Do they? it’s not clear to me....
A ADC has Analogue input, digital output - converse a DAC has Digital input but Analogue output so how can it be claimed categorically by anybody that they are analogue – for sure they have Analogue sections, as they do Digital!
In a DAC, most silicon area (ignoring the I/O Pads) is used up by the digital section....
The statement "Digital - In which the data is represented by combinations of discrete pulses usually denoted by Os and Is" - pulses of 1's & 0's says nothing about the Energy content of these 1's & 0's, but they are still classified as digital - as you can see it comes down to personal interpretations....
John
This begs the question what a static alias product is, apart from a design flaw.JohnW said:
If one is coupling a problematic DAC to an amp, extra filtering is in order. One cannot blame an amplifier module for being designed to work well with a good source. I'm therefore not saying that filtering is always unnecessary, but that your example was not a case of the normal DAC output spectrum causing problems, which is what started the discussion. Instead, we're looking at a much less obvious culprit.JohnW said:
That said, did you verify that the CD player itself did not produce these spuriae in the base band? It wouldn't be the first case of a better amp laying stuff bare that went unnoticed before.
Ow c'mon you know you're stretching things thin here.JohnW said:
ie. it's incorrect to classify them as digital. Commonly they're called mixed signal devices.JohnW said:
Interesting.JohnW said:
Suppose I have a signal that says "10101010... ad infinitum", like a digital mute signal in DSD. This signal is represented as a nice square wave. I superimpose a small 1kHz sine wave. The digital signal says "silence", but the analog filter says "small 1kHz sine wave". Alternatively you can add 1kHz jitter and get the same outcome.
In other words, this signal encodes silence in saying but a sinewave in deeds. You call it digital because it says "silence" I call it analogue because in the end result it does "1kHz". This shows that the encoded content is very relative matter to the analogue filter. In this extreme case the encoded content is flatly ignored because the encoded content says "absolute dead silence" whereas the post filter makes something quite different from it.
What I'm saying is that whether a signal is digital is patenly not a matter of opinion. Having said that, you're entitled to yours.
However as I understand, you are just reconfirming the description of "digital" from my dictionary and my own perspective - in each case of the DSD datastream you mentioned, the signal was represented by a pulse stream of 1's & 0's - this does not violate the description of “digital” which you earlier agreed “is entirely right”.
“Digital - in which the data is represented by combinations of discrete pulses usually denoted by Os and Is"
The dictionary statement of “discrete pulses 0s & 1s” it does not differentiate the energy content of these 1’s & 0’s – only “which the data is represented by combinations of discrete pulses 0’s and 1s”
You have agreed that it's incorrect to classify ADC’s / DAC’s as digital as they should correctly be called mixed signal devices – which again was my VERY point that they are not ONLY analogue contrary to the statement: -
The Digital amplifier is a Mixed signal device just the same as any ADC / DAC, its function is to translate the Digital data in analogue – where this happens seems to be a point of contention – I’ve stated my perspective that the point the digital signal becomes “analogue” is the point it can be recovered WITHOUT the aid of any further processing / filtering – and in my Digital amplifiers this happens ONLY after the LPF on the output of the HBridge.
Now I fully understand, and you are also correct with your point of view that the “Reference Node” of my designs is a critical “Analogue” point, as any factor that changes the Energy CONTENT* of this discrete digital pulse stream of 1’s & 0’s be it Clock jitter, PSU noise / modulation, RDSon mismatch / non linearity & Rise / fall time errors etc will effect the analogue information encoded within these pulses – however they are still Digital in the purest sense that the Data “is represented by combinations of discrete pulses usually denoted by Os and Is” – as clearly defined by the dictionary as the definition of “Digital”.
* Energy content = Area of digital waveform
John
While one could agree that a "digital amplifier" (don't like that expression either) is a mixed signal device, the actual amplifying process is a perfectly analog operation.
With digital circuits one can make logic decisions, calculations and storage of data (digital filtering for instance is a combination of the last two). The output of digital circuits is therefore LOGIC STATES and/or NUMBERS.
The output of an amp is consisting of VOLTAGE and CURRENT, delivered by a PSU and controlled by a steerable device (or a combination of many of these), whether the latter is operating in a linear mode or switching mode doesn't matter - the function is analog in nature.
