SNR ratings on disc players

[sacguy231]

Maybe this is a stupid question - the SNR figure that is listed for a given CD player (98 dB, 104 dB, etc.), is that only for the analog outs? Rather, if you use an optical output from the CD player is there a different SNR for that output, or is SNR solely determined by the DAC that it's being processed by? I ask because I notice that manuals only list one SNR even though there might several different output methods (analog, S/PDIF, Toslink). Example - if I run a lowly CD player through a fabulous DAC via Toslink, is the SNR determined by the DAC or is something in the output/path of the original source device what determines or limits SNR. Just a curiosity that I have seen little to clarify this point on. Thanks!

 

[restorer-john]

The Signal to Noise ratio specified for CD players is with reference to full output via the Line Out jacks (analog) and usually with respect to the nominal (2.0V) output. The SPDIF (coax or optical) jacks have different characteristics and S/N is not applicable for this consumer interface over short distances.

As an example, a 96dB S/N with reference to 2.0V @0dBFS, suggests a residual noise of approximately 31.6uV.

When using a digital connection to an outboard D/A converter, the ultimate signal to noise is determined by (a) the source content and (b) the external D/A converter characteristics. The "transport" is all but irrelevant.

PS. Not a stupid question at all.

Maybe this is a stupid question

 

[sacguy231]

Fantastic response, thanks. I was prompted because I have a lowly JVC CD player (among some high line models as well) that lists a quite measly 94dB SNR...seems quite low for a CD player built in the late 90s. But, fed through a pre/processor with a great DAC I thought it sounded wonderful and got me to thinking if the SNR even mattered since I was bypassing the analog output stages of the source. Thanks again.

 

[cpt_paranoia]

that lists a quite measly 94dB SNR

If it's a 16-bit DAC, the best it can achieve is 98dB

https://www.edn.com/electronics-blogs/bakers-best/4329947/What-does-the-ADC-SNR-mean-

 

[E-Stat]

The effective limit is set by that of the recording and the power supply of the player. Don't forget that what you hear also involves the pre and power amps.

I've experienced wide variations with the preamp used.

 

[wyn palmer]

The relevant spec is something called SINAD- the ratio of maximum signal (usually at about 1db below full scale) to noise and distortion. As was written before, CDs are limited to about 98dB SINAD, and they generally are a bit worse.
I believe that the amplifier with the highest SINAD that is commercially available is the Benchmark AHB2 which has been measured at 113dB.
The human auditory system is purported to have a SINAD of c. 120dB or about 20 bits, although it gets worse as you get older.
The best commercial DACs manage to hit about 120dB SINAD. Combining the better ADCs with the better DACs and amplifiers produces a SINAD of about 110dB, or about 10dB (1.5bits) worse than the human auditory system and about 2 bits better than the best 16 bit CD player.
LPs have SINADS at best in the 70s region. 15IPS 1/4" analog tape decks are in the upper 60s to lower 70s effective SINAD.
Combining an RME ADI-2 PRO FS ADC/DAC with a Benchmark AHB2 gives a typical SINAD of about 110dB.
The tube amps that I have measured have SINADs in the 80dB region- they are both noisy and have pretty high distortion levels.
Whether this matters or not is a matter of opinion.

 

[tcdriver]

...The tube amps that I have measured have SINADs in the 80dB region- they are both noisy and have pretty high distortion levels...

The Fisher SA-300, tube amplifier, is specified at "More than 100dB below rated output..." and I have measured a little better than that on the one I owned.

 

[wyn palmer]

The Fisher SA-300, tube amplifier, is specified at "More than 100dB below rated output..." and I have measured a little better than that on the one I owned.

That's certainly possible, tube amps can be designed like solid state amps with sufficient feedback to linearize the generally rather high open loop distortion and particular care can be taken in the design to reduce noise, but that's certainly not universal. For example, SET amps may have very low SINADs, but are beloved of many people.
As I said, whether it matters or not is up to individual tastes, for example the Nelson Pass amps seem to measure quite poorly, but are generally perceived as sounding quite good.

 

[restorer-john]

The Fisher SA-300, tube amplifier, is specified at "More than 100dB below rated output..." and I have measured a little better than that on the one I owned.

You cannot compare S/N of a CD player and an amplifier. Also the Fisher SA-300 specifies 0.5% THD (which is terrible) and the residual is hardly great at -100dB (167uV) with respect to 16.7V RMS (35W). Consider a decent D/A converter has a residual of <3uV, around 55 times better.

The relevant spec is something called SINAD- the ratio of maximum signal (usually at about 1db below full scale) to noise and distortion.

Why people confuse the issue by talking about SINAD, I don't know. The OP asked about S/N. One test is residual noise vs maximum output in the absence of signal, the other is noise and harmonics in the presence of a notched out signal. One test stimulates harmonics, the other does not. They are not interchangeable.

CD players have always specified S/N not SINAD. That is since day one.
Amplifiers have always specified S/N not SINAD. Noise (20-20K or A-WTD) was also often quoted separately in uV for completeness. Amplifiers with higher power outputs got a bit of an easy ride with S/N numbers due to their larger voltage swing. Again, S/N numbers cannot be directly compared between amplifiers with different output powers.
The oft quoted amplifier THD spec was essentially THD+N until the last 25 or so years when fundamental analysis (FFT) allowed extraction of the harmonics from the noise to give separate THD and N and THD+N numbers.

