INA1651 input resistors

[ccaudle]

There is one instance in which a super-high common mode impedance can be problematic with an active input.

That's when you have two devices at different chassis potentials: One mains grounded at or near 0V and the other floating at 60V (or 120V in 240V countries) because it has a 2 wire cord.

True, but that is easy to solve, just connect the shield at both ends. Or tie the chassis together some other way, like an additional conductor between chassis.

 

[thor.zmt]

In the absence of component values, it's hard assessing any kind of performance.

Well, if I wanted to publish them, I would have. Mind you, it's pretty obvious what to expect from this design.

Cascoded J-Fet into Folded Cascode loaded by bootstrapped input Mosfet follower.

Basically J-Fet Trasconductance directly into a near infinite load.

Single device transfer function plus added looped feedback of the J-Fet operated in the "triode" region. J-Fet operates at effectively minimal input voltage (circuit is inverting) and constant voltage.

Open loop gain 83dB, -3dB point for open loop gain 6.5kHz. Closed loop gain 0dB, so any HD will be reduced by the full loop gain.

Doesn't it depends on perfect matchiong of a few resistors? And maybe capacitors?

The limited open loop gain of the single stage active circuit is actually more limiting than resistor matching.

This also applies to Op-Amp based circuits at higher frequencies, at lower frequencies there often is a lot more open loop gain. On the other hand my circuit (despite being unity gain stable) offers almost 20dB more available negative feedback than common unity gain stable Op-Amp's.

Thor

 

[abbey road d enfer]

Well, if I wanted to publish them, I would have. Mind you, it's pretty obvious what to expect from this design.

Obvious for you, probably because you've spent some time about it.
Remember, this is group DIY, not the IEEE journal.
The circuit is somewhat obscured by the remnants of probable earlier experiments.
Like the two 1G resistors that only make sense in considerations of other simulations.

The limited open loop gain of the single stage active circuit is actually more limiting than resistor matching.

Don't R30+R4 vs. R17 define unity gain? And more significantly R6 vs. R17 define CMRR (assuming perfect gain matching from both stages)?

my circuit (despite being unity gain stable) offers almost 20dB more available negative feedback than common unity gain stable Op-Amp's.

I understand that. What are the practical consequences?

 

[Newmarket]

LT/AD have duals and quads at a cost of about $2 per resistor (1ku list price)
https://www.analog.com/en/parametricsearch/11449#/

Thanks. Those are more along the lines of the precision thin film resistor pairs in SOT23 or similar that I referred to.
Vishay MPM series ?- but long time ago so don't quote me !
Rather than "Standard" 0.01% tolerance resistors that I think Thor meant.
But anyway, I hadn't looked for similar with more resistive elements so thanks for putting those on my radar.
Some specs' on there that I'm not familiar with but only scanned a datasheet - "Side to Side Matching" etc.

 

[Newmarket]

Thor: "By not using laser trimmed on chip polysilicone resistors we get overall better resistors with lower noise and HD. It is all a tradeoff."
IIRC THAT use SiCr not polySi.

TI has an app note regarding their on-chip thin film.
On-Chip Thin Film Resistors Enable High-Performance Audio Circuitry https://www.ti.com/lit/an/sboa312/sboa312.pdf

wrt SiCr that's what I understood.
But I'm always wary of being definite on these things since I once insisted consistently that when a casual acquaintance was talking about having to change all BOM resistors to Tantalum that he was actually talking about capacitors. I was....WRONG (Squared)

 

[thor.zmt]

Thor: "By not using laser trimmed on chip polysilicone resistors we get overall better resistors with lower noise and HD. It is all a tradeoff."
IIRC THAT use SiCr not polySi.

Ok, but not NiCr. I am sure if SiCr resistors were better if used outside a chip we would see them en masse.

If buying a 5,000 pcs reel of thin film or MELF resistors ex. factory, 0.01% about doubles the cost over 0.05% and doubles again over 0.1% which are cheap enough that as a rule I specify 0.1% Thin Film for the audio path of even for extremely budget oriented products and MELF in anything with a slight extra budget for quality parts. Using Thick film or wider tolerance savings are so miniscule as not to be material.

Thor

 

[Newmarket]

If buying a 5,000 pcs reel of thin film or MELF resistors ex. factory, 0.01% about doubles the cost over 0.05% and doubles again over 0.1% which are cheap enough that as a rule I specify 0.1% Thin Film for the audio path of even for extremely budget oriented products and MELF in anything with a slight extra budget for quality parts. Using Thick film or wider tolerance savings are so miniscule as not to be material.