So there is no such thing as a digital amplifier (at least in my world).
Regards
Charles
With digital circuits one can make logic decisions, calculations and storage of data (digital filtering for instance is a combination of the last two). The output of digital circuits is therefore LOGIC STATES and/or NUMBERS.
The output of an amp is consisting of VOLTAGE and CURRENT, delivered by a PSU and controlled by a steerable device (or a combination of many of these), whether the latter is operating in a linear mode or switching mode doesn't matter - the function is analog in nature.
So there is no such thing as a digital amplifier (at least in my world).
Regards
Charles
As to the aliasing issue I experienced with a 3rd parties product, the demodulated spurie was very audioable as "birdies" and was not within the “Audio Band” of the CD player - where and how this fold back in the audio band occurred I’m not sure, however extra LPF on the input if the UCD reduced the audio effects.
However, lets not loose track of the point of my posting – there is a VERY worthwhile reason to lock the DAC & the Class D amp modulator (also SMPS), as no matter how much “Practical” filtering you have on the front of Analogue input class D amp – you will always find instances of spurie fold-down. I’ve seen cases of fold-down spurie in poor designs via mains interference between unlocked units!
This is an issue for ANY non-locked Class D be it Fixed frequency or UCD / hysteresis type approaches - I see to varying degrees time over with my fixed frequency’s designs - however as you would expect they are fixed spurie, and don't modulate Freq. with signal content.
The only guaranteed solution for out of band products with a non-locked analogue Class D is a brick-wall filter - but who wants that!
John
However, lets not loose track of the point of my posting – there is a VERY worthwhile reason to lock the DAC & the Class D amp modulator (also SMPS), as no matter how much “Practical” filtering you have on the front of Analogue input class D amp – you will always find instances of spurie fold-down. I’ve seen cases of fold-down spurie in poor designs via mains interference between unlocked units!
This is an issue for ANY non-locked Class D be it Fixed frequency or UCD / hysteresis type approaches - I see to varying degrees time over with my fixed frequency’s designs - however as you would expect they are fixed spurie, and don't modulate Freq. with signal content.
The only guaranteed solution for out of band products with a non-locked analogue Class D is a brick-wall filter - but who wants that!
John
And what follows that node is not an analogue control loop? Again you seem to be confusing form with content, electrical phenomena with data, a printed page with the information it carries.JohnW said:
I am not going to argue against convoluted arguments intended only to twist a perfectly sensible notion "digital/analogue" into something intractable that anyone can fish their own dinner out.
For me the argument is schluss. I don't need to prove that digital is numbers and that voltages and time are analogue. These things can fend for themselves without my help.
I can't remember saying there wasn't a good reason for doing so. I remember explaining to you, during your visit (and also on the forum only recently), how the synchronous operation of a DAC and a power amp such as in your design can even linearise the modulation process and yield better results than using a filtered analogue input signal.JohnW said:
That said, good reasons abound for making self-oscillating amplifiers, a point which I'm sure you will agree as well.
Well, I'm sure the diy community here would be most appreciative of such a design, so don't let us stop you
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
I heartily second that encouragement Sander.
Not to veer too far off topic, but your comments a bit ago in this thread regarding buzz words and marketing hype is on target, although I would add that from my recent experience an even more significant influence on purchasing decisions is the product appearance or so-called Industrial Design.
Harman has for example gotten an astonishing product life out of speakers made primarily from clear plastic. I thought they had reached the end of their appeal when they were redesigned to accept analogue inputs (as opposed to USB only, as stipulated by the first intended customer, Apple) and sat on store shelves for the most part. But I notice now that they are being pushed as an iPod accessory and getting attention on iPod user websites, and are doubtless streaming out the door once again.
...getting back on track....
Thank you Bruno and John for helping me clarify the distinctions among terms in my own mind. It is nice to know that even talented designers of this stuff can have disagreements---it indicates to me that the field is healthy.
Not to veer too far off topic, but your comments a bit ago in this thread regarding buzz words and marketing hype is on target, although I would add that from my recent experience an even more significant influence on purchasing decisions is the product appearance or so-called Industrial Design.
Harman has for example gotten an astonishing product life out of speakers made primarily from clear plastic. I thought they had reached the end of their appeal when they were redesigned to accept analogue inputs (as opposed to USB only, as stipulated by the first intended customer, Apple) and sat on store shelves for the most part. But I notice now that they are being pushed as an iPod accessory and getting attention on iPod user websites, and are doubtless streaming out the door once again.
...getting back on track....
Thank you Bruno and John for helping me clarify the distinctions among terms in my own mind. It is nice to know that even talented designers of this stuff can have disagreements---it indicates to me that the field is healthy.
Bruno Putzeys said:
Why you expect that cross-over distortion (or the lack of it) only comes into play when using digital PWM is something of a mystery to me
To clear up the confusion:
I think in terms of whether there is a sine wave in the system or not. If a conventional DAC is used to re-create a sine wave, then there's probably a linear opamp involved in the signal path (especially if I have a volume control after the DAC). I was assuming that the opamp would be class AB and thus, prone to crossover distortion. There's a significant difference between a signal at -80dB and at -10dB in terms of crossover distortion, this point can't be argued with. I guess the point I'm not considering is to use low crossover distortion class A amps in the signal path.
If I don't re-create a sine wave, and instead drive an analog-domain-PWM signal into a PEDEC stage, then there's no sine wave present in the system except at the post-filtered output. The difference between a -80dB signal and a -10dB signal is simply a change in the duty cycle. Crossover distortion is not quite the same with a switching amp as with a linear amp.
Bruno, I presume you mean load the output with a resistor to one rail or the other, to enforce class A operation of the output stage? Or are you talking about providing a moderately low impedance to local common when the output would be at the common potential?
Discrete class A does have many benefits, especially when you can take advantage of knowing your impedances to optimize operating points for noise and bandwidth. In a cost-challenged powered speaker design I did, which also was challenging from a signal-to-noise perspective, I used a number of bipolars with fairly high standing currents appropriate to the impedances involved. An unexpected side benefit was high immunity to external r.f., even including cellular phone emissions.
Discrete class A does have many benefits, especially when you can take advantage of knowing your impedances to optimize operating points for noise and bandwidth. In a cost-challenged powered speaker design I did, which also was challenging from a signal-to-noise perspective, I used a number of bipolars with fairly high standing currents appropriate to the impedances involved. An unexpected side benefit was high immunity to external r.f., even including cellular phone emissions.
I meant not driving a too-low-impedance load. Most IC op amps, certainly those we consider suitable for audio, have a standing current in the output stage of several milliamps. So if you drive a high-impedance load, or at least not a too heavy one, you shouldn't worry about the device entering class B.bcarso said:
Biasing to a rail is sometimes a good idea, sometimes not (if you use a bias current that is insufficient to unbias one of the output transistors, you've actually decreased the signal level at which the op amp goes class B).
The last project I'm involved in at philips (last week before I leave) involves a headphone amplifier for battery power. Has to run at less than 500uA. Discretes come in handy then.
---before you leave?? Not for good I hope--the glow lamp fabricators need people like you ;-)
Thanks for the clarification. Yes, the forcing of class A is not a panacea.
(off topic warning):
Low current operation is challenging. The last thing (more or less) I did for H*rman was the On Tour portable powered speaker, which manages to play pretty loud for about 5 hours on four AAA alkalines. With more reasonable levels it will go more than 24, especially with a bit of rest between sessions. It uses a plethora of discretes, mainly for cost, since good performing low current op amps were too expensive. Filterless class D parts from M*crosem* were used for the power amps, managing about 3W/ch with fresh batteries.
Thanks for the clarification. Yes, the forcing of class A is not a panacea.
(off topic warning):
Low current operation is challenging. The last thing (more or less) I did for H*rman was the On Tour portable powered speaker, which manages to play pretty loud for about 5 hours on four AAA alkalines. With more reasonable levels it will go more than 24, especially with a bit of rest between sessions. It uses a plethora of discretes, mainly for cost, since good performing low current op amps were too expensive. Filterless class D parts from M*crosem* were used for the power amps, managing about 3W/ch with fresh batteries.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.