 

[wyn palmer]

You cannot compare S/N of a CD player and an amplifier. Also the Fisher SA-300 specifies 0.5% THD (which is terrible) and the residual is hardly great at -100dB (167uV) with respect to 16.7V RMS (35W). Consider a decent D/A converter has a residual of <3uV, around 55 times better.



Why people confuse the issue by talking about SINAD, I don't know. The OP asked about S/N. One test is residual noise vs maximum output in the absence of signal, the other is noise and harmonics in the presence of a notched out signal. One test stimulates harmonics, the other does not. They are not interchangeable.

CD players have always specified S/N not SINAD. That is since day one.
Amplifiers have always specified S/N not SINAD. Noise (20-20K or A-WTD) was also often quoted separately in uV for completeness. Amplifiers with higher power outputs got a bit of an easy ride with S/N numbers due to their larger voltage swing. Again, S/N numbers cannot be directly compared between amplifiers with different output powers.
The oft quoted amplifier THD spec was essentially THD+N until the last 25 or so years when fundamental analysis (FFT) allowed extraction of the harmonics from the noise to give separate THD and N and THD+N numbers.

Well, CD players may have only specified S/N, but there was always an implied near odB FS signal level to compare the noise to (how can there be a signal to noise ratio without a signal, either real or implied, after all) - it is mostly quantization noise and hence is entirely relative to the full scale signal, and distortion was also specified relative to the full scale signal, so designers of converters always specified SINAD, even if the SINAD was/is decomposed in the specs of CD players.
Oh, and these measurements/designs were made using FFTs (Fast Fourier Transforms) from the early 80s- I know, I sat through design reviews on early ADI converters, and all of our opamp designs were simulated using FFTs to assess distortion.
Also there are a number of modern amplifiers that have equal or greater SINAD values (and indeed S/N ratio) to that of even 24 bit converters- The Benchmark AHB2 and the various Pudseys Class D amps for examples. Many people say S/N when they really mean useful signal to the impairments- or SINAD, so the argument goes both ways.
S/N therefore is but a component of SINAD, and the process of generating SINAD also generates the separately determined noise spectrum and the distortion components, which are reported in the analysis, and combined to produce SINAD.
S/N can be entirely independent of output power. After all power amplifiers are essentially gain stages with large output swings, and if they have the same input referred noise then they will have the same S/N ratio, independent of the gain, and independent of the output power.

So, a fair degree of conflation is reasonable.

 

[restorer-john]

S/N can be entirely independent of output power. After all power amplifiers are essentially gain stages with large output swings, and if they have the same input referred noise then they will have the same S/N ratio, independent of the gain, and independent of the output power.

Signal to noise specification in amplifiers, unless referenced to an arbitrary and consistent number (ie 1W@8ohms) are referenced to full rated power or in the case of preamplifiers and digital sources, with respect to the full rated output voltage. Conflation is not advisable.

A 200W/ch@8ohms amplifier with 25uV of residual and a 40W/ch@ 8ohms with the same 25uV residual will have S/N ratios of 124dB and 117dB respectively. A 9dB difference...

The Benchmark is quiet, but it's not a terribly high powered amplifier. It's rated residual is (A-Wtd) is 7.1uV and with respect to its highest rated voltage swing of 28.28V (@ 8 ohms), it therefore achieves a rated S/N of 132dB.
The Class Ds from BP are high powered, but noisier by a long shot. Even with the AES filtering for out of band, they make up for the poorer residual noise by swinging higher voltages.

CD players did not pick 2.0V nominal output for any other reason than marketing. The original industry standard specification was 1.4V and it was changed at the absolute last minute (March 1983) so target S/N numbers could be hit.

 

[wyn palmer]

Signal to noise specification in amplifiers, unless referenced to an arbitrary and consistent number (ie 1W@8ohms) are referenced to full rated power or in the case of preamplifiers and digital sources, with respect to the full rated output voltage. Conflation is not advisable.

A 200W/ch@8ohms amplifier with 25uV of residual and a 40W/ch@ 8ohms with the same 25uV residual will have S/N ratios of 124dB and 117dB respectively. A 9dB difference...

The Benchmark is quiet, but it's not a terribly high powered amplifier. It's rated residual is (A-Wtd) is 7.1uV and with respect to its highest rated voltage swing of 28.28V (@ 8 ohms), it therefore achieves a rated S/N of 132dB.
The Class Ds from BP are high powered, but noisier by a long shot. Even with the AES filtering for out of band, they make up for the poorer residual noise by swinging higher voltages.

CD players did not pick 2.0V nominal output for any other reason than marketing. The original industry standard specification was 1.4V and it was changed at the absolute last minute (March 1983) so target S/N numbers could be hit.

True, but the input referred noise plus gain is what's important as that is what allows you to make consistent comparisons.
The benchmark delivers c. 180w into 4ohms, and much more if bridged with no significant change in input referred noise, so what exactly is your point?
Increasing the output voltage on a converter essentially leaves the SINAD and S/N unchanged, except for the contribution of the final analog stages. Professional converters with adjustable input and output conversion levels have essentially unchanged SINADs and S/N ratios despite their changing absolute voltage levels.
Who cares about out of band noise, so of course AES is applied for the Pudseys.
This is all getting a bit too technical.
In essence everything is analysed using voltage gains and equivalent input noise/distortion in the relevant measurement bandwidth. All else is obfuscation.