Digital Audio and Amplifier Noise Floor Comparison - Is 16bit/44.1kHz All We Need ???

[manisandher]

Can someone help me out with a calculation? The specs of my mono amps are:

 

- maximum unclipped output +/-20 Volts

- noise ~60 uV unweighted, 20-20 KHz

 

Would SNR = log(20/0.000006)*10? That gives me ~65dB, which doesn't seem right - the amps seem much quieter to me.

 

(Edit: the distortion spectrum in the manual shows the noise floor sitting at -135 dB below the 1 watt 1 kHz signal.)

 

Mani.

[esldude]

1 hour ago, manisandher said:

Can someone help me out with a calculation? The specs of my mono amps are:

 

- maximum unclipped output +/-20 Volts

- noise ~60 uV unweighted, 20-20 KHz

 

Would SNR = log(20/0.000006)*10? That gives me ~65dB, which doesn't seem right - the amps seem much quieter to me.

 

(Edit: the distortion spectrum in the manual shows the noise floor sitting at -135 dB below the 1 watt 1 kHz signal.)

 

Mani.

I'll assume that is peak to peak. So 28 volts rms.  You should use 20 instead of 10 in your calculations for voltage. You'll get a touch over 113 db. 

 

Spectrum views are deceptive because FFT will lower the view of the noise floor. You'd need to know FFT size to judge from that.

[manisandher]

15 minutes ago, esldude said:

I'll assume that is peak to peak. So 28 volts rms.  You should use 20 instead of 10 in your calculations for voltage. You'll get a touch over 113 db. 

 

Spectrum views are deceptive because FFT will lower the view of the noise floor. You'd need to know FFT size to judge from that.

 

Thanks Dennis - much appreciated.

 

Mani.

[sandyk]

10 hours ago, psjug said:

Even though you were trying to reduce noise, and did so, how do you know that the improvement that you are hearing can be attributed to noise reduction, and not an improvement in something else like some form of distortion?

psug

I can only go by what I saw on the screen of the C.R.O both before and after the changes. The original DC offset corrector from 1987  that mine was based around  used a TL071 I.C. and you could also see and hear the improvements simply by changing to lower noise opamps.

Initially I changed to a AD744, then finally a low noise OPA134.  Another constructor of my design suggested that I change from a 2n2 capacitor across R8  to a  4n7 as he had noted a further small improvement;. I also noted a further minor improvement, although the C.R.O. was unable to confirm this visually.

Any residual noise from the Offset Corrector goes into the base of a transistor in the differential  input pair.

Incidentally, this amplifier and the Preamp versions have a distortion of <.0006 %, so I doubt that you would readily  hear any further reduction in distortion figures.

 I use DC Offset Correctors because both my Class A 15W/Ch. and Class A Preamp/HA are fully DC coupled.

i.e. there are no input capacitors or a capacitor in the gain setting areas.

Alex

[fas42]

15 hours ago, esldude said:

My next question would be are there recordings with this kind of dynamic range?  Theoretically it would be possible to get close.  I think in practice much of the processing and microphone limitations will mean you can't get to 120 db in most recordings.  I think you'll struggle to actually get more than 80 db in recordings honestly.  

 

Where getting clean behaviour over a wide range is beneficial are recordings like this,

 

 

The intro may not start at a true 80dB or so down, but the impact of the opening is that much greater if the full range of the start is realised.

[fas42]

The theoretical reasons for why a wider range, bigger numbers is "nicer" all seem reasonable enough - however, in the real world of listening to music it's remarkable how well the human hearing system can adjust to what it hears, if all the "right things" are in place ... as someone who has heard the most 'technically brilliant' setups  and recordings come across as being banal, uninteresting; and the most unlikely, "rough at the edges" captures of a musical event take my breath away - I know where I would place my priorities, .

[esldude]

10 minutes ago, fas42 said:

The theoretical reasons for why a wider range, bigger numbers is "nicer" all seem reasonable enough - however, in the real world of listening to music it's remarkable how well the human hearing system can adjust to what it hears, if all the "right things" are in place ... as someone who has heard the most 'technically brilliant' setups  and recordings come across as being banal, uninteresting; and the most unlikely, "rough at the edges" captures of a musical event take my breath away - I know where I would place my priorities, .

Yes, but that isn't really the topic.  I'll take rough at the edges breath taking music.  I'll also take very well recorded versions of the same music.  

[fas42]

But, 16/44.1 can handle both rough at the edges, and very well recorded - as a means of storage of "all that matters" - the noise floor of 16/44.1 I have never found to be a limit; poor judgement, mistakes in mastering have always been the causes of something that seemed not quite right that in what I've come across.

 

Things like needing dither ... one has to create an extreme listening situation, to clearly hear the effects of not applying dither to CD tracks - a situation which would never happen in any listening to a music CD.

[psjug]

On 6/21/2019 at 7:20 PM, sandyk said:

psug

I can only go by what I saw on the screen of the C.R.O both before and after the changes. The original DC offset corrector from 1987  that mine was based around  used a TL071 I.C. and you could also see and hear the improvements simply by changing to lower noise opamps.

Initially I changed to a AD744, then finally a low noise OPA134.  Another constructor of my design suggested that I change from a 2n2 capacitor across R8  to a  4n7 as he had noted a further small improvement;. I also noted a further minor improvement, although the C.R.O. was unable to confirm this visually.

Any residual noise from the Offset Corrector goes into the base of a transistor in the differential  input pair.

Incidentally, this amplifier and the Preamp versions have a distortion of <.0006 %, so I doubt that you would readily  hear any further reduction in distortion figures.

 I use DC Offset Correctors because both my Class A 15W/Ch. and Class A Preamp/HA are fully DC coupled.

i.e. there are no input capacitors or a capacitor in the gain setting areas.

Alex

Did you only change the op amps in one of the circuits, or did you replace the bottom circuit with the top circuit?  I would think it is possible that what you hear is related to better power supply rejection, or from the top circuit providing more complete removal of the AC in the input.

[semente]

On 6/22/2019 at 2:46 AM, fas42 said:

But, 16/44.1 can handle both rough at the edges, and very well recorded - as a means of storage of "all that matters" - the noise floor of 16/44.1 I have never found to be a limit; poor judgement, mistakes in mastering have always been the causes of something that seemed not quite right that in what I've come across.

 

Things like needing dither ... one has to create an extreme listening situation, to clearly hear the effects of not applying dither to CD tracks - a situation which would never happen in any listening to a music CD.

 

Have you tried these audibility tests?

 

https://www.audiocheck.net/audiotests_dithering.php

[fas42]

2 hours ago, semente said:

 

Have you tried these audibility tests?

 

https://www.audiocheck.net/audiotests_dithering.php

 

Over the years I've examined what dithering does to the signal ... there's nothing like reducing 16 bit music to 8 bits, and then applying every variant of dithering to that, listening to what the subjective result is, . You then understand why the precise nature of dithering is critical - just applying some arbitrary dither may in fact make the situation worse.

 

Also having a test CD, with music passages attentuated by up to 60dB, with no dithering, is very instructive - it demonstrates that one can hear 16 bit digital artifacts, but they are buried so low down in level that only an extreme listening setup can make them obvious.

 

Only recordings which have faulty mastering on them have shown to be a problem - dithering is one of the last issue to have relevance if one is after convincing sound ...

 

So, no, I haven't bothered listening to those tests ...

[sandyk]

7 hours ago, psjug said:

Did you only change the op amps in one of the circuits, or did you replace the bottom circuit with the top circuit?  I would think it is possible that what you hear is related to better power supply rejection, or from the top circuit providing more complete removal of the AC in the input.

The bottom circuit was part of the original design from 1987 that I constructed. There weren't that many low noise opamps to choose from back then. My own Class A fully DC coupled preamps and 15W Class A were constructed many years later ,where I realised that the original circuit could be markedly improved in order to reduce noise going back into the differential pair,  as well as provide a much smoother correction with abrupt volume changes by using the 2 stage input filter.  Opamp noise becomes a problem with such high gain used. I also used a very low noise 10 x battery battery preamp ahead of my CRO to confirm the initial improvements with my Preamp.

 Later improvements could only be verified by extended listening, as well as reports from other constructors of my design.

 The Offset Corrector can be seen in the bottom right hand corner of the Preamp/HA PCB

[psjug]

13 hours ago, sandyk said:

The bottom circuit was part of the original design from 1987 that I constructed. There weren't that many low noise opamps to choose from back then. My own Class A fully DC coupled preamps and 15W Class A were constructed many years later ,where I realised that the original circuit could be markedly improved in order to reduce noise going back into the differential pair,  as well as provide a much smoother correction with abrupt volume changes by using the 2 stage input filter.  Opamp noise becomes a problem with such high gain used. I also used a very low noise 10 x battery battery preamp ahead of my CRO to confirm the initial improvements with my Preamp.

 Later improvements could only be verified by extended listening, as well as reports from other constructors of my design.

 The Offset Corrector can be seen in the bottom right hand corner of the Preamp/HA PCB

So to get back to whether you are hearing the improvement because of lower noise or something else... 

Don't you think that greater attenuation in the audible band in the revised circuit  should result in lower overall distortion from the power amplifier?  And so maybe this is the improvement you are hearing, not the lower noise?

[sandyk]

10 hours ago, psjug said:

So to get back to whether you are hearing the improvement because of lower noise or something else... 

Don't you think that greater attenuation in the audible band in the revised circuit  should result in lower overall distortion from the power amplifier?  And so maybe this is the improvement you are hearing, not the lower noise?

I believe that it is both, because the added 4n7 capacitor across the 1M resistor was mainly aimed at further reducing the HF noise  from the opamp itself at such a high gain, and  resulted in a further small improvement. The original version didn't have this capacitor, but a later version had 2n2 added. Feedback from other members resulted in me increasing this to 4n7.

Not too many are likely to agree that we can hear a further reduction in distortion below the already very low level of < .0006%

 

[fas42]

Trouble is, anomalies in the sound don't always present themselves as nice, neat, "measurable whenever you want to" behaviours. As a scenario, consider a power amplifier that uses a transformer which is barely good enough to ensure that the specs that look good on the glossy brochure are met - in real world use, the core of the transformer overheats, saturates as one tries for "big sound", and the power supply characteristics change quite significantly - and the subjective SQ suffers. Switch off, giving the circuitry time to restabilise, and then do some 'proper measuring' ... "Nope, the amplifier's working perfectly - all the number's look good!' ... The real problem is never caught ...

 

So many issues in substandard audio are like that - they don't sit obediently on the test bench, waiting to be 'trapped' in a "Ah- hah!" moment ... sometimes, lots of just thinking and idle fiddling with aspects of the whole give one clues as to where focused attention is worthwhile - and good movement forward then occurs ...

[psjug]

2 hours ago, sandyk said:

 

Not too many are likely to agree that we can hear a further reduction in distortion below the already very low level of < .0006%

 

Is this the 0.0006% figure for the op amp?  The distortion I was talking about would be from the residual of the filtered power amplifier's waveform (audible frequencies), which is part of what is sent back to the amp for the purpose of offset correction.  In the original circuit (bottom image), I think this has a 90 degree phase shift in the audible band, which can't be good.  In the updated (top) circuit, the two pole input filter gives a 180 degree phase shift, so I would guess this and also the much greater attenuation by the input filter would give a very good improvement, much more than the lowering of the noise.  Correct me if my thinking is wrong; I am not very experienced in audio circuit design.

[sandyk]

35 minutes ago, psjug said:

Is this the 0.0006% figure for the op amp?  The distortion I was talking about would be from the residual of the filtered power amplifier's waveform (audible frequencies), which is part of what is sent back to the amp for the purpose of offset correction.  In the original circuit (bottom image), I think this has a 90 degree phase shift in the audible band, which can't be good.  In the updated (top) circuit, the two pole input filter gives a 180 degree phase shift, so I would guess this and also the much greater attenuation by the input filter would give a very good improvement, much more than the lowering of the noise.  Correct me if my thinking is wrong; I am not very experienced in audio circuit design.

 

No, the <.0006% figure is for both the Class A Preamp/HA and the 15W Class A amplifiers.

The original Silicon Chip 15W Class A  article that it was partly based around, but never as a Preamp/HA, gave those figures as typical, but my version uses way better matched devices and resistors etc . as well as dual mono construction, DC coupling,  separate external transformers and a very low noise PSU PCB (<4uV) for the front ends as well in the 15W Class A P.A. The Preamp/HA uses the same low noise PSU PCBs for the individual channel PCBs .

 A friend of mine confirmed my Preamp as <.0006% distortion , but test equipment used couldn't go lower, so it is likely to be better than .0006%. The photo was of course just the PCB for a single channel.

 

Click on the image several times for a larger than life image.

[fas42]

A rule of thumb I use is that an amp design has to be capable of intrinsically doing 200KHz; that is, I would look at what the waveform at a reasonable voltage swing was like at that frequency - if it was a complete mess, skip that one; a very decent looking sine wave tells one that the circuitry is "fast enough", and is likely to do well in "real life".

[sandyk]

2 minutes ago, fas42 said:

A rule of thumb I use is that an amp design has to be capable of intrinsically doing 200KHz

 

A U.S.A. constructor of my Class A Preamp/HA measured it as -3dB at 1.5mHz

[sandyk]

2 hours ago, sandyk said:

 

A U.S.A. constructor of my Class A Preamp/HA measured it as -3dB at 1.5mHz

Correction.

-3dB  @ 1.5MHz

 

[sandyk]

 

For know-it-alls like Ralf11 who obviously thinks that I am making this up:

 My Class A Preamp is biased into Class A at a little over 100mA, although it can go much higher, and  It is capable of directly driving a 75 ohm correctly terminated cable, and will even pass through VHS quality Video.

[fas42]

The preamp is not the problem - it's pretty trivial to get RF performance at low currents, a cheap as chips opamp can do that. The fun starts when you want push out amps, into a speaker - you either use something which also serves to keep your house warm in winter, or, start to have battles getting a class AB output stage to behave itself - or use class D.

[Panelhead]

On 6/20/2019 at 11:10 AM, psjug said:

Just out of curiosity, have you or @Shadders seen measurements of finished amplifiers made from these kinds of kits?  I'm not sure I've seen measurements of any amplifier other than the AHB2 that meets the 96dB THD+N at 1W.  Maybe nCore is right on the edge of meeting this with a 22KHz filter.  Maybe there are a few others, but I bet they are rare.

  This is a measurement performed with an AP analyzer on my amp. It is dead quiet connected to 105 dB efficient Klipsch La Scala II speakers. 

Noise.pdf

[Ralf11]

6 hours ago, sandyk said:

 

For know-it-alls like Ralf11 who obviously thinks that I am making this up:

 

 

what's with the personal attacks, twinkie?

[sandyk]

25 minutes ago, Ralf11 said:

 

what's with the personal attacks, twinkie?

 

 Work it out for yourself Muffy from your posted reaction to my posted measurement.