This is an analog section for the PCM1794 i put together some time ago. Perhaps some of you may be inspired.
The DAC is intended to be run in the DSP mode, that is, as a NOS-DAC. Some background.
Some people prefer NOS-DAC's since they don't have that aggressive FIR reconstruction filter that are mandatory nowadays. The midrange is reported to be more digestible and the whole presentation is considered by many to be more involving, etc.. But I have heard some grumbling about the treble in NOS-DAC's. And of course, there are a lot of overtones present, especially when run at 44.1khz. My subjective impression is that NOS-DAC's has a tendency to blow up treble details, at about the same manner as a aluminium tweeters do( which I can't stand ).
So I equipped this analog section with a 3:rd order Butterwoth LP filter that cuts at approximately 50khz.
The nice thing with a 3:rd order Butterworth is that it doesn't ring at all, watch the attached image. It's the severe ringing in ordinary oversampling DAC's that probably is the reason for sound degradation. Personally I'm particularly skeptic towards pre-ringing. Is there any sounds that has pre-ringing in nature?
OK. The section has been designed so that it will operate on only one 12V supply. The inputs will have a DC bias at around 1.2V which the DAC doesn't object against.
Some DIY-projects actually takes the analog signal directly from the DAC outputs. The DC may be as high as around 2.5V.
I actually tried at first to take the signal from the DAC directly but didn't think it was a good idea. I simply didn't sound any good. Probably there is a reason for the constructors to recommend a dedicated analog section.
One "funny" thing. At first I decoupled the bases of Q1 and Q4 with a capacitor to ground. That made the whole thing oscillate at some 60mhz with an amplitude of appr. 2V at the input. ( It's a couple of millivolts normally) Still it produced quite decent sound, at least so good that it wasn't obvious that something was wrong. I listened to the prototype the whole first day before I found this oscillation, and I can tell you that it didn't sound very good.
About the output section. This is a single ended variation. Like an LTP with one of the legs twisted around. Why so? Well I prefer single ended designs, they have a more natural harmonic spectrum.
If some of you want to build it, take contact and I will give some advices. There are several things that has to be considered. I currently have no proper PCB layout, since the prototype is heavily modified.
BTW, about that Butterworth LP filter. I also have implemented that in my PCM1704 multi bit DAC. It makes the typical NOS-sound more "domesticated", so to speak. It doesn't have so much hf artifacts. I use to call it a D-NOS - a domesticated NOS.
That DAC is of course in an other league. A sigma delta DAC doesn't have the abilities to present all small nuances in the treble.
The DAC is intended to be run in the DSP mode, that is, as a NOS-DAC. Some background.
Some people prefer NOS-DAC's since they don't have that aggressive FIR reconstruction filter that are mandatory nowadays. The midrange is reported to be more digestible and the whole presentation is considered by many to be more involving, etc.. But I have heard some grumbling about the treble in NOS-DAC's. And of course, there are a lot of overtones present, especially when run at 44.1khz. My subjective impression is that NOS-DAC's has a tendency to blow up treble details, at about the same manner as a aluminium tweeters do( which I can't stand ).
So I equipped this analog section with a 3:rd order Butterwoth LP filter that cuts at approximately 50khz.
The nice thing with a 3:rd order Butterworth is that it doesn't ring at all, watch the attached image. It's the severe ringing in ordinary oversampling DAC's that probably is the reason for sound degradation. Personally I'm particularly skeptic towards pre-ringing. Is there any sounds that has pre-ringing in nature?
OK. The section has been designed so that it will operate on only one 12V supply. The inputs will have a DC bias at around 1.2V which the DAC doesn't object against.
Some DIY-projects actually takes the analog signal directly from the DAC outputs. The DC may be as high as around 2.5V.
I actually tried at first to take the signal from the DAC directly but didn't think it was a good idea. I simply didn't sound any good. Probably there is a reason for the constructors to recommend a dedicated analog section.
