If you want to call it ASRC or maybe resampling that seems a bit better, but semantics aren't that important. My first point was that the the ASRC introduces the exact same things that the non-OS crowd does not like about the internal oversampling filters. It would not meet the criteria as described by Kusunoki. Not that I agree with him, but he appears to be among the first to promote this idea.
My second point was regarding your statement that AD1955 can be run as a non-OS DAC. It's definitely NOT non-OS in any configuration. AD1955 runs internally at at least 128*Fs I believe. This is true even if you use external filter mode and do not use the internal FIR filters. As rfbrw said, of the DACs you mentioned, the PCM1704 is the only one that could actually be run at the input data rate. The others are delta-sigma converters...
Last edited:
All tweeters I know of are capable of reproducing ultrasonics to some extent. They behave as electroacoustical filters and may well have a half-power (-3.01 dB) point at or below 20 kHz, but their response is anything but brick wall.
If you guys still think SRC is the same as OS, you might be interested in how Cambridge Audio describe the 851D, a dual AD1955 384 dac, "ATF2 upsampling – All stereo sources can be upsampled to 24-bit/384kHz, kept jitter-free by proprietary, second generation ATF Adaptive Time Filtering algorithms co-developed with Anagram Technologies of Switzerland."
Semantics are very important because the term upsampling is accurate, distinct, and in common use.Cambridge 851D
As for the rest of your post, well, if I may repeat myself
¯\_(ツ)_/¯ .
Last edited:
Full-range drivers, just like tweeters, also don't behave as brick wall filters. Are you worried about ultrasonics causing intermodulation in the loudspeakers? If so, abstain from everything that audiophiles like: don't play high sample rate recordings, don't use a non-oversampling DAC without reconstruction filter and definitely don't play DSD.
Last edited:
I'm not concerned -. I already tested my ears when I was younger and found ultrasonics were irrelevant to me, given its filtered off.
On the topic of upsampling, I found the same discussion on pink fish media. And I also found Pavel's description of their development of dual mono nos ad1955 asrc dac, "We have developed and implemented a new upsampling filter algorithm which doesn't use zero stuffing. This new technique relaxes the filtering requirements and also allowed us to add Non-Oversampling option. There are now 4 filter characteristics available. These include a traditional linear phase SHARP , our proprietary "no-ringing" FLAT, carefully designed MINIMUM PHASE and unique NOS (Non-Oversampling). Also the precision of the upsampling engine has been increased to 66bit accumulator" www.audiopraise.com/forum/read.php?9,210
Clearly, OS is differntiated from upsampling, and nos doesn't automatically imply no filter. Again, accurate, distinct, in common use.
On the topic of upsampling, I found the same discussion on pink fish media. And I also found Pavel's description of their development of dual mono nos ad1955 asrc dac, "We have developed and implemented a new upsampling filter algorithm which doesn't use zero stuffing. This new technique relaxes the filtering requirements and also allowed us to add Non-Oversampling option. There are now 4 filter characteristics available. These include a traditional linear phase SHARP , our proprietary "no-ringing" FLAT, carefully designed MINIMUM PHASE and unique NOS (Non-Oversampling). Also the precision of the upsampling engine has been increased to 66bit accumulator" www.audiopraise.com/forum/read.php?9,210
Clearly, OS is differntiated from upsampling, and nos doesn't automatically imply no filter. Again, accurate, distinct, in common use.
Last edited:
Now it's my turn - I'm not sure you understand how Pavel and others have succeeded in making a NOS dual mono AD1955 using programmable SRC.
I don't actually care. I'm more interested in the way people on these forums interact. Some people like Abraxalito are genuine gift givers and raise their status by doing so. Others are takers and seek an ego boost. And likely there's everything in between.
And this is now so way off topic it's just silly. Do feel free to have the last word.
Last edited:
A non oversampling delta sigma converter doesn't work. Sorry to disturb your fantasy. You and your buddies can call whatever misguided creation you have non-oversampling, but that doesn't make it true. You've simply bypassed the 8x interpolation filter.
Why don't you ask Abraxalito if he thinks you can make a truly non oversampling AD1955. There is a reason he is using converters like TDA1387. Even if I don't agree with his philosophy, he understands what he's doing.
Whatever, if you think bypassing the internal filter does the trick for you, then call it what you will.
Last edited:
Either I don't understand your answer, or you don't understand the question, or both.
Regarding the question, in post #119 you wrote: "Worth considering is whether a tweeter is capable of reproducing ultrasonics. If it isn't, they aren't there to hear. They might, however, be messing with it can do, depending on the design."
