Possibly because the tweek is really a tweek depending on types of components, which op-amp, and layout. The AC CMRR of the op-amps seems to matter too.
Here are the links George posted. This one was the most interesting.
http://lavryengineering.com/pdfs/lavry-sampling-oversampling-imaging-aliasing.pdf
The other two just for continuity.
http://lavryengineering.com/pdfs/lavry-sampling-theory.pdf
http://www.lavryengineering.com/pdfs/lavry-white-paper-the_optimal_sample_rate_for_quality_audio.pdf
Remember? How could I possibly forget? You were very supportive. Thank you again for that.
jn
OK I'll read that I promise. I'm immediately put off by the practical vs theoretical stuff. Theory is based on the Dirac delta function, zero time extent and integral of 1. Introducing arbitrary sampling intervals that might be practical muddies up the waters from the very start.
Serious?? How about a million point FIR filter?
Last edited:
I was just giving him the benefit...I suspect at the time a million point FIR was out of the question.
Maybe with a cray?
Jn
You don’t need that many to make a halfway decent filter.
Here is a 1990 example implementation from BB:
http://tech.juaneda.com/download/DF1700.pdf
Its a total of 199 taps if I read the datasheet correctly.
Why didn't Lavry use it? Or does he use it but just refers to it as a brickwall?
Jn
My initial thought is that he was speaking conceptually or theoretically when bringing up those high-order analog filters and the examples without oversampling. That is probably how the very first CD players worked - although I highly doubt they had 16-pole Butterworth filters.
He introduces oversampling and then in the interpolation and imaging paragraphs describes the process of inserting zeros and filtering, which is the role of filters like the one I had linked.
All of his recent products do use filters very similar to the one linked as far as I know. They are typically integrated into the DAC chip now instead of being a separate chip. Some designers implement their own in an FPGA or DSP and bypass the on-chip filter. I didn't see the term brickwall in that one PDF (lavry-sampling-oversampling-imaging-aliasing.pdf) - but that's the typical aim of those filters.
Last edited:
Oh boy! Jn, with all due respect, but this remark shows that you have no clue what was/is? going on in digital audio playback field.
That DF1700 (identical to SM5813 by NPC) is one of the early filters that gave the "digital sound" its bad reputation. Its outdated, it accepts only 16 bit input data and its internal resolution (20x22 bit multiplier, 25 bit accumulator) is only so so. And it does not work at 88.2/96kHz sample rates. No way Lavry would use it. What Lavry uses in his top converters is some custom silicon with his own algorithms.
(Many moons ago I made a pretty good pocket money by designing and manufacturing all sorts of digital filter upgrades for various DACs and CD players, so I've seen and have listened to them all - from Yamaha to Mitsubishi, Philips, NPC, BB, PM and I know what I'm talking about). 😉
Last edited:
It's not great, but remember, it is from 1990! I had to search for an example from pre-1997 and I just happened to find that datasheet.
Even with all its shortcomings, I would take it over some 7th order+ analog filter.
The overall point was that you don't need a billion taps to implement it. I'm pretty sure most filters still use a similar number.
Last edited:
You dont know what you are talking about. yes, you can and it works better than you might guess.
Then as i said also, you can use headphones too.
THx-RNMarsh
Yo buddy, I resemble that remark!!
My two data points are DSP from late 70's, and Lavry's paper, with an occasional tidbit from EDN.
I really enjoyed the Lavry read, but did come away noticing a few internal inconsistencies but, more importantly, some external ones w/r to sampling theory fundamentals. Hence my questions as posed here.
The 1700 internals you mention do seem light at 25 bit AC. The motion boxes I use now run 48 bit math, I believe a 64 AC set, and very specific rules of engagement to prevent overflow as well as dipping too low in the mud lest encoder noise swamp the calcs, that and low bit error accumulation/truncation.
I can see him going the oversampled route to get away from chips like that.
Thanks for being gentle on me...😀
Jn
Last edited:
That is IT. That is all this comes down to. the rest is almost interesting and would not exist if back in the day, the standards were set higher. A very thorough AES study by many of the industries best and finest set the 'real' standard. Just because we consumers got stuck with CD is no proof itself of its adequacy to be accurate enough to be undetectable.
Its been one work-around after another with CD. Just to keep 16/44 going. Forget trying to defend it. CD is better in many ways than an LP system but worse in some other areas. 24/96 plus fixes all my issues. 24/192+ is wonderful if mastered well. JN got the jist of why. Then, just keep the HF under better control. All done. The rest is on the record side to do well by us. IMO the internet is a game changer for audio. We are not stuck with 16/44 CD any more. the consumers work around 🙂
THx-RNMarsh
Last edited:
I'm glad you understand now, I hope it will follow with other.
-----------------------------------------------------------------------
About opinion: You can doubt other opinion, but it can do with respect.
I'm sorry that I can not much express my self in this discussion, because my limited English language. But I can follow and understand about 90% of this thread.
RNM has mentioned the issues of >20KHz artifacts even in conventional CD player context. The first couple of generations of McIntosh CD players were (the contemporary best) Philips transports and boards, with Mc style, packaging and their additional 2 x 5532 LPF board. They apparently didn't consider out-of-band attenuation sufficient without the additional LPF. Judgement call.
All good fortune,
Chris
All good fortune,
Chris
Not a million points, but my first FIR filter from 1984.
We used it to filter ultrasonic test signals on the inner
containment of nuclear reactors, not important things
like audio. The bottleneck was Z80 processor I/O.
It is also an example that electronics tends to use up
all the available space.
Attachments
I certainly can but I would verify it to see if it's sufficient or not, which you didn't do.
As for "how high of frequency were you able to hear well enough to evaluate the accuracy?" per your claim "I said the CD sound is not accurate as you go towards the 20KHz limit" 😕
Is that what you did when comparing live vs recorded sound to figure out the accuracy?
- Status
- Not open for further replies.