Zoudio didnt claim that. I did. The specific TI chip they use for the amplifiers is I2S input, direct digital to PWM modulation - no feedback.
On both my AIO4CH amplifiers, I have a USB to I2S converter connected successfully. I dont use the BT, but have it available as an option. Nowhere in the amplifier chip signal path I can find is there an analog path - unlike other TI class D amplifiers which have "Burr Brown" DACs inside, a bit of analog signal path before returning to digital as PWM for output. On some, you can feedback the analog signal from the speaker output terminals. It's amazing they make parts for every SQ preference; realizing some customers want analog negative feedback and its sound - even on an I2S input amp chip.
BT is getting better as it evolves. Isnt the 5.0 version 24 bits? How many bits do you need for audio? 24 going to compress something? Unless in the BT chip itself, they DAC to analog, then use that to ADC modulate the BT signal - but that just doesnt make sense to me.
However, I know nothing about how BT actually works, so, maybe. If it compresses, I dont know where in the signal chain it would do so, or why. If what some term as compression actually refers to is 48kHz max sample rate @ 24 bits / sample, so it'll "compress" your 96 or 192 rates down to just > CD quality. The marketing people at Qualcomm want you to believe you'll never hear the difference anyway. I'm 65, so I probably wouldnt.
Running I2S, the Zoudio amps sound great to me. Their latest offering does BT 5.0. I like that you get 4 channels of amplification, with DSP generated just about any crossover you'd like. No coils, resistors and caps in my speakers; tweeter and woofer (actually FR and bass-helper) each get their own dedicated amp, driving their own, dedicated frequency range.
On both my AIO4CH amplifiers, I have a USB to I2S converter connected successfully. I dont use the BT, but have it available as an option. Nowhere in the amplifier chip signal path I can find is there an analog path - unlike other TI class D amplifiers which have "Burr Brown" DACs inside, a bit of analog signal path before returning to digital as PWM for output. On some, you can feedback the analog signal from the speaker output terminals. It's amazing they make parts for every SQ preference; realizing some customers want analog negative feedback and its sound - even on an I2S input amp chip.
BT is getting better as it evolves. Isnt the 5.0 version 24 bits? How many bits do you need for audio? 24 going to compress something? Unless in the BT chip itself, they DAC to analog, then use that to ADC modulate the BT signal - but that just doesnt make sense to me.
However, I know nothing about how BT actually works, so, maybe. If it compresses, I dont know where in the signal chain it would do so, or why. If what some term as compression actually refers to is 48kHz max sample rate @ 24 bits / sample, so it'll "compress" your 96 or 192 rates down to just > CD quality. The marketing people at Qualcomm want you to believe you'll never hear the difference anyway. I'm 65, so I probably wouldnt.
Running I2S, the Zoudio amps sound great to me. Their latest offering does BT 5.0. I like that you get 4 channels of amplification, with DSP generated just about any crossover you'd like. No coils, resistors and caps in my speakers; tweeter and woofer (actually FR and bass-helper) each get their own dedicated amp, driving their own, dedicated frequency range.
Last edited:
Depends on Codex
Bluetooth not much more than 264 to 320 Kbps that is max bitrate.
Devices can default much lower
So depends on how cheap it is.
Even if expensive still cheap, price mark up LOL
Sony brought it up to 990 kbps
but you likely have to change settings in device.
usually default to 320/330
useless numbers most people listen to lossy music anyways from Mp3
or whatever horrible file uploaded to internet.
CD quality only 16 bit 44khz
You can make Wave File at 24bit 192khz
which is likely highest rate of most soundcards.
Software can go way up to 32 bit or higher.
dont matter unless your soundcard has same resolution.
Mp3/Mp4 is good enough for most, but is compressed garbage non the less.
But it most common because of storage space and internet streams.
If you want high quality, it would be computer with high resolution soundcard
using low loss files like Wave or Flac.
Then again, only as good as original source file.
Phone dont need service, get old phone very cheap.
connect to wifi, hot glue to old CD player. Dream come true.
Analog and bluetooth output
Or use Car stereo with touch screen.
most have bluetooth and usb.
so if you did have high quality wave file. can still play from usb stick
Bluetooth not much more than 264 to 320 Kbps that is max bitrate.
Devices can default much lower
So depends on how cheap it is.
Even if expensive still cheap, price mark up LOL
Sony brought it up to 990 kbps
but you likely have to change settings in device.
usually default to 320/330
useless numbers most people listen to lossy music anyways from Mp3
or whatever horrible file uploaded to internet.
