Sometimes pedagogy can harm the didactic
Thanks for that,
yup this story of LP filter operating after Nyquist frequency helping to smooth what is below in the 20-20K Hz is conceptually odd.
One has to understand the upsamplings touch this 20-20K Hz range and the filter is here to attenuate the effects of the added up-samples resulting at the end in a more accurate 20-20K Hz band than a NOS one.
I believe the problem stands in the word "Low Pass filter" as if there is a frontier with a true split between each side of the Nyquist frequency...and where the nasty HF stuffs you're talking about were "only" just after this Nyquist frontier: one can simplify and think the problem stands only there and the ears anyway are not concerned because it's above 22K! Perhaps this confusion is a part of the NOS concept success ?
It's conceptually easier to "image" when one talks for instance about pre ringing in the 20-20K Hz range.
I have no idea about Chord dacs and Dave model in particular... and if they are the today' paradigm. I just meant the filter software sounds better to my ears due to my wallet paradigm.
.
Edit : Chord Dave better than a Total DAC with its simplier filter ?
Thanks for that,
yup this story of LP filter operating after Nyquist frequency helping to smooth what is below in the 20-20K Hz is conceptually odd.
One has to understand the upsamplings touch this 20-20K Hz range and the filter is here to attenuate the effects of the added up-samples resulting at the end in a more accurate 20-20K Hz band than a NOS one.
I believe the problem stands in the word "Low Pass filter" as if there is a frontier with a true split between each side of the Nyquist frequency...and where the nasty HF stuffs you're talking about were "only" just after this Nyquist frontier: one can simplify and think the problem stands only there and the ears anyway are not concerned because it's above 22K! Perhaps this confusion is a part of the NOS concept success ?
It's conceptually easier to "image" when one talks for instance about pre ringing in the 20-20K Hz range.
I have no idea about Chord dacs and Dave model in particular... and if they are the today' paradigm. I just meant the filter software sounds better to my ears due to my wallet paradigm.
.Edit : Chord Dave better than a Total DAC with its simplier filter ?
Last edited:
Some (for example Hans van Maanen) argue that any listening test done with sine waves is inconclusive because the human auditory system is not linear and time invariant. That is, intermodulation between ultrasonic signals just might be audible.
Besides, there is an article from a Japanese research group about an experiment with EEG equipment and a super tweeter that could be switched on and off. They claim that Japanese gamelan players have different brainwaves when listening to gamelan music band limited or not band limited to 26 kHz.
On top of that, domestic audio systems are often listened to by domestic cats who can easily hear up to 85 kHz - although most cats are not fussy about audio equipment, despite their superior hearing (Show off your audiophile cat).
Many way to "smooth" a cat
I agree about the cats, doggies are more a problem with hifi...
It's a strange hearing test as gamelans are made from many ringings stuffs than overlaps eachothers so it's hard to concentrate oneself on highest harmonics with such instruments as your attention is focused on the short (in time) ringings that answer from each other... I will say it's a rythmic music. Easier to test from our western perspective a standalone triangle or jazz cymbal for treble experiments.
I'm very surprised by the result of this nippon scientist team... but as they measure it, what else to say? I assume they double tested without the players knowing if it was placebo or switched off/on, cause scientists know what is bias and ABX... That might explain the ear sensivity with the different LP filters applied in the high end of the human ear sensivity? But as you said evolution made us not as pets : our ears are more focused on immediate close dangers : short & near sounds in the M&F bump, so around 2K Hz to 5K Hz in relation to the spl applied. It's already treble range btw
Saying that I most prefer indian tablas though more low in frequency and not metalic stuff but I'm not surprise indonesian gamelan dancers for instance are in trans due to disconance, fast rythm and maybe HF content as well. Notice it would be harder to test them with EEG
BTW, may you have some Lobby behaviour to make us a TDA1541L please... L as legacy : 21 linear bits this time with same architecture than the former A grade : best trebles I ever heard are from a selected TDA1541A chip in NOS and unfiltered dac layout.
Sorry tottaly off topic...
I agree about the cats, doggies are more a problem with hifi...It's a strange hearing test as gamelans are made from many ringings stuffs than overlaps eachothers so it's hard to concentrate oneself on highest harmonics with such instruments as your attention is focused on the short (in time) ringings that answer from each other... I will say it's a rythmic music. Easier to test from our western perspective a standalone triangle or jazz cymbal for treble experiments.