One "funny" thing. At first I decoupled the bases of Q1 and Q4 with a capacitor to ground. That made the whole thing oscillate at some 60mhz with an amplitude of appr. 2V at the input. ( It's a couple of millivolts normally) Still it produced quite decent sound, at least so good that it wasn't obvious that something was wrong. I listened to the prototype the whole first day before I found this oscillation, and I can tell you that it didn't sound very good.
About the output section. This is a single ended variation. Like an LTP with one of the legs twisted around. Why so? Well I prefer single ended designs, they have a more natural harmonic spectrum.
If some of you want to build it, take contact and I will give some advices. There are several things that has to be considered. I currently have no proper PCB layout, since the prototype is heavily modified.
BTW, about that Butterworth LP filter. I also have implemented that in my PCM1704 multi bit DAC. It makes the typical NOS-sound more "domesticated", so to speak. It doesn't have so much hf artifacts. I use to call it a D-NOS - a domesticated NOS.
That DAC is of course in an other league. A sigma delta DAC doesn't have the abilities to present all small nuances in the treble.
Attachments
Last edited:
Great to see a discrete-based I/V and filter stage for a change 
I also use a discrete I/V for my NOS DAC (though based on TDA1387) and a fairly steep filter. Because its steep it has a fair amount of ringing but I so far haven't found that objectionable in the slightest. So this leads me to disagree that 'the severe ringing in ordinary oversampling DACs' is the problem with them. I sometimes use an upsampling filter to get 88k2 into my DAC, in that case I've not noticed any pre-ringing. Its generally at ultrasonic freqs to my ears anyway....
I also use a discrete I/V for my NOS DAC (though based on TDA1387) and a fairly steep filter. Because its steep it has a fair amount of ringing but I so far haven't found that objectionable in the slightest. So this leads me to disagree that 'the severe ringing in ordinary oversampling DACs' is the problem with them. I sometimes use an upsampling filter to get 88k2 into my DAC, in that case I've not noticed any pre-ringing. Its generally at ultrasonic freqs to my ears anyway....
OK, but why have a NOS-DAC when the analog filter is steep?
Maybe this supports my assumption that pre-ringing is particularly bothersome for our ears. A Butterworth filter has only post-ringing.
Abraxalito, how steep is your filter?
I think this topic is interesting since I read a document from the Meridian guys who put together the new MQA format. They show fairly convincingly that even slight "smearing" effect will be audible and that our hearing is particularly sensitive to time related distortion. Our hearing isn't so bothered by amplitude but more of attacks and decays and how a music signal rapidly changes. A FIR filter has a tendency to be in the way when the sine wave in question starts to change it's amplitude rapidly.
Maybe this supports my assumption that pre-ringing is particularly bothersome for our ears. A Butterworth filter has only post-ringing.
Abraxalito, how steep is your filter?
I think this topic is interesting since I read a document from the Meridian guys who put together the new MQA format. They show fairly convincingly that even slight "smearing" effect will be audible and that our hearing is particularly sensitive to time related distortion. Our hearing isn't so bothered by amplitude but more of attacks and decays and how a music signal rapidly changes. A FIR filter has a tendency to be in the way when the sine wave in question starts to change it's amplitude rapidly.
I originally designed a steep filter for my NOS DAC to see if it would help the treble. I found, just as you did that the HF was less than impressive coming straight out of a NOS DAC. And the filter did indeed fix it up nicely. But it didn't just fix the HF....
My filter doesn't have a constant slope like a Butterworth type, its more similar to an elliptic. The transition band goes from +3dB @ 18kHz to -47dB @ 34kHz so that's about 54dB/8ve or equivalent slope to a 9th order Butterworth.
I agree the topic is highly interesting - though I've not been convinced by anything I've so far heard on the subject of MQA. I agree with Bruno Putzeys that MQA is trying to have its own cake and eat it - if aliasing isn't a problem and hence we don't need steep slope anti-aliasing (or -imaging) filters, then how can ringing be a problem?