CD quality only 16 bit 44khz
You can make Wave File at 24bit 192khz
which is likely highest rate of most soundcards.
Software can go way up to 32 bit or higher.
dont matter unless your soundcard has same resolution.
Mp3/Mp4 is good enough for most, but is compressed garbage non the less.
But it most common because of storage space and internet streams.
If you want high quality, it would be computer with high resolution soundcard
using low loss files like Wave or Flac.
Then again, only as good as original source file.
Phone dont need service, get old phone very cheap.
connect to wifi, hot glue to old CD player. Dream come true.
Analog and bluetooth output
Or use Car stereo with touch screen.
most have bluetooth and usb.
so if you did have high quality wave file. can still play from usb stick
Last edited:
@jjasniew
Sorry mate, it seems like you've fallen for the Class D marketing, or just misunderstand it. PWM is at its heart an analogue process and does not even necessarily have to be digitally controlled. Just because it has a digital front end, does not make it a 'digital amplifier' there is no such thing; regardless of what the marketing departments say. When something attempts to switch something on, or off, that doesn't mean its a digital process. that would make a light switch digital ... it would also define any 'square wave' as being digital, but we all know that isn't true either (even without taking into account the realities imposed by switching losses in the real world and that if you look at any square wave that has made it into the physical world, it isn't square). Class D isn't digital and it never was. thankfully at least the wiki has been updated to reflect this these days.
Just as Bluetooth is lossy and it would still be lossy if it was 64bit. the bit depth has nothing to do with it. lossy is lossy and you can certainly encode even the most horrendously compressed signal in 24bit. Bluetooth or at least APTx et all, are much better than they used to be, but I certainly wouldnt want it to be my only input.
Sorry mate, it seems like you've fallen for the Class D marketing, or just misunderstand it. PWM is at its heart an analogue process and does not even necessarily have to be digitally controlled. Just because it has a digital front end, does not make it a 'digital amplifier' there is no such thing; regardless of what the marketing departments say. When something attempts to switch something on, or off, that doesn't mean its a digital process. that would make a light switch digital ... it would also define any 'square wave' as being digital, but we all know that isn't true either (even without taking into account the realities imposed by switching losses in the real world and that if you look at any square wave that has made it into the physical world, it isn't square). Class D isn't digital and it never was. thankfully at least the wiki has been updated to reflect this these days.
Just as Bluetooth is lossy and it would still be lossy if it was 64bit. the bit depth has nothing to do with it. lossy is lossy and you can certainly encode even the most horrendously compressed signal in 24bit. Bluetooth or at least APTx et all, are much better than they used to be, but I certainly wouldnt want it to be my only input.
Last edited:
Don't fret about it though. there is a lot of noise on the internet that seems determined to still misrepresent this 70 year old process (PWM) as being some sort of modern panacea and we all know to be truly modern, something has to be digital. Marketing departments are rarely staffed by engineers and it would appear that they hardly ever consult their engineers before writing the ad copy, either. Its possible some of these people simply think the D in class D, stands for digital. Pro tip: it doesn't.
What I'm referring to, re the Zoudio amp, is this little fellow in the particular amp chips used - outlined in red;
I've seen the diagrams of other TI chips, where the output of a DAC is brought out of the chip as an analog signal on a pin. It goes through a nice, cheap MLC cap (in the implementations I've seen) and then back into the amp (via a second pin) to be converted to PWM for the output drive. That's the difference I'm illuminating between this design and such.
Looking at the diagram I can see I was wrong about this amp having no feedback. It does and I have no idea how it works. I've never seen anything more descriptive about it than the above "block", from TI's TAS5825 spec; it doesnt go into details about what the two lines "Close Loop Feedback" exactly are.
I like how it sounds; I wont be going off it anytime soon. Not for tubes, hot JFet class A, nor someone else's fancy pants DAC with analog outs. I'd drive a push-pull tube output stage with this amp as a DAC if I wanted to have more fun playing around with some connect half A to half B system topology first. How's that for just sayin' I'd recommend it, because of what it is and what it can do.
Lossy 'ol BT as the only input, yeah - but this is DIY audio, not off the shelf finished product audio. Which is why I extended myself to get I2S connected up, to hop around the BT part. I'm sure with a little diligence that could be done with the latest Zoudio product. Mine has a switch on the backside; "BT / USB"...
Still not understanding how BT - with say FLAC ripped from a CD as input, is "lossy" - except in the sample rate dimension for "HD" audio sources. The say the latest even has bit error correction - so there's one type of "loss" they fixed in 5.0... So all the bits sent are corrected to be received as sent.