I'm very surprised by the result of this nippon scientist team... but as they measure it, what else to say? I assume they double tested without the players knowing if it was placebo or switched off/on, cause scientists know what is bias and ABX... That might explain the ear sensivity with the different LP filters applied in the high end of the human ear sensivity? But as you said evolution made us not as pets : our ears are more focused on immediate close dangers : short & near sounds in the M&F bump, so around 2K Hz to 5K Hz in relation to the spl applied. It's already treble range btw
Saying that I most prefer indian tablas though more low in frequency and not metalic stuff but I'm not surprise indonesian gamelan dancers for instance are in trans due to disconance, fast rythm and maybe HF content as well. Notice it would be harder to test them with EEG
BTW, may you have some Lobby behaviour to make us a TDA1541L please... L as legacy : 21 linear bits this time
with same architecture than the former A grade : best trebles I ever heard are from a selected TDA1541A chip in NOS and unfiltered dac layout.Sorry tottaly off topic...
I also confirmed that unfiltered output (leaking mirror image, non-interpolated) is not as harmful as people say. But this is probably only valid for certain audio system. I use custom transformer output DAC, which is intrinsically attenuating -3dB to -6dB above nyquist.
Is it possible that the tda1541 dac in nos mode produces different spectrum of Image that normal R2R dac?
Tda1541a is multibit dac but not r2r type like a pcm56 or equivalent analog device chips
Has anyone look at the image of wave form of nos tda1541a?
Tda1541a is multibit dac but not r2r type like a pcm56 or equivalent analog device chips
Has anyone look at the image of wave form of nos tda1541a?
BTW, may you have some Lobby behaviour to make us a TDA1541L please... L as legacy : 21 linear bits this time
Yes, I also prefer tda1541a over the pcm56, just wondering if the different architecture inside the tda1541a make it working well without a digital filter i.e. nonoversampling.
I tried the tda1541 in NOS and I quickly put back the 4x OS, it is way better if done properly. But I understand some people prefer the more blurry, cluttered sound of the non-os tda1541, to me it is not as good and seriously deficient.
To be honest the PCM56 K which is the best version of the PCM... is 8x OS and it sounds very poor, low-fi compared to modern dacs but it is analog like... true, like poor analog rig.
The tda1541 in NOS sounds almost as poor as the pcm56k in 8x OS... this is true. I can rate TDA1541 a 7/10 and PCM56k a 4/10 , the TDA1541 in NOS is like a 6/10
O and btw a 9 or 10 on 10 would be a DYNAVECTOR XV-1T CARTRIDGE or a Koetsu or Jan allard.
Last edited:
Very interresting. Thanks to all the explanation of this thread and good advices I'm going to try oversampling by myself.
Btw do you hear a difference between different sampling rates please ? As far I know the TDA1541A for instance can go x8 for the 44.1 K Hz. (not for the 44 K Hz but materials are more rare with this rate... or perhaps x4 upsampling is enough at playback ?)
Or a filter parameters advised if you're using software and not hardware ??
Btw do you hear a difference between different sampling rates please ? As far I know the TDA1541A for instance can go x8 for the 44.1 K Hz. (not for the 44 K Hz but materials are more rare with this rate... or perhaps x4 upsampling is enough at playback ?)
Or a filter parameters advised if you're using software and not hardware ??
the idea behind OS is to get rid of the encoding, like in a Class D amp.
4x is good at the time, I don't think the TDA supports 'well' other higher rates. Modern DACS support higher OS without any issues.
I use I think 4399 AK dac and it supports DSD 22Mhz and has 128X OS
I only have one DSD recording so far and It plays well with custom software but it is not a night and day difference, just good.
4x is good at the time, I don't think the TDA supports 'well' other higher rates. Modern DACS support higher OS without any issues.
I use I think 4399 AK dac and it supports DSD 22Mhz and has 128X OS
I only have one DSD recording so far and It plays well with custom software but it is not a night and day difference, just good.
Tda1541a can run up to 384khz sample rate (16x) in simultanous data mode. Pedja has done this.
It isn't clear to me from the description in their article whether the test was double blind or only single blind. Single-blind tests can be quite unreliable due to unintended nonverbal communication.
By the way, there is a thread somewhere on this forum about interpolation filters in a small Spartan 6 FPGA. The starter of that thread also sells boards with his filter. The advantage compared to standard hardware interpolation filters is that you can easily avoid imperfections such as intermediate rounding and that you can play with coefficients. If I remember well his filters support two coefficient sets, by default a linear-phase and a minimum-phase set.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.