Thor

Yes - Thin Film or MELF.
Of course with MELF you have to factor in the "roll away!" losses
But given the DIY aspect of this forum I don't think many here are buying reels of 5000 !

 

[thor.zmt]

Obvious for you, probably because you've spent some time about it.
Remember, this is group DIY, not the IEEE journal.
The circuit is somewhat obscured by the remnants of probable earlier experiments.
Like the two 1G resistors that only make sense in considerations of other simulations.

Yes, shown is the common mode test. This should be obvious.

Don't R30+R4 vs. R17 define unity gain? And more significantly R6 vs. R17 define CMRR (assuming perfect gain matching from both stages)?

Any active circuit has an amount of error. This limits CMRR and makes CMRR follow the open loop gain.

Resistor Ratios only determine the CMRR in the first order approximation. The usual papers and equations tend to work on this first order approximation, which in practice is fine.

I understand that. What are the practical consequences?

If operated in the "triode region" a J-Fet is a pure square law device, while a BJT is closer to a cube law device.

By making the circuit to have a single device transfer function and operating the device with minimal signal, inherent distortion is made to be low and strictly 2nd harmonics.

By making a circuit with relatively greater feedback at high frequencies and less high order distortion, rise of distortion towards high frequencies is reduced and overall distortion can be low with a very simple circuit. And the parts used are "jelly beans" that are available off the page, instead of listing 104 Weeks lead time and zero stock.

Thor

 

[abbey road d enfer]

If operated in the "triode region" a J-Fet is a pure square law device, while a BJT is closer to a cube law device.

By making the circuit to have a single device transfer function and operating the device with minimal signal, inherent distortion is made to be low and strictly 2nd harmonics.

By making a circuit with relatively greater feedback at high frequencies and less high order distortion, rise of distortion towards high frequencies is reduced and overall distortion can be low with a very simple circuit.

Didn't you recently wrote distortion was " just someone looking for his lost keys under a streetlight because the light is better there, compared to where he lost them."

 

[thor.zmt]

Didn't you recently wrote distortion was " just someone looking for his lost keys under a streetlight because the light is better there, compared to where he lost them."

No, that is not what I wrote. Please make sure to include context and the precise text, it matters.

I am not against "low HD" and even well below audibility limits. I am less certain that more feedback is the best way to achieve it and that it is necessary to achieve extremely low subaudible distortion.

Thor

PS, here is what I wrote in a discussion on another board and in response to what:

Though agree, that it seems that there is a race for achieving the lowest THD and much of the medicine is fb (a bit like the 70's)

I do not consider "lowest THD" as a valid design goal.

There is no proven correlation between low (lower than needed to be inaudible) THD and perceived "good sound" or a perception of "reduced fidelity impairment" or a perception of "improved fidelity".

Thus it is pointless.

It reminds me of the old story of the drunk guy apparently looking for something under a streetlight. When the Cop asks him he tells the cop he is looking for his house keys.

The Cop helps looking, after a few minutes the cop ask: "Are you sure you lost the keys here?".

The drunk guy points somewhere into the darkness and says: "No. I lost them over there!".

The Cop asks: "Then why are you looking here?".

The drunk guy replies: "The light is better is here, easier to search!".

 

[abbey road d enfer]

No, that is not what I wrote. Please make sure to include context and the precise text, it matters.

My quote was taken directly from your post #69 in HT Transformers for Vacuum Tube Preamps

 

[abbey road d enfer]

I do not consider "lowest THD" as a valid design goal.

So, what is a "a valid design goal"?

There is no proven correlation between low (lower than needed to be inaudible) THD and perceived "good sound" or a perception of "reduced fidelity impairment" or a perception of "improved fidelity".

But you can't deny that there is a level of distortion that is excessive, just like too narrow BW. These things can be measured objectively.
What other measurable parameters are you suggesting as "valid"?
Of course THD is a very crude method, but no other performance markers that can be measured have ever succeeded in replacing the basic BW and THD measurements.
You can cite IMD or slew-rate or rise-time measurements, but nobody ever agreed on a consensus regarding their level of audibility.

 

[thor.zmt]

My quote was taken directly from your post #69 in HT Transformers for Vacuum Tube Preamps

Nope, it was not.

Didn't you recently wrote distortion was " just someone looking for his lost keys under a streetlight because the light is better there, compared to where he lost them."

Here is what I actually wrote:

"Low THD? Is that a valid design goal, knowing the THD of microphones and speakers? Or just someone looking for his lost keys under a streetlight because the light is better there, compared to where he lost them."

Thor

 

[thor.zmt]

So, what is a "a valid design goal"?

Fidelity impairments low enough to be reliably inaudible plus a decent safety margin.