I've seen the diagrams of other TI chips, where the output of a DAC is brought out of the chip as an analog signal on a pin. It goes through a nice, cheap MLC cap (in the implementations I've seen) and then back into the amp (via a second pin) to be converted to PWM for the output drive. That's the difference I'm illuminating between this design and such.
Looking at the diagram I can see I was wrong about this amp having no feedback. It does and I have no idea how it works. I've never seen anything more descriptive about it than the above "block", from TI's TAS5825 spec; it doesnt go into details about what the two lines "Close Loop Feedback" exactly are.
I like how it sounds; I wont be going off it anytime soon. Not for tubes, hot JFet class A, nor someone else's fancy pants DAC with analog outs. I'd drive a push-pull tube output stage with this amp as a DAC if I wanted to have more fun playing around with some connect half A to half B system topology first. How's that for just sayin' I'd recommend it, because of what it is and what it can do.
Lossy 'ol BT as the only input, yeah - but this is DIY audio, not off the shelf finished product audio. Which is why I extended myself to get I2S connected up, to hop around the BT part. I'm sure with a little diligence that could be done with the latest Zoudio product. Mine has a switch on the backside; "BT / USB"...
Still not understanding how BT - with say FLAC ripped from a CD as input, is "lossy" - except in the sample rate dimension for "HD" audio sources. The say the latest even has bit error correction - so there's one type of "loss" they fixed in 5.0... So all the bits sent are corrected to be received as sent.
As for BT with FLAC as input, it’s more like forcing an egg through a garden hose, might be able to make a darned good scramble with it , but just isn’t going to be good for a poached arrangement.
Not sure if I should be happy that some of the newer technologies sound as good as they do out of the box after spending a few years now working with the older stuff!
Not sure if I should be happy that some of the newer technologies sound as good as they do out of the box after spending a few years now working with the older stuff!
the picture is exactly what I expected. Class D, with digital input stage and everything else analogue. So, using an internal wirebond or trace vs an external trace makes it more digital? PWM is analogue. Everything apart from the input stage is analogue, the feedback is analogue, so thats how the feedback works. just like any other feedback. some of the output signal is fed back to a node in the amplifier and compared/contrasted with the input. This process can of course be digital or analogue, but this is analogue; otherwise it would require a further ADC and that would also add latency.
https://soundsightheadphones.com/wireless/bluetooth-5-0-codecs-and-other-bluetooth-formats/ they are getting better, but still lossy; its a matter of how lossy each codec is. Hopefully LDAC becomes more widely supported, as its getting very close.
https://soundsightheadphones.com/wireless/bluetooth-5-0-codecs-and-other-bluetooth-formats/ they are getting better, but still lossy; its a matter of how lossy each codec is. Hopefully LDAC becomes more widely supported, as its getting very close.
Last edited:
I'm somewhat familiar with the ramp-comparitor analogue circuit for generating a PW from an analog input. Within the "Digital to PWM Conversion" box, how do you know the ramp part isnt implemented by a counter and the comparison isnt done by digital logic; the input word value vs the counter value? Seems the obvious way to do it to me - you already have the digital word for the sample point of the audio signal. Why convert it to analog to compare with an analog ramp? Make the ramp part fit the context of what's already there...
Perhaps I'm just naive to assume when a manufacturer like TI says "Digital to PWM" that could possibly be what's going on inside.
Again, I have noooo idea how the feedback part actually works, so maybe they are just pulling the wool cap over my eyes / imagination...
From the article, I see the "loss" problem is that BT wont do HD audio, so you lose your 96K (or 192k) bitrate. LDAC being the best bitrate performer, "gets very close" to high bitrate fluency, while APT-X-HD is still down in the 48k scrambled egg territory ;')
So it would seem Zoudio is limiting the products ultimate performance through the Bluetooth V5.0 APT-X / A2DP / AAC, as its primary connection channel. The TAS5825 amplifier chips "Supports 32, 44.1, 48, 88.2, 96, 192 kHz sample rates" which I assume is achievable with a DIY USB to I2S connection, as I have done with mine. From what I've read, Zoudio will eventually offer USB connection implemented via the same chip that does the BT - which may support these higher sample rates as well - as a firmware update.
Whether it does that as well as, say, a JLSounds USB to I2S board - with its USB galvanic isolation and super low jitter clocks - remains to be seen. I have that on one of my Zoudio amp based systems. I think it sounds great, even at pedestrian sample rates.
