Just wait and see Ed, your sample of seven (participants) is still a bit small - not yet representative of the community of all record listeners.
HBt.
I think the poll has a selection bias - people who could hear a difference are more likely to vote.
Ed
Ed
In my book, this would be call for the extra $$$ & complexity of doing DBLTs to check if this is a true golden pinna(e).
I know some of my DBLT panel can hear a 1kHz sine wave -20dB BELOW the noise level of a properly dithered 16 bit signal chain. Not sure this would hold in the presence of your impulse chain.
Headphones - the only choice. IMO 20dB below noise level is audible. Do not forget we speak about integrated noise level over 20Hz - 20kHz, not about noise density as seen as a “noise floor” in spectrum analysis.
That's only possible if the Digital Signal Chain is properly dithered. When I did the test, end of the last Millenium, there were many Digital Signal Chains which weren't properly dithered. For these, the signal got very 'crackly' well before the noise level and usually disappeared completely when it got below the noise.
If you want to test if a Digital Signal Chain behaves properly ie like a good analogue chain, at low levels, play a good piano recording and 'record' it at lower & lower levels until it goes well below the noise. It should still sound like a very noisy but good piano recording. 😊
You need a matched attenuator / amplifier to properly account for the recording 'level'. ie attenuate the signal -60.0 dB before 'recording' or passing it through your Signal Chain .. and amplify it cleanly by 60.0 dB at the other end. This gets around the need for headphones, quiet room bla bla
You need to consider this 'matching' very carefully at very low levels below -60dB Full Scale cos the noise upsets simple calculations.
I'm open to that possibility but would want to test it on a DBLT 😲
Last edited:
Thank you for your comments, but please let me remind you that the ADC's have passed a long way of improvements in the past 25 years. In my case, I get results >>20dB better than 20 years ago, regarding SNR and distortion residuals. SNR > 120dB, in real life conditions, is a fact. Together with 32-bit floating.
This is also possible to be done. I still feel a smell of voodoo, talking about differences in vinyl playback, with its physical limitations, that are well known to both of us.
Perhaps people that wanted to hear a difference but could not were more likely to vote...
Perhaps people that know what you're looking for, understand the poll, and have a personal bias toward a result, never listened to the file and voted in order to steer you off track.
.
.
.
It's a poll, not a study, after all. You have to assume some good will / truthfulness and rigor from the participants in order to draw any conclusions (real or otherwise).
Even your statement above seems odd to someone (me) that has no idea what you're trying to accomplish. I haven't owned a TT in decades, and I haven't the foggiest of clues about the engineering / equalization.
The first post describes the signals. The poll asks us if we can "hear" them. Sure, I can hear every single one. However, what I think you want is if we can hear what you've "added" to the clicks in samples 2 and 4? Above, you state that you're looking for people to hear a "difference" => a difference between X? and Y? What?
I didn't enter my results b/c I wasn't sure I was 'properly' listening for what you wanted and exactly what I was suppose to "hear" or declare that I heard if in-fact I heard "it". So, as a thoughtful participant, I thought "bad" data would be worse than entering a guess. So, I did not offer my "results".
There were 16b converters which formed textbook 16b digital signal chains in the early 90s. I think it is fairer to say there are more of them today than 25 yrs ago.
I was pleasantly surprised that the 16b performance of a Lenovo T21 laptop Line Input was 'textbook' when I emerged from a decade of being a beach bum. Sadly, this is no longer the case in laptops or even soundcards; the makers prioritizing fancy bells & whistles like built-in reverb, zillion (claimed) bits, MegaHz sampling etc. over textbook performance
Guru Scott Wurcer tells the story of a 24b Matsushita converter he was involved with which had textbook 24b performance. He said they concluded it wasn't feasible for production. The best '24b' converters today have about 20b performance. I'm not sure 32b FP is an advance on '24b' at the present SOTA.
Quite so. I was just making an OT comment on the audibility of stuff below noise.
Easy can hear 2 and 4 on 30EUR over-ear headphones. Even with moderate levels. On low casual background level cannot hear 4.
OK, I did, by a precise stepped attenuator. Would you tell me if it is "good enough" to prepare a test with vinyl?
I would really like to know what is your goal, or better please show me YOUR examples.
And I do not understand why to work with such attenuation. We are about to check OUTPUT of the phono preamp, thus not signals in mV.
PMA, as I said earlier, i was just commenting on item 4 in EdGr's test which involved signals near or below the noise. I was explaining how I did this repeatably in Jurassic times to determine good and bad Digital Signal chains.
So no. it has little to do with phono preamps ... but perhaps some relevance when people claim to hear such stuff in a none DBLT.
It appears svp is the one who can detect the low level signal in item 4 of EdGr's test. I must now try to find some 30EUR over-ear headphones to see if I'm as discriminating as him or if nearly 30 yrs as a beach bum has trashed my hearing
So no. it has little to do with phono preamps ... but perhaps some relevance when people claim to hear such stuff in a none DBLT.
