INA1651 input resistors

[thor.zmt]

I still don't know what are these requirements according to you or whatever deity deems "appropriate to the application".

Well, I didn't volunteer to do the work.

But still you seem to consider they are irrelevant, since they're looking for the keys under the lamppost.

Irrelevant past a certain point. I gave an example of what I consider "past this certain point".

Going further is pointless.

AFAIK, there is still a huge debate about the audibility of these effects.

Is there? I see a lot of graphs showing the limits in a range of publication. So I think it is actually empirically researched.

I'm not so sure. AFAIK, no major audio manufacturer has included the works of Zwicker and Fastl in their evaluation procedures.

And? There is a lot of ground covered by their work, I did not observe anything I was unaware of.

Golden Pinnae is an ironic reference to people who think they are gifted with superior audition. Quite common in the audiophool world. People who design stuff with their ears as only metrics.

You mean they are looking in the dark without a flashlight because that is where the keys are?

Thor

 

[abbey road d enfer]

Well, I didn't volunteer to do the work.

This is a discussion group. You profer unsubtstantiated opinions. It is my duty, as a moderator to ensure the subject is discussed in a contradictory manner. But you always escape, which may be a sign that you're not willing to back up your opinions with hard fact.

Is there? I see a lot of graphs showing the limits in a range of publication.

Such as? Do you mean the EN 61305standard?

You mean they are looking in the dark without a flashlight because that is where the keys are?

They're not even looking for keys, because they think they don't have a lock...

 

[thor.zmt]

This is a discussion group. You profer unsubtstantiated opinions.

Really, you seriously think that reducing THD way, way below any remote chance of audibility if the HD was the actual signal, never mind as HD of the signal with masking is something that needs doing and is a good thing?

It is my duty, as a moderator to ensure the subject is discussed in a contradictory manner.

Did you mean constructive manner?

But you always escape, which may be a sign that you're not willing to back up your opinions with hard fact.

What facts are needed? You mentioned Zwicker and Fastl, you should be well that their work rather supports my points.

Such as? Do you mean the EN 61305standard?

No, you mentioned Zwicker and Fastl, who have an excellent book with many such graphs. They seem to largely confirm what I found in separate papers on Psycho/Physio-acoustics previously.

They're not even looking for keys, because they think they don't have a lock...

You profer unsubtstantiated opinions.

Thor

 

[abbey road d enfer]

Really, you seriously think that reducing THD way, way below any remote chance of audibility if the HD was the actual signal, never mind as HD of the signal with masking is something that needs doing and is a good thing?

I didn't say that. Never. I say that THD must be evaluated, and that in all likeliness, the lower the better.

Did you mean constructive manner?

Contradictory as in CONTRADICTORY JUDGMENT. By this term is understood, in the state of Louisiana, a judgment which has been given after the parties have been heard, either in support of their claims or in their defence. A judgment is called contradictory to distinguish it from one which is rendered by default.

What facts are needed? You mentioned Zwicker and Fastl, you should be well that their work rather supports my points.

So, are you advocating using their works to propose standardized method of audio performance assessment? AFAIK nobody has done that, not even the AES.

You profer unsubtstantiated opinions.

So you deny the fact that there are some "designers" that assess their produts in a mostly subjective manner?

 

[ccaudle]

So what do you suggest to replace or complement the usuall trilogy (BW, S/N & THD) as objectve measurements?

I would suggest as a start to completely throw out the "T" in THD, there is no need to use a single value for something which is not only known for decades to be frequency sensitive, but which is now easy to measure using e.g. FFT based instruments.
Thor has used a couple of times so far the phrase "Distortion H2 dominated and monotonic" which is a nice goal, but of course not all equipment behaves in the same way. 1% 2nd HD is quite a bit different than 1% 7th HD.

I still don't know what are these requirements according to you or whatever deity deems "appropriate to the application".

Perhaps I shouldn't jump into the middle of this, but this is starting to seem argumentative without giving Thor the benefit of the doubt here. I don't know that a web forum post is the place to rehash all the information in e.g. "An Introduction to the Psychology of Hearing" by Brian C.J. Moore or review all of the relevant literature on auditory masking which led to the development of the various audio coding standards (mp3, aac, vorbis, opus, etc.), but I don't think there should really be much argument with Thor's basic premise that there is a lot of settled science around audibility, and you can make some reasonable assumptions about maximum likely playback SPL and come to some conclusions about whether a particular effect will be audible or not.
I would argue for a quite wide guardband, there have been some reports of quite surprising audibility, but most of those fell into the category of "hmm...didn't realize the human brain could distinguish that" more than into the "contradicts previous experiments" category. See for example one of my favorite preprints "Are We Measuring the Right Things? Artefact Audibility Versus Measurement" by Paul Frindle.