My filter doesn't have a constant slope like a Butterworth type, its more similar to an elliptic. The transition band goes from +3dB @ 18kHz to -47dB @ 34kHz so that's about 54dB/8ve or equivalent slope to a 9th order Butterworth.
I agree the topic is highly interesting - though I've not been convinced by anything I've so far heard on the subject of MQA. I agree with Bruno Putzeys that MQA is trying to have its own cake and eat it - if aliasing isn't a problem and hence we don't need steep slope anti-aliasing (or -imaging) filters, then how can ringing be a problem?
Actually it does because you bias the protection diodes and this has been shown to raise thd quite a bit. Voltage at the pins of the pcm1794 should be kept close to 0 (200 or 300mV either way is fine). Have you made distortion measurements ?
edit: passive I/V is fine with the pcm1794 btw but with small resistors value and a second stage to amplify to 2Vrms or so.
Last edited:
Abraxalito, that's a fairly steep filter. I hope you haven't realized it with OP amps.
I have the Tidal streaming and I'm very impressed by the MQA format. Of course Meridian want's the cake for themselves.
Who says aliasing isn't a problem? Is I see it, a high sample rate isn't there for bats to be happy. With a high SR, the filters - both aliasing and reconstruction - can be made much less aggressive.
00940, interesting about those diodes. Can these diodes cause a lot of THD, despite the actual signal is measured in millivolts? No, I haven't measured any THD, I don't have good enough equipment.
But those diodes probably was responsible for the poor subjective outcome when I took the signal directly from the circuit, that is, used a 200 ohm resistor as I/U converter.
Though, the DIY DAC "dddac" uses that approach. And the builders seems happy. But on the other hand, they haven't heard what a good analog section can do.
One more thing about the PCM1794. In DSP mode, the circuit expects the system clock to be not more than 64*fs, otherwise the signal will be crap. So I experimented and fed the circuit from a DSP so that it received four consecutive copies of each sample, hence I could turn up the system clock to 256*fs.
It worked as planned ( The goal was to enhance hf detailing ) but I noticed one frustrating thing.
The output from the current segments was still updated with a factor of 64*fs, so the precision wasn't boosted at all.
That made me draw the conclusion that the circuit actually detects absolute frequency.
Why did the constructors do this? I can imagine two reasons.
1. They thought 64*fs is more than enough.
2. The output current segments have problems turning off and on at a higher rate than appr. 3mhz.
This put together I have more or less given up hope on sigma delta devices.
I have the Tidal streaming and I'm very impressed by the MQA format. Of course Meridian want's the cake for themselves.
Who says aliasing isn't a problem? Is I see it, a high sample rate isn't there for bats to be happy. With a high SR, the filters - both aliasing and reconstruction - can be made much less aggressive.
00940, interesting about those diodes. Can these diodes cause a lot of THD, despite the actual signal is measured in millivolts? No, I haven't measured any THD, I don't have good enough equipment.
But those diodes probably was responsible for the poor subjective outcome when I took the signal directly from the circuit, that is, used a 200 ohm resistor as I/U converter.
Though, the DIY DAC "dddac" uses that approach. And the builders seems happy. But on the other hand, they haven't heard what a good analog section can do.
One more thing about the PCM1794. In DSP mode, the circuit expects the system clock to be not more than 64*fs, otherwise the signal will be crap. So I experimented and fed the circuit from a DSP so that it received four consecutive copies of each sample, hence I could turn up the system clock to 256*fs.
It worked as planned ( The goal was to enhance hf detailing ) but I noticed one frustrating thing.
The output from the current segments was still updated with a factor of 64*fs, so the precision wasn't boosted at all.
That made me draw the conclusion that the circuit actually detects absolute frequency.
Why did the constructors do this? I can imagine two reasons.
1. They thought 64*fs is more than enough.
2. The output current segments have problems turning off and on at a higher rate than appr. 3mhz.
This put together I have more or less given up hope on sigma delta devices.