But you can't deny that there is a level of distortion that is excessive, just like too narrow BW. These things can be measured objectively.
What other measurable parameters are you suggesting as "valid"?

I am suggesting that with all parameters there is a point where improving them no longer results in a reduction of perceived fidelity impairment, even if at much larger levels for such impairments reducing them causes an audible improvement of fidelity.

Of course THD is a very crude method, but no other performance markers that can be measured have ever succeeded in replacing the basic BW and THD measurements.

Ok, let me be facetious.

Given are:

My speaker (say) is:

30Hz-25kHz @ +/-3dB 3'rd octave smoothing,
87dB/0dBW/1m,
~ 0.1% (-60dB) @ 97dB/1m > 200Hz
< 1% (-40dB) @ 97dB/1m @ 50Hz

It will be hard to find such a speaker on the real world.

Amplifier 1,
1% THD (-40dB) @ +20dBW,
0.01% THD (-80dB) @ 0dBW,
Distortion H2 dominated and monotonic
5Hz - 80kHz @ -3dB
90dB SNR @ 0dBW

Amplifier 2,
0.0001% THD (-120dB) @ +20dBW,
0.0001% THD (-120dB) @ 0dBW,
0.1Hz - 10MHz @ -3dB
120dB SNR @ 0dBW

Which amplifier shows "appropriate design goals"?

You can cite IMD or slew-rate or rise-time measurements, but nobody ever agreed on a consensus regarding their level of audibility.

This is what my little quip about looking under the streetlight means. We design lower THD because it is (now) easy to measure (streetlight) not because we have reliable and significant proof that reducing THD reliable produces a reduction of perceived fidelity impairment.

Thor

 

[abbey road d enfer]

Fidelity impairments low enough to be reliably inaudible plus a decent safety margin.

How do you characterize "Fidelity impairments low enough..."
What are your metrics for fidelity?

I am suggesting that with all parameters there is a point where improving them no longer results in a reduction of perceived fidelity impairment, even if at much larger levels for such impairments reducing them causes an audible improvement of fidelity.

I agree; that's exactly what I meant when I mentioned "IMD or slew-rate or rise-time measurements".

Ok, let me be facetious.

Given are:

My speaker (say) is:

30Hz-25kHz @ +/-3dB 3'rd octave smoothing,
87dB/0dBW/1m,
~ 0.1% (-60dB) @ 97dB/1m > 200Hz
< 1% (-40dB) @ 97dB/1m @ 50Hz

It will be hard to find such a speaker on the real world.

Don't I know...

Amplifier 1,
1% THD (-40dB) @ +20dBW,
0.01% THD (-80dB) @ 0dBW,
Distortion H2 dominated and monotonic
5Hz - 80kHz @ -3dB
90dB SNR @ 0dBW

Amplifier 2,
0.0001% THD (-120dB) @ +20dBW,
0.0001% THD (-120dB) @ 0dBW,
0.1Hz - 10MHz @ -3dB
120dB SNR @ 0dBW

Which amplifier shows "appropriate design goals"?

Very hard to say, but for example cross-over distortion, which happens at low level, can be detected easily on speakers that have more THD locally.
As I wrote earlier, distortion from speakers is different enough to electronics distortion that a decently trained listener can make the distinction.

This is what my little quip about looking under the streetlight means. We design lower THD because it is (now) easy to measure (streetlight) not because we have reliable and significant proof that reducing THD reliable produces a reduction of perceived fidelity impairment.

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

 

[abbey road d enfer]

Nope, it was not.

You are actually reiterating...

 

[thor.zmt]

How do you characterize "Fidelity impairments low enough..."
What are your metrics for fidelity?

I do not propose any metric. Many exist.

I simply suggest that any fidelity impairments need to below audibility or we have a problem with fidelity. By allowing a bit extra we make sure that multiple devices ins series remain inaudible. I also simply suggest that going beyond "no audibility plus a bit extra in case" confers no further audible advantaged.

I am deliberately non-specific as this is a foundation of audio/acoustics issue. We can debate what these fidelity impairments are and what limits are once we acknowledge the basic proposition as true. Much literature and material on the subject is extant.

I agree; that's exactly what I meant when I mentioned "IMD or slew-rate or rise-time measurements".

Any required, naturally. You are expecting me to come out with something "this and that is what really matters", I will not.

I don't know. We do know that with music 12% HD with H2 dominant are audible and not objectionable for an average SPL in the 90's dB, for example. Also that 12% H3 are audible and objectionable but 3% are not objectionable but audible under identical conditions. On the other hand under the same conditions 0.1% H2 and 0.025% H3 are not audible.

