John Curl's Blowtorch preamplifier part II

[gpapag]

Originally Posted by simon7000

Look at the THD curve for a 10mV signal. According to the data sheet there is a lot more distortion at that level. Would you hear that or is it all noise? Then look at the distortion level for 10 volts out into 600 ohms would a pro system see those levels and what would it sound like?

Distortion is not a fixed number.

FFT can tell how much of it is signal correlated distortion. The rest has to be uncorrelated noise.
As for the ”hearing it” no one can tell, no use in assuming.
Just for the masking effect, the lower the level of the signal (and the higher the freq.), the less the masking effect.

I noticed we are repeating ourselves –and each other- here.

Originally Posted by PMA

Do we care enough about noise background, noise "bottom"? The "signature" for low level signals. That means noise, power supply residuals, interference residuals. Not just harmonic distortion.

George

[gpapag]

Originally Posted by Esperado

when you wife scream …and you just see her lips moving :-)

My wife screaming "George, decrease the volume” and me just seeing her lips moving? You are kidding me!
I would have taken my wet pants off and swear to her having fallen in my knees that this will never happen again.

Christophe, here, women RULE.


George

[Esperado]

Originally Posted by simon7000

Distortion is not a fixed number.

Ed, you did not answered my previous questions: Did -you think you will offer-us a comparison of the same resistance's distortions measurements, done with your selected parts, with an average metal film set, in order we can have a basis to analyze the "quality parts" effects ?
(I know, it is time consuming to pair them).
Right now, i'm looking to my box of thousands of blue 1/4 w resistances in a suspicious way with a guilty feeling.

Originally Posted by gpapag

Christophe, here, women RULE.

They do everywhere: i'm divorced :-)

[scott wurcer]

Originally Posted by simon7000

Where? I really am being concerned about you. .

ES

Where, at the very beginning post #32485, in response the a very gentle nudge to actually pay attention to what the circuit is doing you said…

C'mon Scott, you know the drill. They are looking at the difference across a 10 ohm resistor with a gain of 100. The noise from just the resistor should be (back of envelope calc's) 100pV/Hz at the input. Add this to the 2.7 nV/Hz and you get 2.702 nV/Hz. (For those who have trouble with this addition please do a bit of homework before commenting.) With a gain of 100 that is 270.2 nV/Hz of noise. Now .00003% of 5 volts is 1.5 uV. So (1.5 uV/270.2 nV)^2 is a measurement bandwidth of 31 Hz.

This is wrong the distortion is ~150uV at the output (sees the same gain as the noise) not 1.5uV. You have at the output a ratio of 150,000/270 (~55dB). You don’t need much more than 20dB spurs above the noise floor to get an almost exact read on the distortion components. I don’t think you need more than a 2 or 4k FFT to do this. Note you also show that the noise is dominated by the op-amps own input noise the resistors contribute virtually nothing.

Now they are measuring using an AP System 2, something I am familiar with. If I just go to the FFT analysis without stepping up the default resolution this level wouldn't even show up. Now at maximum rate the AP does go to -150ish re 1 V rms. But that takes 45 minutes or so. Without details on the settings of the AP one does not know if they were looking at noise or signal.

Since this circuit amplifies the THD by 40dB you don’t need much more than -100dB. Did you actually try this circuit? I could duplicate their results with a sound card and CoolEdit 2000.

When they first did their setup they used higher value resistors. (So please don't give me they know what they are doing line... everyone makes missteaks.) I did tease BP about it and never got a reply, but the next time I looked they were using much lower value resistors.

The text and accompanying figure on the TI data sheet have no mistakes of any note, they say what they are doing quite precisely. Problems though the CMRR effects are included with the distortion (CMRR is probably more linear and does not contribute over the op-amp especially under heavy load). The gain of the op-amp is falling with frequency, the GBW of the op-amp is high enough that this is probably not a factor.

In another thread I said I don't like THD+N graphs, but the audio industry wants to see them. We prefer THD vs. level/load/frequency separated into 2nd's and 3rd's.

[john curl]

Well, everybody is again carrying on with their opinions of what audio quality is.
For me, WHAT WORKS, WORKS!
I spent a good number of years, both designing and using analog tape recorders. Noise was a PROBLEM, not a fix. The LESS NOISE that we could obtain in an analog tape recording process, the better the overall sound, and we got up to 80dB/CCIR weighted.
The reason for this is that there IS useful and accurate information at levels 20-40dB below operating level. Tape recording theory predicts this, and no 'flim-flam' coverups like dithering are necessary to make an analog tape signal virtually perfect as it falls in level, except for the noise which is much like the general background, anyway, when done to the greatest fidelity. Dolby will lower the noise floor, BUT it will invariably foul up the sound quality. It is just another 'fix' or 'sweetener' , much like Aphex is, properly applied, or a 'noise gate'.
Why I bring this up here is because many here have limited or poor experience with magnetic tape, while using skinny tracks (more channels), low speed (cost savings), and poor quality tape (cheaper). In these conditions, 'good enough' was OK, but not for hi end audio. Here we rely on direct disc vinyl, 30ips, or 18ips 35mm film, full track analog master recording, or the VERY HIGHEST DIGITAL (96K or better) as our references. Then it is like putting our designs on a test track and noting the performance, rather than driving in the suburbs and noting that the time to the store and back was about the same as something cheaper and less finicky.
Hi end is all about making the equivalent of Porsches, or as close to their quality as we can, with what we have to work with, sort of like Hondas.
It is not like putting forth a cheap 'American' badged car and saying that it is good enough for 'anything' because it will seat 4 people and can be driven on the freeway when necessary.