I'm trying to figure out what you meant by the last sentence, "They might, however, be messing with it can do, depending on the design." I gather from the context that "They" means the ultrasonic signals, so apparently the ultrasonics might cause some problem in the loudspeaker.
As long as the loudspeaker behaves as a linear time-invariant system, any ultrasonics that are applied to it are either reproduced as ultrasonics or not reproduced at all, so I guess that's not what you meant by "They might, however, be messing with it can do, depending on the design." Hence, you must be referring to non-linear and/or time-variant behaviour of the loudspeaker. As far as I know long-term drift is the only time-variant behaviour loudspeakers have, so it must be the nonlinearity that causes the potential problem you refer to in post #119.
If the ultrasonics are strong enough and the loudspeaker non-linear enough, intermodulation in the loudspeaker could cause difference frequencies in the audible range. Hence my question whether that was what you meant. But then the answer about testing your ears doesn't make any sense, because the issue would be in the loudspeaker and not in your ears.
I was wondering whether to dive into this, but since I've been mentioned I'll say that its complete nonsense to claim a D-S DAC like the AD1955 can be non-oversampling. The whole premise of D-S DACs is to move the excess quantization noise out of the audio band which is impossible without large amounts of oversampling.
Just because its become common parlance to claim DACs based on such chips as PCM179X and AD1955 are 'NOS' doesn't make it correct. Just as almost all commentators on DACs made with chips such as TDA1543 and TDA1541 call them 'R2R'.
Last edited:
Can I just clarify if you mean oversampling in the broad sense that Chris does, i.e. any sample rate change, or in the more narrow definition that might be explained as "8x oversampling in the DAC IC" ? The reason I ask is I keep bumping up against different interpretations of established nomenclature and it seems DiyAudio has it's own socially constructed (terms we all agree on) way of talking about this?
Regarding upsampling language:
Some text books make some terminology distinction depending on whether only zero stuffing is done, or if interpolation filtering has also been done.
Other than that, different terms would only seem to distinguish how a signal is intended to be used, not the state it is in.
Nanoloop, is your preference based on state of the signal or some other basis?
Some text books make some terminology distinction depending on whether only zero stuffing is done, or if interpolation filtering has also been done.
Other than that, different terms would only seem to distinguish how a signal is intended to be used, not the state it is in.
Nanoloop, is your preference based on state of the signal or some other basis?
Don't worry about it Mark.I seriously think I will look back at this and think I wasted the time I had by spending it on the this forum.
Imagine you want to make and sell sigma-delta audio DAC ICs. You have noticed that some of the potential customers are fond of DACs that have no digital filtering except for the zero-order hold function (that is, keeping the output constant until the next sample arrives). In order to assure a nice, clean clock for the sigma-delta, you want to include an asynchronous sample rate converter, so the DAC's clock need not be synchronized to the source clock.
What you can then do is this:
1. Depending on a configuration bit set by the customer, do or don't pass the input signal through a normal digital interpolation filter.
2. Then pass it through an asynchronous sample rate converter that behaves as similar as possible to a zero-order hold. This can be done when the sample rate is converted upwards by a large factor: ensure that most of the output samples of the ASRC are 1:1 copies of its input samples, only when a new input sample arrives, one or a few output samples get interpolated values. ESS has a patent on an asynchronous sample rate converter like that, US 7330138 B2.
3. Then put the signal into a sigma-delta / noise shaper with unity signal transfer function.
Except for the inherent ultrasonic quantization noise of the sigma-delta, this will produce waveforms very similar to those of a multibit non-oversampling DAC chip when the filter of step 1 is disabled (even though the actual sigma-delta does and must run on a much higher rate than the input sample rate, for the reasons abraxalito explained in post #131).
What you can then do is this:
1. Depending on a configuration bit set by the customer, do or don't pass the input signal through a normal digital interpolation filter.
2. Then pass it through an asynchronous sample rate converter that behaves as similar as possible to a zero-order hold. This can be done when the sample rate is converted upwards by a large factor: ensure that most of the output samples of the ASRC are 1:1 copies of its input samples, only when a new input sample arrives, one or a few output samples get interpolated values. ESS has a patent on an asynchronous sample rate converter like that, US 7330138 B2.
3. Then put the signal into a sigma-delta / noise shaper with unity signal transfer function.
Except for the inherent ultrasonic quantization noise of the sigma-delta, this will produce waveforms very similar to those of a multibit non-oversampling DAC chip when the filter of step 1 is disabled (even though the actual sigma-delta does and must run on a much higher rate than the input sample rate, for the reasons abraxalito explained in post #131).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.