On the surface (read, to a novice) the benefits of direct digital amplification over class D seem obvious. By direct digital amplification I mean an amplifier that takes a digital signal (SPDIF/PCM - a square wave) converts it to PWM (another square wave), amplifies it, and converts it to analog at the output. This as opposed to a Class D amp which takes an analog signal, converts it to PWM, amplifies it, and converts back to analog at the output. (Yes I know this is an over simplification, but I hope it will do for what I want to ask.) The obvious point being that the direct digital amp eliminates the DAC, which has got to be a huge advantage, especially considering how much people will invest in a good quality DAC.
Now, there are a handful of direct digital amps out there. Some very inexpensive and some very expensive, and I know of one (and only one) amp module available to the DIYer. This is opposed by a huge number of class D amps available, both as commercial amplifiers and DIY modules. So, my question is, why are there so few direct digital amps? It doesn't make sense to me, given what seems to me to be the obvious advantage that they possess. I am sure there must be some technical reason why the direct digital technology has not been pursued any more than it has been. But what? I would love to hear people's take on this.
Now, there are a handful of direct digital amps out there. Some very inexpensive and some very expensive, and I know of one (and only one) amp module available to the DIYer. This is opposed by a huge number of class D amps available, both as commercial amplifiers and DIY modules. So, my question is, why are there so few direct digital amps? It doesn't make sense to me, given what seems to me to be the obvious advantage that they possess. I am sure there must be some technical reason why the direct digital technology has not been pursued any more than it has been. But what? I would love to hear people's take on this.
SPDIF is not PWM and will not work with class D directly. That is why analogue is chopped up in to PWM.
Yes I don't believe any digital audio is PWM. And you also have to seperate the 2 or more channels.
Class D amps with analog input are dead simple - usually a single chip with single power supply. If you want a digital amp with SPDIF input, you will usually need at least 3 chips: SPDIF receiver, the amp itself and a microcontroller. You will also need a step-down power supply on-board for digital domain and some other small bits. While all this is fairly simple to design, it complicates things a lot for most cheap manufacturers who can't even do a proper layout 🙂
SPDIF is PCM, not PWM. But PWM is a square wave, like any digital signal. Whether or not this makes PWM "digital" is probably a different conversation. That's not my question.
An example of the kind of amp I am talking about is the DDX320 by HiFiMeDIY. I have one and it sounds very good. It takes SPDIF input and outputs an amplified analog signal at the speaker posts. No special speakers are required. And no DAC is needed. To my uneducated mind that seems like a huge advantage, both in terms of sound quality and from a financial standpoint - though it's the question of sound quality that I want to better understand. The DDX320 sells for only $114, so implementing the technology isn't costly, and how complex could it be? Companies like Hypex have invested huge sums into R&D of class D amplification and their units sell for much more than the DDX320. I'm wondering why they don't invest in direct digital as well. Seems like a no-brainer. But since they don't, I think there must be a technical reason that direct digital is inherently inferior to class D. I'd like to know if that's true and if so, what it is.
An example of the kind of amp I am talking about is the DDX320 by HiFiMeDIY. I have one and it sounds very good. It takes SPDIF input and outputs an amplified analog signal at the speaker posts. No special speakers are required. And no DAC is needed. To my uneducated mind that seems like a huge advantage, both in terms of sound quality and from a financial standpoint - though it's the question of sound quality that I want to better understand. The DDX320 sells for only $114, so implementing the technology isn't costly, and how complex could it be? Companies like Hypex have invested huge sums into R&D of class D amplification and their units sell for much more than the DDX320. I'm wondering why they don't invest in direct digital as well. Seems like a no-brainer. But since they don't, I think there must be a technical reason that direct digital is inherently inferior to class D. I'd like to know if that's true and if so, what it is.
SPDIF input to analog out does use a DAC.
Analog input to class D out doesn't use an ADC.
Is this not correct?
Analog input to class D out doesn't use an ADC.
Is this not correct?
Well fundamentally one of these amps is just a combination D/A converter and amplifier, but with the D/A conversion process happening in the power stage instead of in a separate upstream DAC.
With the current state of the art, poking a small signal, high quality DAC chip next to an analog domain amplifier produces better results, and with a simpler design. I don't expect this to change any time soon.
With the current state of the art, poking a small signal, high quality DAC chip next to an analog domain amplifier produces better results, and with a simpler design. I don't expect this to change any time soon.
Don't understand fully how it works, but as I can see the STA320 is what converts the spdif to 2.1 and makes it usable for the amp section. And by doing that you are "stuck" with one converter. And there is a conversion that can be done well or bad. Like when dac is integrated into the amp (without analog input). You can't test different DAC for your amp. It is a way, but many may not like it.