[SY]

Flim-flam? You're clearly not understanding dither. One can easily hear signal well into the noise floor, just like analog tape- but with a lower noise floor.

Trying to think about modern digital technology with 1968 concepts will drive you to incorrect conclusions. In your words, what works, works, and that includes dither.

[Esperado]

John, prices of tapes had never been an issue, prices of studio hours was high enough. I preferred Ampex ones for their high dynamic.
Tape speed (15 or 30) was a matter of preference, between high bandwidth and low frequencies linearities (you know magnetic heads sizes resonnances...). On my side, I preferred the sound of 15"/s.

Multi tracks and 1/4" machines were mainly two brands, Studer and Ampex, all over the world. Both with advantages and inconveniences.
Standard was 2" 16 or 24 tracks (i preferred 2X16 synchronized machines, when available)
I agree with the Dolby/Dbx artefacts i hated and needing a lot of maintenance and tuning. But its success demonstrated the noise floor problem obviously.
Noise floor of the multi-tracks were not low enough, as noises of each tracks added themselves (+3db each X2) during mix down, always a painful concern.
Added with electric instrument amplifiers noises and hum, and noise of the studio itself. On some very acclaimed records, done with the best equipment available at this time, you can clearly notice the tape noise difference, before and after each tune. Not at all the dreamed 'high end' world you describe.
They were not "virtually perfect" at all and much too noisy, in real use, reason of the success of Dolby. But we tried our best with what we had, including noise gates or fade down of tracks between used parts on them.
Not to forget the losses of sound quality (Dynamic) on the recorded tracks during each passage in front of the heads, during so many re-recordings. I used, when possible, two multi track machines: the first one with drums, staying apart during re-recording, while a copy of the drums and basses on two tracks was used on the second machine just for monitoring during re-recordings.
Only noise floor of the mixing machine was, as you said, OK.

Digital arrival was just a miracle. and very soon, we do not had to worry about recording machines limitations any more: It 'WORKED'.
Dither is not adding listening noise to the signal, it is about getting MORE informations of very low levels signals near the first bits.

[Esperado]

Notice i did not said a word of 35mm film tapes, large tracks sizes, but *poor* middle aged machines.

[simon7000]

Originally Posted by scott wurcer

Where, at the very beginning post #32485, in response the a very gentle nudge to actually pay attention to what the circuit is doing you said…



This is wrong the distortion is ~150uV at the output (sees the same gain as the noise) not 1.5uV. You have at the output a ratio of 150,000/270 (~55dB). You don’t need much more than 20dB spurs above the noise floor to get an almost exact read on the distortion components. I don’t think you need more than a 2 or 4k FFT to do this. Note you also show that the noise is dominated by the op-amps own input noise the resistors contribute virtually nothing.



Since this circuit amplifies the THD by 40dB you don’t need much more than -100dB. Did you actually try this circuit? I could duplicate their results with a sound card and CoolEdit 2000.



The text and accompanying figure on the TI data sheet have no mistakes of any note, they say what they are doing quite precisely. Problems though the CMRR effects are included with the distortion (CMRR is probably more linear and does not contribute over the op-amp especially under heavy load). The gain of the op-amp is falling with frequency, the GBW of the op-amp is high enough that this is probably not a factor.

In another thread I said I don't like THD+N graphs, but the audio industry wants to see them. We prefer THD vs. level/load/frequency separated into 2nd's and 3rd's.

Scott you are mixing things up. If you see what you think is a mistake in arithmetic mention it. Don't go off on a tangent. I showed the limit of the AP just a while back. I have attached it again.

The issue was and is that using the AP on autopilot and presenting a "THD" graph when it is THD + Noise and then quoting the numbers in a "Headline" is wrong.

My point was and is that the noise dominates until just before clipping sets in.

The first point that is getting lost in the noise is that the passives can have more distortion than the active parts.

The second point I was trying to show is that if a chip such as this is used for a low level amplifier you can parallel them to reduce the noise. If the issue was distortion as the data sheet suggests paralleling them would not reduce that.

So the question for the interested is what is the actual distortion using such a chip as a low level preamp (say .01 volt input max) and how many would you parallel to get the noise equal to the actual distortion. (That is where I was going before you tried to derail things.)

I thought at first you were trying to be humorous, then when you continued to stay stuck at the noise gain circuit you worried me. I am not sure what to make of you now. (You still seem to be stuck at the noise gain circuit.)

[simon7000]

Originally Posted by Esperado

Ed, you did not answered my previous questions: Did -you think you will offer-us a comparison of the same resistance's distortions measurements, done with your selected parts, with an average metal film set, in order we can have a basis to analyze the "quality parts" effects ?

Already in this thread but way back or just buy a copy of Linear Audio #1