So clearly Harmonic Distortion is a fidelity impairment, but it needs a lot of HD by modern standards to be audible and more to be objectionable.

Don't I know...

Yes, so clearly Speakers subject of very high levels of fidelity impairments in the traditional categories. Why does nobody demand -120dB SIAND (0.0001% THD&N) for speakers but routinely expects the same from a DAC?

Very hard to say,

I disagree. Very easy to say. At +20dBW (100W input) I want to see a 87dB/0dBW/1m speaker that has less than several percent HD at any frequency and single number percentages of HD below 50Hz.

Based on how HD works and adds, the speaker dominates HD, neither amplifier provides any audible contribution.

The same holds for all other "standard" parameters. Non of these amplifiers will cause audible fidelity impairments.

Obviously Amplifier 1 has "appropriate specifications" while Amplifier 2 is severely overengineered.

Likely Amplifier 2 will not only reduce the HD by 40dB over Amplifier 1 but also reduce your wallet content by 40dB more than Amplifier 1, while you do not need the improvements in HD, noise and frequency response it offers. That is what I mean with "valid design goal".

but for example cross-over distortion, which happens at low level, can be detected easily on speakers that have more THD locally.
As I wrote earlier, distortion from speakers is different enough to electronics distortion that a decently trained listener can make the distinction.

I mostly disagree. If we correctly evaluate distortion (THD & N is not it) then we can be clear what source of distortion dominates at a given harmonic.

However I think we can both agree that with a speaker having 87dB/0dBW/1m (0dBW= 2.83V) sensitivity (even a distortionless one) the following HD is inaudible:

1% THD (-40dB) @ +20dBW,
0.01% THD (-80dB) @ 0dBW,
Distortion H2 dominated and monotonic

0.0001% THD (-120dB) @ +20dBW,
0.0001% THD (-120dB) @ 0dBW,

It does not matter what is the cause of the HD, at these levels and with, in the first case the qualified distortion profile HD will be at audibility levels.

With a stereo pair of speakers the 0dBW SPL at 3m will be 84dB with a 2nd harmonic at 4dB ONLY in presence of a 84dB fundamental. That is guarateed to be masked.

With -120dB THD&N under the same condition any distortion would be at -36dB SPL, in other words 36db below hearing threshold.

All other parameters listed can be shown to be equally inaudible.

So for the purpose of listening to music using loudspeakers, both amplifiers can be - for the parameters discussed to be free from audible fidelity impairments, despite offering different levels of signal fidelity.

Do they still sound different? Good question, but that was not the question stated.

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

I suggest to use whatever audible fidelity impairment can be shown by suitable scientific evidence to be relevant, which may or may not include the "usual trilogy" and to include limits for audible fidelity impairment for each metric also backed by scientific evidence.

Otherwise we literally behave like our drunk looking for keys under street lights because that is where the light is, but not the keys, or like the practitioners of what Richard Feynman calls "Cargo Cult Science":

Cargo Cult Science

Thor

 

[abbey road d enfer]

I do not propose any metric. Many exist.

Correct me if I'm wrong, but a designer must have a set of measurable requirements to start work. AFAIK, that's how engineers work.

I simply suggest that any fidelity impairments need to below audibility

According to whom? The Golden Pinnae standard?

 

[thor.zmt]

Correct me if I'm wrong, but a designer must have a set of measurable requirements to start work. AFAIK, that's how engineers work.

Correct. And these requirements must be valid in the sense of appropriate to the application, or you end up with equipment that is over or under engineered.

According to whom?

There is extensive well documented scientific work on audibility of (for example) frequency response, harmonic distortion and noise. There is also such work on other subjects, e.g. Phasenoise (decorrelated Jitter), Jitter induced spuria (correlated jitter), phase-shift, impulse response etc. et al.
In other words, we have a pretty decent idea, backed by reliable science on what is audible. Nobody claims this body of work is complete, but it is qute substantial.

The Golden Pinnae standard?

Never heard of it. Could you provide a link to the standard documents? And why should one consider this unknown standard?

Thor

 

[abbey road d enfer]

Correct. And these requirements must be valid in the sense of appropriate to the application, or you end up with equipment that is over or under engineered.

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

There is extensive well documented scientific work on audibility of (for example) frequency response, harmonic distortion and noise.

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

There is also such work on other subjects, e.g. Phasenoise (decorrelated Jitter), Jitter induced spuria (correlated jitter), phase-shift, impulse response etc. et al.

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

In other words, we have a pretty decent idea, backed by reliable science on what is audible.

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

quote Golden Pinnae
Never heard of it. Could you provide a link to the standard documents? And why should one consider this unknown standard?

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.