Perhaps I'm just naive to assume when a manufacturer like TI says "Digital to PWM" that could possibly be what's going on inside.
Again, I have noooo idea how the feedback part actually works, so maybe they are just pulling the wool cap over my eyes / imagination...
From the article, I see the "loss" problem is that BT wont do HD audio, so you lose your 96K (or 192k) bitrate. LDAC being the best bitrate performer, "gets very close" to high bitrate fluency, while APT-X-HD is still down in the 48k scrambled egg territory ;')
So it would seem Zoudio is limiting the products ultimate performance through the Bluetooth V5.0 APT-X / A2DP / AAC, as its primary connection channel. The TAS5825 amplifier chips "Supports 32, 44.1, 48, 88.2, 96, 192 kHz sample rates" which I assume is achievable with a DIY USB to I2S connection, as I have done with mine. From what I've read, Zoudio will eventually offer USB connection implemented via the same chip that does the BT - which may support these higher sample rates as well - as a firmware update.
Whether it does that as well as, say, a JLSounds USB to I2S board - with its USB galvanic isolation and super low jitter clocks - remains to be seen. I have that on one of my Zoudio amp based systems. I think it sounds great, even at pedestrian sample rates.
I'm confused as to what you are arguing. what is your point? PWM is analogue, thus ... analogue and it is not 'digital all the way to the speaker posts' thats nonsense. or are you just attempting to move the goalposts? I dont know whats going on exactly in the feedback any more than you do, but its irrelevant. you have a digital input, followed by a 'digital to PWM' loosely translated as 'digital to analogue' its a mixed signal device with analogue output.
I guess the point is there's more than one way to generate a pulse width, corresponding to a time varying level;
You can take a digital word, convert it to an analog level, compare that to a ramping analog voltage and get a pulse. Check.
You can take a digital word, compare it to a word in some register that's counting up at a rate - and also get a pulse. Check.
One method happens in the analog domain, the other in the digital. That's all I'm trying to say. I agree what the pulse is used for after its generated is to make an analog signal.
You can take a digital word, convert it to an analog level, compare that to a ramping analog voltage and get a pulse. Check.
You can take a digital word, compare it to a word in some register that's counting up at a rate - and also get a pulse. Check.
One method happens in the analog domain, the other in the digital. That's all I'm trying to say. I agree what the pulse is used for after its generated is to make an analog signal.
Both of the pulses are analogue, but I think we've spent enough time on it 🙂 my apologies for ragging on about it; its a pet peeve. re bluetooth, depending on the hardware codec support, it doesnt matter if its 44.1 or 192. its lossy it IS getting there though.
hacking boards to get i2s inputs they dont come with is a pretty good way to add a bunch of jitter and you may find if you are trying to support higher sample-rates, it wont cooperate. Dont get me wrong, i'd love it if they solve this problem, not that it really matters to me; my streamer/dsp is built into my DAC and I cant see any amplifier I use not being built by me, for a long long time. I have effectively unlimited jfets/mosfets/chips at my disposal to build amps for the rest of my life. there is only a short multicore cable connecting to each speaker. i'm not building into my speakers, as i'd rather not build a DSP/DAC/AMP of this quality into every speaker I build. If I settle, I may eventually build it into the mains, but as WAF is of no consequence, I dont really care. bluetooth is a long way from having the bandwidth I would need, anyway.
hacking boards to get i2s inputs they dont come with is a pretty good way to add a bunch of jitter and you may find if you are trying to support higher sample-rates, it wont cooperate. Dont get me wrong, i'd love it if they solve this problem, not that it really matters to me; my streamer/dsp is built into my DAC and I cant see any amplifier I use not being built by me, for a long long time. I have effectively unlimited jfets/mosfets/chips at my disposal to build amps for the rest of my life. there is only a short multicore cable connecting to each speaker. i'm not building into my speakers, as i'd rather not build a DSP/DAC/AMP of this quality into every speaker I build. If I settle, I may eventually build it into the mains, but as WAF is of no consequence, I dont really care. bluetooth is a long way from having the bandwidth I would need, anyway.
Reality is more nuanced than this.
All signals are analog, whether a signal's also digital depends on how its interpreted. A certain kind of PWM can be interpreted digitally. That doesn't take away from it also being an analog signal. A ClassD amp is analog simply because no interpretation of any of the signals digitally is taking place.
