?? Want the algo. which was used.
The EE's just came out with a streaming DAC to sell. I am going to throw the spec analyzer on it soon to see how well they did above 20KHz They didnt have access to my machine but now it is here and I am here and I am curious. If it isnt good enough, I will surely let them know.
-RNM
Last edited:
Gamelan music from Java island (Indonesia). I live at Java island from born until now. My father love to play gamelan, but he was not professional.
I was called Dick Marsh at work until the woman at the answering desk who had to page me in the building... she called and said she hated to page she wanted Dick. So, we agreed to page me as Richard after some 15 years of being Dick
🙂
-RM
🙂
-RM
Is that the grammar or sense of a page? My first boss was Dick Jewel, imagine the possibilities (no one ever complained).
Ask TNT, I am 100% sure it’s already in the DSP libraries. I can’t think of any building blocks that are not already pre-built there. Your guy has to write the glue code between the library and the I/O of your choice.
The questions that beg answering are:
Where did the sampled data come from, as I'm sure Scott didn't use my JPEG file.
When up sampling, what was the data length. Sufficiently long, it's trivial to understand that the full flat sine can be reconstructed. And was the resultant unity amplitude, Lavery's was about half amplitude.
But, that was not the answer to my question, which is being avoided by diversion.
How long a dataset is required to recover the pure sine of 22k sampled at 44.1k, as recall Scott did say too small a dataset will violate nyquist.
Nowhere is anybody addressing my original other question, which was, is NRZ linear.
I am not easily diverted, so would love to get clear answers to my actual questions. Lavry at least provided explanations, diagrams, some theory, and results. I prefer content that is testable and reviewable, so would like actual technical content. Thanks
Jn
Last edited:
I'm not sure you are reading the same thread I am. The issue was extended response recording of an acoustic instrument that has harmonics that are ultrasonic. I don't remember any pretentions about acoustic accuracy in reproduction.
Personally I think that would be a pointless pursuit anyway. Even if you got flawless reproduction of a specific live instrument in your living room you would have had to remove any room sound from the recording. I'm not sure how that would be relevant to reproducing a larger more complex group of musicians in a real space. This is where reverb tails are important.
In any case if time and circumstances permit I could make a trip to Auburn and make some recordings with Mark using whatever he has or I can also bring several options. Maybe even a Pacific Microsonics but those a real hassle to deal with. Real issue is DAW software that won't muck up the recording.
I may have some other easier, quicker options. Let me check.
I was thinking the same. Cymbals splash at least 360 degrees planar symmetrical, and vertically slightly different up/down. A mike can only pick up energy in one vector. Room reflects would totally change the sound, majorly mess up the mike capture.
Even an anechoic recording isn't the sane as live.
Jn
It is a FIR filter. The algorithm is multiply, accumulate, and shift.
I can't tell from the screenshot what software generated the filter coefficients, but they are almost all implementations of the Parks–McClellan algorithm.
Well, the output is modeled as an LTI system wherever I find it. I suspect if it wasn't linear, it would be a topic of discussion since it seems to be the standard output implementation.
That's something I have a lot of experience with. No mucking up will happen with a recording done here.
Also have a couple of different good quality 24/192 ADCs, and a couple of Grace 101 mic preamps (two different vintages on the Grace boxes though).
Last edited:
Mic directionality is actually more complicated. So it is with ears for that matter, although different from mics
Last edited:
Looks like intuition fails in the long and short cases. It takes an infinitely long filter to do it with infinitely small error, long or short sample set. We can never get to an absolutely perfect brick wall filter in practice. Its always a question of how good is good enough.
Also, if there were a phase error due to ZOH, it would be a linear error, not non-linear. That's because unwanted phase shift is a type of linear distortion.
Last edited:
No thats not it at all. I said the CD sound is not accurate as you go towards the 20KHz limit. 16/44 is not enough. A higher sampling rate seems to help a lot in regards to accurate sound. Now we are onto the effects of 20KHz limit and filters affects both within the AF range and spurious HF noise above 20Khz.
The comparison is to live known sounds. Recorded/playback 16/44 vs live musical instrument in the room. It is easy to disregard the room as i have told many times, you need to listen in near-field conditions anyway to avoid serious room issues. head phones can help as well.
Recorded close mic'ed as well as listening close to sources will tell a lot about accuracy of the sound thru listening.
Now we also thought about filters and the whole process has come into question. We have imaging accuracy, as well.
Finally, I hope we get to my question re single shot pulse repro with a system maybe more suitable for CW apps.... at least with CD using 16/44.1
In my opinion, I hypoth'ed that a BW similar to what we learned in all analog is needed... at least 40KHz BW, not 20. And higher sampling. AES long ago came to similar conclusion which was not about CD, per say. 24/96KHz was set as the PRO standard and i agree it is better. It sounds more accurate.
So, record play and compare 16/44 to 24/192+ to live. I am trying to tell the 'CD is fine' people, no it isnt fine as an accurate music source. Never has been. Only the high bit and sampling rate is accurate.
THx-RNMarsh
Last edited:
Would this be a fair rephrasing of your question? : How significant is the inevitable droop of D/A sample and hold? At this point, we're back in the magic kingdom of analog. The slope of droop isn't straight line either, so: how good is good enough?
Much thanks, as always,
Chris
I'd suggest that the recording and playback of music over a couple of loudspeakers is so fraught with peril that it wouldn't even help if the listener could have instant switching on demand between live and repro. Supersonic subtleties are of such a different order of magnitude that *any* recording (that includes supersonics) will work equally well, and be available to all interested.
All the best fortune,
Chris
I wasn't reacting to your issue. There was a question about ultrasonic output from musical instruments which is also of interest to me so lets see what is there and how much. Also how it is affected by recording and reproducing chains.
I'll bring my RTX which is quite good to 90+ KHz as well. I'll need to make some cable adapters for the B&K universe.
RE reconstruction filters- The usual analog filter on the output of oversampling DAC's doesn't seem to work all that well. The digital filters stop fine in the audio band but they do nothing for the images from the "over' sample rate. The typical VCVS analog filter expect enough opamp gain at the high frequencies to do its thing. Maybe an active filter for the range from 500KHz up is not the best idea. I switched to L/C filters which work quite well for me. I can't find the spectral plots right now but it was significant.
Yes. Dick was quite and acceptable name up until about 20 yrs ago before it was turned into a derogatory term. My wife has an uncle who’s name we’ve now changed to ‘Dirk’
Last edited:
- Status
- Not open for further replies.