Abraxalito, that's a fairly steep filter. I hope you haven't realized it with OP amps.
I have the Tidal streaming and I'm very impressed by the MQA format. Of course Meridian want's the cake for themselves.
Who says aliasing isn't a problem? Is I see it, a high sample rate isn't there for bats to be happy. With a high SR, the filters - both aliasing and reconstruction - can be made much less aggressive.
00940, interesting about those diodes. Can these diodes cause a lot of THD, despite the actual signal is measured in millivolts? No, I haven't measured any THD, I don't have good enough equipment.
But those diodes probably was responsible for the poor subjective outcome when I took the signal directly from the circuit, that is, used a 200 ohm resistor as I/U converter.
Though, the DIY DAC "dddac" uses that approach. And the builders seems happy. But on the other hand, they haven't heard what a good analog section can do.
One more thing about the PCM1794. In DSP mode, the circuit expects the system clock to be not more than 64*fs, otherwise the signal will be crap. So I experimented and fed the circuit from a DSP so that it received four consecutive copies of each sample, hence I could turn up the system clock to 256*fs.
It worked as planned ( The goal was to enhance hf detailing ) but I noticed one frustrating thing.
The output from the current segments was still updated with a factor of 64*fs, so the precision wasn't boosted at all.
That made me draw the conclusion that the circuit actually detects absolute frequency.
Why did the constructors do this? I can imagine two reasons.
1. They thought 64*fs is more than enough.
2. The output current segments have problems turning off and on at a higher rate than appr. 3mhz.
This put together I have more or less given up hope on sigma delta devices.
First, MQA is a pure marketing play. It has no value.
On your conclusions - Clearly a nearly 15-year-old device defines an entire category of DAC architecture... You don't have to worry about things like this if you use it as it was intended. Using it without an interpolation filter and with passive I/V is stupid. If you want to do that, use an IC designed to operate in those conditions.
A few things to think about:
If you think pre-ringing is a problem, consider using an external digital filter but implementing it as minimum phase FIR or IIR.
The latest AKM, Cirrus, and ESS DACs all have similar or better specs than PCM1794 and for sure the AKM and Cirrus parts output less out-of-band noise than the PCM179x. The AK4490/97 also have an external filter interface mode. The ESS DACs have a low output impedance and will be more amenable to passive I/V.
If you think pre-ringing is a problem, consider using an external digital filter but implementing it as minimum phase FIR or IIR.
The latest AKM, Cirrus, and ESS DACs all have similar or better specs than PCM1794 and for sure the AKM and Cirrus parts output less out-of-band noise than the PCM179x. The AK4490/97 also have an external filter interface mode. The ESS DACs have a low output impedance and will be more amenable to passive I/V.
Abraxalito, that's a fairly steep filter. I hope you haven't realized it with OP amps.
I did try that approach, several years back. See here : Active elliptic anti-imaging filter - diyAudio
But it didn't take long to discover that it didn't sound much good at all. So nowadays its all passive filters, or if active then discrete implementations.
Could you elaborate this a bit for me please?
This is what I assume from a very limited set of info I have so far (from this thread)...
The analog input stage will produce (apparently, pls correct me if I'm wrong...) around 1.2V DC of the same potential (positive) to the IoutL- and the IoutL+ pins. How is this going to affect the protection diodes in any way? Do you have 1794 analog-out stage internal drawing?
Regards,
Nick
You'll have to do a google search for "Calvin" and pcm1794. He did some research on this but it's not all on this forum. smms73 who did extensive work with the pcm1794 also suggested that there were protection diodes but connected to the rails. In which case, a small positive DC offset wouldn't be that much a problem. That's worth checking anyway. When I find some time, I'll try to reinstall my esi juli@ and see if I can test a pcm1798 dac I've got with various offset (but still an active I/V to avoid AC at the output).