PWM that can be interpreted digitally needs to be quantized in time, meaning all pulse widths are N*t where N is integer and t is the quantum of time. It seems likely that the TI chip is producing that kind of PWM at the output of its 'Digital to PWM' block. Thus the output signal can be termed 'digital' in that it can be interpreted as such. Feeding such a signal into an H-bridge driver is treating it as analog however.
yeah I already covered that; or rather I linked the wiki that did. So, if I sit here, vocalising zeros and 1's verbally to encode and decode a predetermined signal and you write them down with a pen, then read them out to the next person that comes in, you are digital?
I get what you are saying, but it does seem a bit loose to me 🙂
I get what you are saying, but it does seem a bit loose to me 🙂
If you're in agreement with me then it must follow that some of your earlier statements are false or misleading. For example :
Class D, with digital input stage and everything else analogue
To answer your question - no, I'm not digital.
Class D, with digital input stage and everything else analogue
To answer your question - no, I'm not digital.
I agree that it would have to be, to meet the definition, not that it is here. We dont have enough information to make that statement one way or the other. I would tend to think that the output of any PWM is analogue and certainly any ive worked with have been, but I am far from knowledgeable in this area. My peeve is with the statement 'digital all the way to the speaker posts'. this is clearly untrue in any and all cases; not least this one. So you are saying there is PWM that can be quantised in both time and level? you should update the wiki 🙂 any reference i'm able to find defines PWM as analogue. PCM is obviously digital.
Any signal can be interpreted digitally, even if it isnt; it depends how you define/encode/decode the information. someone standing by a river, putting 2 different coloured rubber ducks into the water at predetermined intervals could be framed as digital; as long as the receiver is able to decode it and the process is able to identify and correct for errors. Would you call that digital signalling? I wouldnt.
Last edited:
Pulse width modulation is a method used in dimmers and UPS systems, to drive what is essentially an analogue device, also here in class D.
Fans are analogue input, so are computer SMPS units, other examples exist.
The control system for doing that can be digital or analog, due to ease of use and versatility, the preferred method is digital.
Never heard of a digital speaker driver...
Fans are analogue input, so are computer SMPS units, other examples exist.
The control system for doing that can be digital or analog, due to ease of use and versatility, the preferred method is digital.
Never heard of a digital speaker driver...
I've already agreed that all signals are analog, so no disagreement here. Just some signals are amenable to digital interpretation. If you go and talk to @MarcelvdG he might tell you how its possible to embed a PWM stage in a purely digital modulator structure. Bruno's Mola-Mola DAC also (as far as I understand it) does PCM->PWM purely in the digital domain.
Yes, I agree there is no 'digital all the way to the speaker posts' unless the thing following those posts interprets the output digitally. No such 'digital speakers' exist to my knowledge and speaker posts don't make good digital interconnections.
I once tried updating a wiki on active filters, it got reverted double-quick so that taught me not to waste time there.
Yes I'm saying there is time quantized PWM which can be considered purely digital.
I'm saying all signals are analog by nature, not all such signals are amenable to digital interpretation. Natural PWM (non-quantized in time) would need to be quantized in time before being read into a computer and therefore its not in my estimation getting a digital interpretation.
A time base reference, or clock cycles would be essential in digital, I think.
Otherwise manipulation of the signal would be difficult.
Otherwise manipulation of the signal would be difficult.
Yes - any digital system would need an infinite clock rate to make a digital interpretation of natural PWM.
thats why I mentioned the predetermined intervals. its a silly example that wildly exaggerates the issue, as all of these things except the code itself are analogue, whether its an electronic oscillator, or a person doing their best to encode in the natural world at regular intervals. These all have varying degrees of accuracy, but as long as its accurate to the degree the RX expects, it may as well be 'perfect'. the intervals and variables are assumed to be accurate until they are not. in a digital system, if they are not, no code is transmitted and it may be retried until it its understood, given the parameters the receiver expects. an analogue system just 'plays' the incorrect/imperfect signal.
Would it have to be regular though? really? or could it just be any interval/pattern that the receiver is set up to read? Storage of digital code has no timebase until its read and the TX is not necessarily the same timebase as what the RX uses. It would have to be repeating in any case I would imagine, so yeah OK, regular. For the most part I think we are all in agreement though.
Regarding the wiki though, it wouldnt just be the wiki. I havent been able to find any examples of PWM with digital output with google, anywhere; but I do think I understand where you are coming from. Googling 'is there such a thing as a digital PWM' does not return anything. I found numbers of pages that claim to cover PWM and similar schemes in detail and none mention it. every example is a method for analogue, or digital control of an analogue device. Here is one example Yes, the analogue signal can be interpreted digitally, but pretty much anything can; doesnt make it digital. Perhaps i'm just being pedantic.