Last edited:
I'm trying to compare direct digital to class D. They have identical output stages. The difference is in the input stage - one takes analog input the other takes digital input. The direct digital amp converts the digital signal (PCM) to PWM while the class D converts analog signal to PWM. After that point the two are identical. Perhaps the PCM->PWM conversion is problematic compared to the analog->PWM conversion, which might explain why direct digital hasn't taken off. But I don't know. What I do know is that with direct digital the DAC is taken out of the signal path entirely which seems to me must be a big advantage, especially when you consider how much money people are willing to spend on a good DAC..
Here is one, price is incredibly low :
FX Audio D802 HIFI Digital Amplifier Remote Control USB Optical Fiber Coaxial Input 192KHZ 80W*2 Blue - Free Shipping - ThanksBuyer
Some guys said it is the best amplifier they have ever listened to.
There have a huge amount of functions but the distortion specs are in the par with those seen in analog car amps :
STA326 Datasheet(PDF) - STMicroelectronics
FX Audio D802 HIFI Digital Amplifier Remote Control USB Optical Fiber Coaxial Input 192KHZ 80W*2 Blue - Free Shipping - ThanksBuyer
Some guys said it is the best amplifier they have ever listened to.
There have a huge amount of functions but the distortion specs are in the par with those seen in analog car amps :
STA326 Datasheet(PDF) - STMicroelectronics
A 1-bit delta sigma DAC is still a DAC, whether the output is 1mW or 1000W.
And pretty much fallen out of favor for multibit implementations.
What if people start making multi-bit current-feedback class-D amps.
And pretty much fallen out of favor for multibit implementations.
What if people start making multi-bit current-feedback class-D amps.
Analog->PWM conversion has infinite time resolution, so there's no fundamental limits on resolution. Additionally, in the analog domain you can compensate for things such as power supply ripple/sag, finite switching delays, etc... quite easily using feedback.
Digital->PWM conversion is done in discrete time. For starters this brings some fundamental limitations - assuming you're doing classical triangle wave modulation and no dithering/noise shaping, at a fixed 400KHz PWM frequency, in theory you need to compare your input signal and triangle wave at a frequency of 400K*(2^16) = 26.2GHz. Which isn't practical, so you have to bring dithering, noise shaping, different modulation methods, etc... to the party in order to knock that GHz frequency down into the tens/hundreds of MHz.
Next problem is latency. Feedback works great and near-instantaneously in the analog domain; in the digital domain you've got sampling latency and computational latency added to your feedback path, this time delay limits the amount of feedback correction you can do without making the feedback loop unstable. As a result you have to do tricks like feedforward correction - predicting what your output error is going to be based on the audio content coming in, and modifying what your feeding the output stage to cancel that error.
Even when you get this full system working, the performance of the system will be limited by the quality of the A/D converter you're using for feedback.
I spent many months working on this exact problem, when I was designing a "digital to PWM" modulation system for a high power AM radio broadcast transmitter several years back. We had practical reasons for going the digital route, but it opened a whole can of worms - took us a while before we managed to reach the performance of the old analog modulation system used in the previous design.
Thank you for that explanation.
Does something like the TI TAS5558 chip, described as an "8-Channel HD Compatible Audio Processor with SRC and PWM Output", or other chips like it, do all of the above? It's a pretty cheap solution if it works good. But I guess the question is how good could it be?
I'm wondering, does a traditional DAC face the same problems that you mentioned, or is there somehow an advantage going straight from digital to analog vs. digital to PWM? Intuitively it would seem to me to be the opposite, since both PCM and PWM are square waves, converting from one square wave to another would be less problematic than converting from a square wave to a sine wave. But I get the feeling that's being naïve.
In the AM transmitter "digital to PWM" project, it sounds like you had challenges but you did manage to find solutions. Once you did reach the performance of the old system was it worth it? I mean, not including the cost of R&D, was the digital solution an overall improvement cost wise or performance wise?
What I'm trying to get at is whether the challenges of digital->PWM specifically and direct digital amplification in general are so great that they out weigh any advantage gained by eliminating the external DAC from the signal path.
The TAS chips do high order noise shaping in order to achieve sufficient bandwidth/dynamic range, but they run open loop - the power stage needs a very low power supply impedance in order to achieve the audio performance that they give in the datasheet. If the power supply voltage is wandering around with the audio envelope, the THD/IMD numbers go through the roof.
In our transmitter case case, we had practical reasons for going the digital PWM route. We had anywhere from 6 to 12 PWM channels coming out of a big FPGA and going into a bank of power modules, each generating RF which was combined to generate the potentially-hundreds-of-kW RF output. The big advantage was that we had individual, precise control of each channel - by tuning the amplitude/phase of each PWM output, which is easy in the digital domain but hard in the analog domain, we could minimize modulator spurious tones in the RF output and minimize RF combining loss, even when power modules were failed or removed. We also had some proprietary tricks we were doing to make the transmitter more efficient at the expense of linearity, which required some other digital domain tricks to correct.
At the end of the day, it was a superior product... but the analog transmitter was a lot simpler, that's for sure.
- Status
- Not open for further replies.