 

[user 133392]

Mods: Do you think we could prune this thread from post #29 onward and move this to the Brewery?

I'm exhausted reading it and feel like I need to take a second shower today.
It's a perfect fit for the Brewery.

ccaudle: I would suggest as a start to completely throw out the "T" in THD, there is no need to use a single value for something which is not only known for decades to be frequency sensitive, but which is now easy to measure using e.g. FFT based instruments.

Thanks for the link.
While we're throwing out letters of the alphabet in specifying THD using modern FFTs can we throw out the "N" in THD+N?

 

[abbey road d enfer]

Mods: Do you think we could prune this thread from post #29 onward and move this to the Brewery?

I don't think it's a subject for the Brewery. Measurements and performance assessment is an appropriate subject.

I'm exhausted reading it and feel like I need to take a second shower today.

Just don't read it.

It's a perfect fit for the Brewery.

No it isn't.

While we're throwing out letters of the alphabet in specifying THD using modern FFTs can we throw out the "N" in THD+N?

That would be more appropriate than completely dismissing distortion measurements.

 

[ccaudle]

can we throw out the "N" in THD+N?

I'm for it. So instead of THD+N report distortion spectrum and noise spectrum?

 

[user 133392]

I'm for it. So instead of THD+N report distortion spectrum and noise spectrum?

Exactly.
Show the FFTs and state the measurement conditions so that the FFTs mean something.
On other fora I often see FFTs but have no clue as to level, load, gain etc.
They look real pretty though.

Going back to the OP I'd like to see how his idea works out.
I thought it was pretty clever use of the 1651's COM pin to raise Zcm.

With the COM and REF pins set to a 2.5V A/D-supplied Vcm, an added polarity inverter on the output biased to Vcm and a U-pad it might make a pretty decent A/D modulator driver like the THAT128X version we discussed elsewhere. balanced line input with THAT 1286

Abbey: Just tryin' to get the derailed locomotive back on the tracks...

 

[abbey road d enfer]

I'm for it. So instead of THD+N report distortion spectrum and noise spectrum?

I think we're back to square one. What is a good distortion spectrum? How do you express it to the guy who's given the task of designing the product?
You can't tell them "make the spectrum nice". Several stages of development later, you may tell them "oh the spectrum there could be better", but before that you have to give them simple goals.

I still think that a simple brief with attention to BW, distortion and S/N is a good start.

 

[user 133392]

thor wrote: I am sure if SiCr resistors were better if used outside a chip we would see them en masse.

There's an interesting article here about SiCr resistors which includes an HD3 distortion comparison of SiCr to through-hole. (And thick film FWIW.)
https://www.edn.com/design-notes-matched-resistor-networks-for-precision-amplifier-applications/
It includes this table:

The SiCr LT5400 array HD3 was 2 dB better than 1% through-hole in their FDA test circuit at 56 mW: At 1/14th power the TH was 2 dB better.

The reason why you don't see SiCr in discrete SMT or TH form is that the absolute value tolerance is very poor.
The "best" LT5400 is +/- 7.5% absolute tolerance - THAT's SiCr is typically +/-20% absolute.
Well-made SiCr have low HD, good initial matching, low tempco and can be laser trimmed to provide excellent matching.

The reader should not infer that when used on-chip SiCr resistors are inherently inferior to discrete NiCr or other alloys.
Each has their application.

 

[Bo Deadly]

We're not engineering spacecraft where every part has to operate within specific tolerances.

Pro-audio engineering is an evolutionary process. You try something different, take measurements and repeat. If distortion and noise is low, that's a feature. If not, that's a feature.

 

[thor.zmt]

Mods: Do you think we could prune this thread from post #29 onward and move this to the Brewery?

Agreed. It's completely off topic and pointless petty bickering.

I'm exhausted reading it and feel like I need to take a second shower today.

Image having to respond.

While we're throwing out letters of the alphabet in specifying THD using modern FFTs can we throw out the "N" in THD+N?

Can we replace by "HD" instead by "D-A" (Distortion audibility), which illustrates the expected audibility of the distortion and "D-D" (Distortion disagreeability)?

Then we may have something that is useful.

Definitely a quality product should score a clean zero on D-D and a product sold for "fidelity" (as opposed to say "good sound") should score a clean zero on D-A.

Thor

 

[thor.zmt]

There's an interesting article here about SiCr resistors which includes an HD3 distortion comparison of SiCr to through-hole. (And thick film FWIW.)
https://www.edn.com/design-notes-matched-resistor-networks-for-precision-amplifier-applications/

It seems to not include classic thin film resistors in TH format or thin film SMD nor MELF. Data for was published by Ed Simmons in Linear Audio. Highly recommended reading.