In any case, the best results achieved with the pcm1794 (actually slightly better than datasheet) were with a passive I/V with 22r resistors.
In any case, the best results achieved with the pcm1794 (actually slightly better than datasheet) were with a passive I/V with 22r resistors.
Chris719, you seem a bit hectic. "Marketing play". "Stupid".
I think the MQA is great on Tidal. It's a nice way to encode a hires material.
The only thing that concerns me when it comes to sigma delta DAC's ( which are mandatory these days ) is ;
1. The number of hardware bits. PCM1794 has 64 of them or 128 in mono mode.
2. How fast the current segments are updated. PCM1794: 64*fs.
With this circuit, the "sample to sample" resoulution will be 64* 128 = 8192 discrete levels.
OK, I know the sigma delta algorithm averages out things so that a dynamic range of 130 db is achieved. But since it's an averaging algorithm, it can only produce those figures after a couple of samples.
A multibit device can immediately make subtle changes between two consecutive samples.
It would be interesting to se how the devices you mentioned behaves in this aspect.
And about those diodes. I think they can affect the signal only if a passive I/V section is used. A typical active one ( like mine ) has an input impedance of around one ohm and the signal will be around 8mV.
I think the MQA is great on Tidal. It's a nice way to encode a hires material.
The only thing that concerns me when it comes to sigma delta DAC's ( which are mandatory these days ) is ;
1. The number of hardware bits. PCM1794 has 64 of them or 128 in mono mode.
2. How fast the current segments are updated. PCM1794: 64*fs.
With this circuit, the "sample to sample" resoulution will be 64* 128 = 8192 discrete levels.
OK, I know the sigma delta algorithm averages out things so that a dynamic range of 130 db is achieved. But since it's an averaging algorithm, it can only produce those figures after a couple of samples.
A multibit device can immediately make subtle changes between two consecutive samples.
It would be interesting to se how the devices you mentioned behaves in this aspect.
And about those diodes. I think they can affect the signal only if a passive I/V section is used. A typical active one ( like mine ) has an input impedance of around one ohm and the signal will be around 8mV.
I checked out the some of the devices you mentioned, Chris. The AKM seemed not to be suited for hi end purposes.
The Cirrus 43198 looked promising, but has a DIY-unfriendly package.
It has a NOS-mode. Nice.
But there are no multi bit DAC's out today? This should open the market for those DAC's with discrete R2R ladders. Very expensive, and I can't see any advantage over IC-solutions such as the old PCM1704.
Any comment on that? Whats the advantage( if any ) with discrete ladders vs IC?
Well I hope that multibit DAC's will come back some day.
The Cirrus 43198 looked promising, but has a DIY-unfriendly package.
It has a NOS-mode. Nice.
But there are no multi bit DAC's out today? This should open the market for those DAC's with discrete R2R ladders. Very expensive, and I can't see any advantage over IC-solutions such as the old PCM1704.
Any comment on that? Whats the advantage( if any ) with discrete ladders vs IC?
Well I hope that multibit DAC's will come back some day.
I am not sure what you are looking at. AK4497 has comparable specs to CS43198 and both are the only chips that come close to the ESS 9028/9038. They are all better than PCM1794, particularly in distortion performance. I will admit that PCM1794 is still very good.
I don't know what self-invented metrics of performance or "hi-end" you are after.
Given your statements on sigma-delta, you may want to go back and study how they operate again.
I stand by my statement that MQA is pointless and you are better off buying a CD or downloading FLAC.
Overall, I don't understand why you need such a high performance DAC if you are going to destroy it's performance by using it "non-OS". Maybe you should try to get some TDA1541A if you believe that multibit and non-OS is better for whatever reason.
Last edited:
To further clarify. MQA is a vehicle for Meridian to collect royalties. It is a proprietary lossy codec. I find it questionable to use a lossy codec with high-res source material. If you must compress material then just use MP4/AAC and not line the pockets of a bunch of snake oil salesmen. FLAC is lossless and would be real "studio quality", not MQA.
Actually there are, just they're not immediately suitable for audio applications. For example Schitt used a 20bit AD part but it had to be deglitched to get a decent SQ. Other multibit parts are in general fewer than 16bits (14bits is a very common number for very high speed communications DACs which are available from ADI and TI at the very least). AD768 is an exception, it is 16bits but the linearity looks more in the realm of 14, it would benefit from some DSP trickery to improve the numbers.
PCM1704 is jolly expensive if you can still get it. I also fail to see what the advantages are for 'discrete R2R' technically compared to say a PCM63 or PCM1702. But discrete R2Rs are exceedingly fashionable nowadays.
Well, Abraxalito, I think the THD figures om modern sigma delta devices makes constructors in general think that a multibit DAC is pointless. Therefor there is currently no market for multibit devices.
But you can hear a constant grumbling among audiophiles; "multibit DAC's simply sound better". And now we have that old placebo issue again.
I think it's amazing that people still persists taking THD figures in a bunch. As if all distortion had the same subjective impact on humans. For example, 2:nd order harms are virtually impossible to hear if it's below 5 or 10 percent.
Sigma delta DAC's uses an averaging algoritm to obtain nice figures. This means that it takes several sample before it can "home in" at the desired signal, so to speak. Also low level hf changes won't get through the same way as with a multi bit DAC.
Now, all of you mathematical guys will object, I know that. But if someone is interested, I can present an example that is totally math-free that shows that a S/D device just haven't that low level performance at high frequency. Note: at high frequency. At lower frequencies the averaging method does it's job well.
Then something about "destroying the performance" with NOS-operation.
This isn't so easy to prove. It's a bit unclear whether NOS does something "right" or if it's just "adding pleasant dist" to our audiophile ears.
Lets look at a FIR reconstruction filter. It acts a bit like a pendulum that adds "mass" to the inbetween samples of the OS-DAC. It's usually of a very high order so that it can make -let's say a 20khz sine wave - look nice at the oscilloscope.
Now comes the point. In order to do so it kind of "resists" an abrupt change of said sine wave. Hence the pre and post ringing. So what it does is to make static sine waves look nice at the oscilloscope. But it opposes rapid changes of the signal amplitude.
OK, now we have a trade. What will bother our ears most, the famous smearing effect of the FIR filter or the large amount of hf artifacts that comes from a NOS-DAC.
Personally I think a "gentle" filtering algoritm is the best compromise. It doesn't make a 20khz signal look political correct, but what the heck - music watched at the scope is an ugly thing isn't it? So our ears has another aesthetics than our eyes, watching the scope.
Some day, hopefully all music will be sampled at 96 or 172 khz. Then all this NOS-discussion will be pointless.
MQA again. I'm not a fan of MQA. But from my point of view, it's so comfortable to sit at home in front of the Tidal interface and just listen to uncompressed music. Perhaps it has some elements of lossy behavior but it's good enough for me.
I don't want to bother downloading a lot of FLAC- files.
So if someone invents some other format, that's OK with me.
Regards, folk's
But you can hear a constant grumbling among audiophiles; "multibit DAC's simply sound better". And now we have that old placebo issue again.
I think it's amazing that people still persists taking THD figures in a bunch. As if all distortion had the same subjective impact on humans. For example, 2:nd order harms are virtually impossible to hear if it's below 5 or 10 percent.
Sigma delta DAC's uses an averaging algoritm to obtain nice figures. This means that it takes several sample before it can "home in" at the desired signal, so to speak. Also low level hf changes won't get through the same way as with a multi bit DAC.
Now, all of you mathematical guys will object, I know that. But if someone is interested, I can present an example that is totally math-free that shows that a S/D device just haven't that low level performance at high frequency. Note: at high frequency. At lower frequencies the averaging method does it's job well.
Then something about "destroying the performance" with NOS-operation.
This isn't so easy to prove. It's a bit unclear whether NOS does something "right" or if it's just "adding pleasant dist" to our audiophile ears.
Lets look at a FIR reconstruction filter. It acts a bit like a pendulum that adds "mass" to the inbetween samples of the OS-DAC. It's usually of a very high order so that it can make -let's say a 20khz sine wave - look nice at the oscilloscope.
Now comes the point. In order to do so it kind of "resists" an abrupt change of said sine wave. Hence the pre and post ringing. So what it does is to make static sine waves look nice at the oscilloscope. But it opposes rapid changes of the signal amplitude.
OK, now we have a trade. What will bother our ears most, the famous smearing effect of the FIR filter or the large amount of hf artifacts that comes from a NOS-DAC.
Personally I think a "gentle" filtering algoritm is the best compromise. It doesn't make a 20khz signal look political correct, but what the heck - music watched at the scope is an ugly thing isn't it? So our ears has another aesthetics than our eyes, watching the scope.
Some day, hopefully all music will be sampled at 96 or 172 khz. Then all this NOS-discussion will be pointless.
MQA again. I'm not a fan of MQA. But from my point of view, it's so comfortable to sit at home in front of the Tidal interface and just listen to uncompressed music. Perhaps it has some elements of lossy behavior but it's good enough for me.
I don't want to bother downloading a lot of FLAC- files.
So if someone invents some other format, that's OK with me.
Regards, folk's
That's only true if the oversampled bandwidth is insufficiently wide for fully relocating the in-band quantization noise in to. And/or the order of the noise-shaping modulator is too low.
I agree with this. This reason still remains something of a mystery.
What the reconstruction filter does is remove the repeating ultrasonic image bands which exist above the Nyquist frequency. That is its only job. When it does that job the originally encoded continuous signal is revealed. No fancy guessing required, regarding what the original signal looked like.
The ringing (whether pre or post) is an artifact of the reconstruction filter having a brickwall frequency response. This must be so, because they are the Fourier transform of each other. The brickwall frequency response slope, however, is itself dictated by wanting or needing to maximize the passband width, yet not violate Nyquist. A circumstance which defines the CD format.
Last edited:
Ken Newton.
"That's only true if the oversampled bandwidth is insufficiently wide for fully relocating the in-band quantization noise in to. And/or the order of the noise-shaping modulator is too low."
We are talking about different things.
If you want I can put together a document with nice drawings that shows exactly what I mean. I have come up with this assertion before here on this forum and people just dismiss my arguments too fast without thinking thoroughly what I mean.
Sometimes the math can make you blind for certain facts.
For example, there are a lot of "Free energy machines" on Youtube and I'm sure the constructors can come up with an impressive set of theories to explain the functionality.
I don't have to bother with that, I just need to say: "What about that thermonucelar law" ( Homer Simpson version ).
So. An S/D device cannot wave some wand. If it has 128 time slices each sample to make it's magic, then it can't represent a -90db signal of fs/2.
It will be zero or around -21db.
"That's only true if the oversampled bandwidth is insufficiently wide for fully relocating the in-band quantization noise in to. And/or the order of the noise-shaping modulator is too low."
We are talking about different things.
If you want I can put together a document with nice drawings that shows exactly what I mean. I have come up with this assertion before here on this forum and people just dismiss my arguments too fast without thinking thoroughly what I mean.
Sometimes the math can make you blind for certain facts.
For example, there are a lot of "Free energy machines" on Youtube and I'm sure the constructors can come up with an impressive set of theories to explain the functionality.
I don't have to bother with that, I just need to say: "What about that thermonucelar law" ( Homer Simpson version ).
So. An S/D device cannot wave some wand. If it has 128 time slices each sample to make it's magic, then it can't represent a -90db signal of fs/2.
It will be zero or around -21db.
Fair enough. Please post that document for us all to consider. I believe in keeping an open mind to new facts, so, I'll reserve further comment until then.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.