It appears svp is the one who can detect the low level signal in item 4 of EdGr's test. I must now try to find some 30EUR over-ear headphones to see if I'm as discriminating as him or if nearly 30 yrs as a beach bum has trashed my hearing
I have re-read fist post if there are any conditions for this test. There are none.
Earphones: Superlux HD681, amazon rubbish. But they are comfortable and I am using EQ APO equalizer with my preferences (basically lower highs and some custom dips on high mids), ~-3dB on 2kHz.
Amp: PreSonus Studio 26c (that thing is average and with parasitic noises...) as headphones preamp on max level analog headphone volume knob, 192kHz.
PC: 40/100 volume.
Yes I can EASY hear "4" on 40/100 volume, but not on 20/100. 28-30/100 is the limit on which I can hear. Usually I background listen music when working on 4-8/100, youtube videos on 6-10/100 so on the normal levels I will hear it. I am protecting myself from loud noises all the time I can.
"2" is more like an ugly feeling than sound. I am more inclined to say, that my earphones or smth in signal path ads smth to signal, but definitely ABX test will be close to 50% when done... It is like the sound went from emetic to more emetic 🙂
My first try was on standard generic bluetooth IEM headphones. I was hearing the beep, but suspected compression, from whatever reason, idk. Today I tried again, with Tin Hifi P1 Plus dynamic planar IEMs and cheap on-cable usb dac and it's there despite the rubbish going from the dac. I'll try again at home with my qnktc dac, and Elac FS187 speakers. But not today, gonna attend a live performance.
My personal anti placebo was randomly skipping in the last 30 seconds to make sure I'm able to tell where the beep is and where not. It was instantly recognisable.
Also I'm still able to admit that it's the lousy hardware I used, particularly dacs, which is causing me hearing something extra. That's the reason I want to try once more on QNKTC which is the best one I have on hand.
My personal anti placebo was randomly skipping in the last 30 seconds to make sure I'm able to tell where the beep is and where not. It was instantly recognisable.
Also I'm still able to admit that it's the lousy hardware I used, particularly dacs, which is causing me hearing something extra. That's the reason I want to try once more on QNKTC which is the best one I have on hand.
Last edited:
I would not use Bluetooth due to the lossy compression. The test should be conducted on a high-fidelity system that can handle the 63dB peak-to-average.
@ItsAllInMyHead - I switched on and off the small difference at a 0.5Hz rate. You are listening for a small change every second.
The poll grew out of a debate about the audibility of defects in some phono pre-amps. I am trying to inject science as best as can be done over the Internet while having fun too. 🙂
Ed
@ItsAllInMyHead - I switched on and off the small difference at a 0.5Hz rate. You are listening for a small change every second.
The poll grew out of a debate about the audibility of defects in some phono pre-amps. I am trying to inject science as best as can be done over the Internet while having fun too. 🙂
Ed
Your click waveform looks like a sinc:
To get such a waveform out of an ideal RIAA preamplifier, you would have to take a series of Dirac impulses, pass them through an ideal 20 kHz low-pass filter and through an inverse RIAA network. The ideal 20 kHz low-pass could be a very rough model for a cartridge, but how the clicks are supposed to get inversely RIAA corrected is unclear to me.
All in all, I don't think you can conclude anything at all. The test is sighted, also known as unreliable, and the waveforms are unrealistic.
@MarcelvdG - The test has shown that some people can hear the +1 error.
The impulse is a reasonable approximation to a click because the phono cartridge responds to velocity, and the RIAA network is mostly an integrator.
Ed
The impulse is a reasonable approximation to a click because the phono cartridge responds to velocity, and the RIAA network is mostly an integrator.
Ed
If I might add another $0.02 from 2 decades of DBLTs ... some with similar test signals to EdGr's ...
Such signals are VERY fatiguing. If you don't pick out the difference within the first few minutes, you rapidly lose the ability to discriminate .. especially under the psychological stress of a DBLT which might just be checking if you are a deaf Golden Pinnae 😊 Remember, you are NEVER told what you are listening too.
Such signals are VERY fatiguing. If you don't pick out the difference within the first few minutes, you rapidly lose the ability to discriminate .. especially under the psychological stress of a DBLT which might just be checking if you are a deaf Golden Pinnae 😊 Remember, you are NEVER told what you are listening too.
As far as I can tell, it only shows that some people think they hear the + 1 error in your click signal. Maybe they might also have misunderstood the test and compared part 2 to part 1 instead of part 2 to part 2.
The peak value of the click does get higher, but only by about 0.015 dB, and the peak shifts somewhat to the left, see the attachment. I've calculated a sinc waveform and what you get when you pass it through a filter with transfer 1 + sτz, that is, the sinc waveform plus τz times its derivative to time, where τz = 1/(2π · 200 kHz).
Fair enough, if the original mechanical click looks like the impulse response of a low-pass filter with a bandwidth of 20 kHz or more.
Last edited: