Vinyl Record Noise Floor - A Call For Measurements

[Petrushka]

All you have to do is record a silent groove and look at the waveform with the vertical scale enhanced. Compare the noise level with a section of peak loudness. I’ve never done this with a recent pressing, but I’ve done it withe hundreds of older ones, say from the 70s.

I found the typical unweighted s/n to be 30-40 dB. That is after vacuum cleaning, and looking only at clean sections with no clicks or crackle. A good noise reduction algorithm can improve this by 5-10 dB without being noticeable.

 

[MatrixS2000]

Could someone tell me what the above mean?

Frequency on the right scale, time on the bottom, level in colour brightness.


A quick tutorial:

 

[Balle Clorin]

Given test records, SNR is given only relative to the 1 kHz mono signal on CA-TRS1007. It is given relative 3.54 cm/s RMS or 5 cm/s peak for the original JVC, and should be the same on the Clearaudio copy. Around -55 dB unweighted. Combined with the music example above, it would range between -55 to -60 dB to peak and -35 to -40 dB to RMS music power. If A-weighted, it is lower. That said noise figures are probably not static but vary with level, so those -55 to -60 dB levels are not reached in practice.

TRS-1007

View attachment 449780

Yes. REW is easy to use and available to all. All we have to agree on is FFT setting (ref FFT Gain) and number of averages to get comparable graphs, and they are more intuitive than spectrograms

 

[sergeauckland]

Frequency on the right scale, time on the bottom, level in colour brightness.


A quick tutorial:

Thanks for that. Isn't interpreting the brightness then a subjective opinion? Doesn't seem as useful or precise as a proper two-axis graph.

S.

 

[MatrixS2000]

Isn't interpreting the brightness then a subjective opinion? Doesn't seem as useful or precise as a proper two-axis graph.

S.

Definitely not subjective. The brightness is derived from the same process that you use to get level for a spectrum, so you are getting exactly the same information, just represented differently. I can generate a spectrum just by selecting any portion of the spectrogram. For example:

I find spectrograms much more meaningful to characterize and understand what is happening in a piece of music.

 

[Thomas_A]

I can do that too

View attachment 448949

So here is a vinyl rip I did a short while ago. Phono pre without needle dropped:

View attachment 448950

Needle dropped but no music yet:

View attachment 448951

Pretty horrible actually.

The silent part is wobbly. What is the spectrum? Do you have excess LF noise?

 

[digitalfrost]

The silent part is wobbly. What is the spectrum? Do you have excess LF noise?

How can you tell it's wobbly? The spectrum is in the picture I guess? Don't know if I have excess LF noise?! If you see something wrong with this I'm happy for pointers what to investigate but I don't really get what you're saying. Please be more verbose.

 

[SuicideSquid]

Agree. The below measurements for are for raw vinyl rips. The only changes are to bring the levels up to approx. 0dB level as I record low and removing the needle drop and remaining songs of the side. This is seen in the bottom right corner - Clip gain and delete(s). Each record spent 20 minutes in an ultrasonic cleaner, except for the used record which also received a scrub stage and a second trip back into the ultrasonic cleaner after the scrub stage.


Here is the first song from King Buffalo’s Regenerator album. I bought this record new and this recording is the first play on the record.

[snip]

Man this is awesome, thanks so much for taking the time. Super helpful and interesting.

 

[Thomas_A]

How can you tell it's wobbly? The spectrum is in the picture I guess? Don't know if I have excess LF noise?! If you see something wrong with this I'm happy for pointers what to investigate but I don't really get what you're saying. Please be more verbose.

I can tell from the first picture, but it would be confirmed only with spectral analysis. How cartridge and tonearm react to the vinyl irregularities can be significantly reduced.

Dropbox

www.dropbox.com

 

[DSJR]

I just like to be able to point to something concrete when discussing this stuff, recently there was a discussion in an FB group about an article (it's been posted here too) claiming "digital masters are ruining vinyl" or something like that, and a few people were claiming vinyl has better dynamic range than 16bit digital, which I think we all know is nonsense, but it'd be nice to be able to cite something solid instead of "yeah noise floor is somewhere between -30dB and -60dB and dynamic range of the medium is probably somewhere in between, but it really varies", which is probably the truth but is so... wishy washy.


I've found it's super dependent on the record, sometimes I don't notice at all and other times surface noise in particular can be very audible.

I can't remember the thread or poster (apologies again for senior moments), but one gent here, took a digital sourced track and 'vinylised' it, removing bass below 40Hz or so, monoing the rest under 120hz approx, de-essing the sibilance region and I can't remember if any other limiting was done in the mids. The DR figure increased quite a bit, simply by removing the excesses of the digital original music file.


When I used to make cassette tapes up for the car and used a Nakamichi for the purpose, the meters (analogue or bar-graph types) always had 'something rumbling away at bass frequencies around -30dB (Linn turntables back then were pretty quiet in themselves as regards drive and bearing noise). The RIAA playback eq could well reduce hf noise down to better than -65dB on a good pressing and decent stylus profile. So an analysis of background noise I suspect, would broadly follow the RIAA curve and the few plots I've seen, seem to indicate this view. Apologies if this has already been mentioned and shown.

 

[MatrixS2000]

I can't remember the thread or poster (apologies again for senior moments), but one gent here, took a digital sourced track and 'vinylised' it, removing bass below 40Hz or so, monoing the rest under 120hz approx, de-essing the sibilance region and I can't remember if any other limiting was done in the mids. The DR figure increased quite a bit, simply by removing the excesses of the digital original music file.

That only spoofed the DR # not the waveform. When you re-added the bass to the digital track, the DR dropped again. When you re-added the bass to the vinyl rip, the DR remained higher because the vinyl lacked the limitation that the digital files had applied to it.

 

[Balle Clorin]

Frequency on the right scale, time on the bottom, level in colour brightness.


A quick tutorial:

still kind of hard to interpret. not very useful, I rarely use it

 

[Audiofire]

What stylus and turntable was used was not clearly defined here. I have seen an expensive turntable with specs at around -80 dB signal to noise. That would then suggest an expensive turntable is not worth buying, except perhaps for other parts of the specs. Are you basically suggesting a relatively expensive turntable is not worth buying?

 

[sergeauckland]

What stylus and turntable was used was not clearly defined here. I have seen an expensive turntable with specs at around -80 dB signal to noise. That would then suggest an expensive turntable is not worth buying, except perhaps for other parts of the specs. Are you basically suggesting a relatively expensive turntable is not worth buying?

An honest -80dB is impossible, I don't care who's spec that is! As it's a Signal to Noise spec, it depends on what is the signal, and what is the noise. It's possible to claim 80 dB, but firstly, if the S part is the peak an LP can be cut to, i.e about 20dB above 5cm/sec. and the N part is filtered to ignore impulsive noise and anything above low frequencies. Without specifying the conditions of measurement, one can claim almost anything, like those 1000 watt ghetto-blasters...

An expensive turntable is likely to look more impressive, provide more bragging rights down at the pub, and maybe just perform as well as a mid-price Direct Drive Technics.

It's why I use vintage broadcast turntables. They were engineered to work 24/7, not just look pretty.

S.

 

[MatrixS2000]

still kind of hard to interpret. not very useful, I rarely use it

Interesting point of view. Why is the spectrogram the default display (switchable or overlayed with waveform) in a tool like iZotope, while the spectrum view is a secondary option that you have to switch to if it not very useful? Wouldn’t Izotope make spectrum view the default if it was the more useful tool?

 

[MatrixS2000]

An honest -80dB is impossible, I don't care who's spec that is!

Agreed…once the needle hits the record, you will not be at -80dB for any record. I think I showed that pretty clearly with the Tool's Lateralus capture where the signal was deeply masked by the noise.

 

[Thomas_A]

The quiet parts can reach -60 to - 70 dB for good vinyl and good setups. But as soon as signals appear noise increases. The -80 dB that is clalmed from some turntables is not from a vinyl record but probably from special jjigs for rumble measuremets. One could claim that rumble from a turntable should be lower than -65 dB not to add noise to the quiet parts when the best vinyl quailty is used.

 

[sergeauckland]

The quiet parts can reach -60 to - 70 dB for good vinyl and good setups. But as soon as signals appear noise increases. The -80 dB that is clalmed from some turntables is not from a vinyl record but probably from special jjigs for rumble measuremets. One could claim that rumble from a turntable should be lower than -65 dB not to add noise to the quiet parts when the best vinyl quailty is used.

Rumble should be measured using a rumble bridge. That separates the rumble measurement from any rumble recorded on a test LP. However, it is arguable that using the rumble bridge isn't the same as playing an LP. Nevertheless, it is a standard, so has merit.

I haven't heard of modulation noise affecting vinyl, although it may affect the tape the LP was cut from. It's very difficult to separate turntable rumble from vinyl roar or indeed cutting lathe rumble, but if it's not audible on the silent section between tracks, then it's Good Enough.

I am very sceptical of any rumble numbers, even -60 or -70dB let alone -80dB unless the measurement conditions are clearly specified as it's far too easy to cheat.

S.

 

[Audiofire]

Fascinating, so turntables became overengineered with time like amplifiers (for competitive marketing).

 

[Thomas_A]

Rumble should be measured using a rumble bridge. That separates the rumble measurement from any rumble recorded on a test LP. However, it is arguable that using the rumble bridge isn't the same as playing an LP. Nevertheless, it is a standard, so has merit.

I haven't heard of modulation noise affecting vinyl, although it may affect the tape the LP was cut from. It's very difficult to separate turntable rumble from vinyl roar or indeed cutting lathe rumble, but if it's not audible on the silent section between tracks, then it's Good Enough.

I am very sceptical of any rumble numbers, even -60 or -70dB let alone -80dB unless the measurement conditions are clearly specified as it's far too easy to cheat.

S.

I just want to clarify that the noise of -60 to -70 dB is from playing grooves with no signal with respect to potential peak output. Looking at the standard 1 kHz signal which is in relation to average RMS of 3.54 cm/s the SNR is no better than -35 to -40 dB for the most quiet vinyl. And this number is not diectly scalable to dBFS. Ie adding 20 dB to relate to 1 kHz peak will not get you to an SNR of -55 to -60 dB.