Thor

 

[thor.zmt]

Thor has used a couple of times so far the phrase "Distortion H2 dominated and monotonic" which is a nice goal, but of course not all equipment behaves in the same way. 1% 2nd HD is quite a bit different than 1% 7th HD.

Yes, I like to have things qualified. As numeric THD is basically not linked in any reliable way to audibility or objectionability of HD, UNLESS it applies to the kind of HD spectrum I specify) we need some other metric to quantify distortion. Of course, this not new. It goes back to Olson in the 1950's at least.

don't think there should really be much argument with Thor's basic premise that there is a lot of settled science around audibility, and you can make some reasonable assumptions about maximum likely playback SPL and come to some conclusions about whether a particular effect will be audible or not.

Thank you.

I would argue for a quite wide guardband, there have been some reports of quite surprising audibility,

I have no problem with that. I usually take a minimum 20dB Guardband (FOR ALL HARMONICS), unless I am not trying to make gear with inaudible distortion or decide to relax this requirement for practicability reasons.

See for example one of my favorite preprints "Are We Measuring the Right Things? Artefact Audibility Versus Measurement" by Paul Frindle.

Good one.

Thor

 

[Newmarket]

It seems to not include classic thin film resistors in TH format or thin film SMD nor MELF. Data for was published by Ed Simmons in Linear Audio. Highly recommended reading.

Thor

Yes. Just staying TH tells us not much about the resistor. Metal Film ?

 

[abbey road d enfer]

Yes. Just staying TH tells us not much about the resistor. Metal Film ?

I would venture carbon film for 5% and metal film for 1%...?

 

[thor.zmt]

With the COM and REF pins set to a 2.5V A/D-supplied Vcm, an added polarity inverter on the output biased to Vcm and a U-pad it might make a pretty decent A/D modulator driver like the THAT128X version we discussed elsewhere. balanced line input with THAT 1286

Then, again, I would propose a discrete option with a quad op-amp. Frontend Bootstraped inputs and Birt circuit on output.

Balanced in, balanced out, easy gain adjustment.

Or we can use a dual as buffer and an FDA as balanced ADC driver.

Thor

 

[sdvd]

Thank you all for the answers.

I played a bit with Bootstrap circuit with LtSpice for simulation.
I posted attached only one side of the balanced input I simulated just to show something I notice that I can't explain with certitude.

- R1/R2 provide a path for DC leackage of input capacitor and a DC path for the polarisation of the input of the OpAmp. They are bootstraped.
- C3/C4 provide common mode and additionnal differential filtering (They are doubled on the other leg not shown here). They are bootstraped through R7 so not totaly bootsraped.
- C7 provide DC isolation
- R3/R4 provide a discharge path for the input capacitor. They are bootstraped.
- R12/R14 provide filtering and current limiting for C7 if DC applied. (A bit of noise added)
- C11 provide differential mode only filtering. (Unique on balanced input)
- R11 provide differential input impedance for convenience...

I was able to see the high impedance provided by the bootstraped circuit through the different stages of the circuit.

On simulation I found that input impedance goes higher than 22kOhm (R11)...
How is it possible? Does that mean that the OpAmp is injecting more current to the input than it should be? Phase phenomenon? Is that a sign of distortion?

The phenomenon disappear if I disconnect C3/C4/R7, so I think it is comming from here but why?

Also, there is a bump on the very low frequencies on the Output magnitude chart. How could we explain that?

Third question, Are C1 and C9 needed to be audio capacitors or simple ceramic capacitors could be enough?

Thank you all.
David.

 

[abbey road d enfer]

On simulation I found that input impedance goes higher than 22kOhm (R11)...
How is it possible?

The circuit topology can be reduced (delta-Y transform) to a phase-shift circuit, which has very small positive gain. You can simulate that by isolating the opamp, driving the bootstrap loop and measuring at the open input.

Does that mean that the OpAmp is injecting more current to the input than it should be?

Yes

Phase phenomenon?

Yes

Is that a sign of distortion?

Why would it be? Of course there's a small positive feedback, but not enough to really change the distortion profile.

The phenomenon disappear if I disconnect C3/C4/R7, so I think it is comming from here

Of course it does.

but why?

See before

Also, there is a bump on the very low frequencies on the Output magnitude shart. How could we explain that?

The response of the phase-shift circuit is selective.

Third question, Are C1 and C9 needed to be audio capacitors or simple ceramic capacitors could be enough?

These capacitors are AC coupling. the cut-off is at about 0.2Hz. At 20Hz, the signal across it is about 40dB below the input signal, so distortion there should be minimal. If distortion is a real concern, you may want to use film capacitors and maybe of lower nominal value. They could probably be 10 times smaller with no degradation in performance, unless you intend to measure infrasonics.