, I have concluded, as have others, that the best measures of speaker quality are frequency response and dispersion pattern. I have not found any credible research showing that most of the differences we hear among loudspeakers cannot be explained by examining these two variables." -Alvin Foster, 22 BAS Speaker 2 (May, 1999)

                    Comment


                    • #11
                      Re: Some background information on Class D amplifiers

                      Great, open and honest assessment/information Neil, thank you very much!

                      I think Class D has come a VERY long way and is certily good upto the mid-fi market at this point, like CD player and DSD, and all the ranges of bit-rate conversion out there this may ultimately boil down to the filters on the output section. MMy best anaolgy is that class D amps have a very similar sound to early CD players. Just not nearly as bad. Maturity in a technology has a lot to do with it. As of now: I would set up my home theater with class D and I do not look elsewhere when it comes to driving subs. Serious music listening...Still a little way to go.

                      Pratical use comes into play too. On my desktop: there are some affordable mini 1-3 watt tube amps I think I would rather drive my desktop speakers with. I will admit, the Lepai is a silly little peice of gold though... Specially with a good power supply. So what does it all mean? For me, anyway, I look forward to someday and likely in the next 5 years replacing my 60lb 500WPC carver amp with something the size of a paperback that sounds as good if not better and is double the power. There may be no replacement for displacement but there is certinly no option by overhead either...At least in my book!
                      .

                      Comment


                      • #12
                        Re: Some background information on Class D amplifiers

                        Originally posted by foxfire3 View Post
                        So at this point, there isn't no one technology that's preferred over the others for a 2-channel amp.?
                        As a practical matter (availability coupled with excellent performance of available devices) there is . . . self-oscillating wins, either Hypex or IRS2092 based designs. Their greatest sonic advantage is vanishingly low distortion at low power (where most music is) that comes from the absence of the "crossover distortion" that plagues all but Class A or the very best AB designs. The rise in "distortion" at low power seen in the curves for these devices does not actually reflect "distortion" . . . it's the "noise" component of THD+Noise, and we don't hear it as "distortion" (if we hear it at all).

                        The biggest design "issue" for the DIYer is that these amps exhibit some sensitivity to load impedance (a result of the output filter necessary to suppress the sampling frequency). It's easy to design around, and should not pose a problem to anyone who understands the other implications of output and load impedance on electronic devices.
                        "It suggests that there is something that is happening in the real system that is not quite captured in the models."

                        Comment


                        • #13
                          Re: Some background information on Class D amplifiers

                          Originally posted by mzisserson View Post
                          MMy best anaolgy is that class D amps have a very similar sound to early CD players. Just not nearly as bad.
                          Entirely untrue of either Hypex or IR2092 amps (which have NOTHING in common with "early CD players"). And they're not just "mid-fi" . . . they are every bit as good as any of the best "hi-fi" amps on the market. Arguably they DEFINE the "hi fi" market . . . and one can simply dismiss the big-bucks hundred pound monsters-of-old out of hand.
                          "It suggests that there is something that is happening in the real system that is not quite captured in the models."

                          Comment


                          • #14
                            Re: Some background information on Class D amplifiers

                            Originally posted by foxfire3 View Post
                            Thank you...VERY nice explanation of the different types of Class D amps.

                            So at this point, there isn't no one technology that's preferred over the others for a 2-channel amp.?
                            Given that the high-powered delta-sigma amps aren't available outside of CSR, I'd say the Hypex modules are preferred. The nCore variation uses very high loop gains to apply large quantities of feedback that reduce distortion to bragging-rights levels. If you can afford the nCore modules, that's a great solution, but their other modules are also very good. However, if you are on a budget and have well behaved speakers that match the output filter, the IRS2092 amps are a good practical choice. They are probably below the distortion and noise thresholds for most humans, and the IRF reference design that is used by Sure and others makes it well protected and robust. 500W for about $50 is hard to beat.

                            I noticed NAD is using Hypex modules in their new D 3020 integrated]
                            Are you sure they are using Hypex? NAD has a close association with CSR, and they are using the CSR aptX Bluetooth in the D 3020. Diodes, Inc, bought Zetex, and then just recently CSR bought the Zetex technology from Diodes, Inc. If you look at this briefing around 7:02 the head of the voice and music group at CSR talks about the acquisition of the Zetex DDFA technology. I don't see where they mention the DDFA technology for the D 3020, but I don't see any reference to Hypex, either.
                            Free Passive Speaker Designer Lite (PSD-Lite) -- http://www.audiodevelopers.com/Softw...Lite/setup.exe

                            Comment


                            • #15
                              Re: Some background information on Class D amplifiers

                              Hey Neil -- you mention a couple times the relative ease by which a topology can be implemented in multi-channel amps. Why is it important to synchronize driver ICs? If the output waveform (post-filter) is an analog to the input, why does the duty cycle or frequency or whatever actually matter? I suspect it has something to do with phase alignment between channels, but that's just a WAG. Am I even close?

                              Comment

                              Working...
                              X