DA is modelled purely as an RC network. If DA is a source of distortion, what is the precise mechanism for that?
Here's distortion measurements for a cheap carbon track 5k linear potentiometer at several attenuation settings, any distortion is below the noise floor or the limits of a QA503 to detect:
Will have a go at a log pot next
Will have a go at a log pot next
As I have already explained Sean Olive unfortunately misunderstood (or did not read) Frank Massa's article in Electronics 9/1938. THD as a single figure (like SINAD) is insufficient but THD presented as separate harmonics is ok. So FFT spectrum representation, which is included in most measurement based reviews, that shows low harmonics throughout the audio range would have been satisfactory according to Frank Massa.
This is where Sean Olive got it right:
http://seanolive.blogspot.com/2009/04/dishonesty-of-sighted-audio-product.html
1/f noise is yet another reason for avoiding electrolytic caps on DAC Vref.
A 100k log pot, PIHER 100KB716M (which I think is carbon) shows clear distortion (I couldn't find a lower resistance log pot alas, so the noise is annoyingly high), showing that log pots are typically poor performers compared to linear (as you might expect - in a linear track any voltage coefficient will cancel out as the track has constant electric field throughout its length).
The 0dB trace is effectively the floor of the QA403, the trace labelled Left should be -10dB. The distortion at 18dBV is a good 20dB up compared to a simple linear pot.
The take-away is use the Baxandall volume circuit and low impedance linear pots, easy and performant.
For completeness the 5k linear pot again but with 220uF 16V electrolytic on the wiper:
There is slightly raised noise on this, I think due to using a different breadboard this time and more mains pick up.
The 0dB trace is effectively the floor of the QA403, the trace labelled Left should be -10dB. The distortion at 18dBV is a good 20dB up compared to a simple linear pot.
The take-away is use the Baxandall volume circuit and low impedance linear pots, easy and performant.
For completeness the 5k linear pot again but with 220uF 16V electrolytic on the wiper:
There is slightly raised noise on this, I think due to using a different breadboard this time and more mains pick up.
Last edited:
Not to forget it already provides a signal-balanced output for free (when the gain is established at the buffer amp and the inverter runs unity-gain).
Which just means the blind A/B has not been set up and used properly. There must be no test stress.
It appears Hierfi is claiming something about individual electrons. An electron has a fixed charge, so it's not too hard to calculate the effect of one electron in a circuit. Here's the value to 10 significant figures:
https://en.wikipedia.org/wiki/Elementary_charge
Bob Pease had good 'luck' modeling DA as R-C networks and compensating for it in sample/hold circuitry.
https://www.edn.com/wp-content/uploads/2016/11/1982-10-13_EDN_Pease-capacitor-soakage.pdf
I don't like pots or stepped attenuators either, for the same reason. As a compromise I have been using a pair of 10 turn 10K/20K Ohm linear wire-wounds for years. The reason is simply that ever turn on the helical is the same, meaning that every loop is mostly the same resistance, capacitance and inductance, meaning absolute identical division. It needs to be set up for maximum volume at perhaps 2-3 turns, otherwise the already nuisance factor goes way up and your spinning the dial. I usually include a mute switch.
I have constructed amplifiers with bias currents below the femtoAmp region (10^-15) / second for doing relaxation time measurements for paint powders, and measured brain related tissue signals in rodents, which employed a Faraday room and required capacitance neutralizing (triaxial cabling) in order to provide the required bandwidth for the extremely high organic impedances involved. The point is simply that viewing signal from the single electron perspective is to begin to consider how single electrons accumulate on a much larger scale and the implications thereof.
DA and R-C networks have things in common, whereupon R-C can easily obscure DA related artifacts as likely Bob Pease realized. His work indicates that normal harmonic distortion testing doesn't necessarily apply, as an indicator of all mechanisms that DA can be responsible for sonically. Although he states that certain materials are far better or worse than others this does not state that any material is remotely innocuous.
The method to mitigate dielectric materials is not to use capacitors or create an infinite RC time constant, whereupon the potential across the material doesn't change. This is the reason for using DC servo networks. It isn't the RC that is the relevant issue its the DA that requires the mitigation.
Therein is the problem. RC implies unburied surface charge that when acted upon by real time fields acts responds in a real time manner. Trapped charges are time stamped in moving back out from the dielectric material at a rate related to the instantaneous field strength much like a delayed tape recording. Each electron is like a tape recording of a signal entered back in time when it crossed the barrier between the conducting stream and into the dielectric. In other words each electron is a time smearing element that in combination with perhaps millions of others creates artifacts of unknown nature.
A new twist on the OLD and WORN tale:
"I changed (capacitors/resistors/tubes/cable holders) and my Wife came running from the kitchen (3 rooms away 🙄) screaming: "WHAT did you do? It sounds SO MUCH BETTER now!" 😲
If at least Wives (and Kids) came with a calibration sheet as measurement microphones do.
U-wife or U-kid instead of U-mic? 😎
Same thing.
Maybe there is NO difference?
Prejudice usually tints results.
So focused listening shows no difference, but loose undedicated type does, specially while doing something else?
Yes, that makes sense.
In English please?
Any measurements or Physics backing that statement?
Books about Alchemy or Astrology are written in easier to understand Language.
Sorry... English is my second language... unfortunately it is also my first.
I am speaking of variance from pure capacitance/pure resistance from a mathematical perspective as pertaining to a pure RC network. DA reflects absorbing something, specifically electrons. These are not absorbed instantly, nor are they released instantly, leaving time related migration caused by electric fields or otherwise drifting around randomly. This can be looked upon from an individual electron perspective being thereafter backed off from there to the behaviour in a macro perspective. Fields from one conductor to the opposing side can be considered instantaneous acting upon both surface and buried electrons. There are two groups of electrons, the unburied surface electrons, being those predictive of RC and those of buried electrons either reentering the zero impedance electrodes or exiting electrons into the dielectric material. The difficulty is in the bulking of both being equivalent in lacking the memory of electrons resulting form DA having relevance. My question is to the correlation between artifacts created by DA not being bulked in and dismissed with conventional RC.
Quote
I am speaking of variance from pure capacitance/pure resistance from a mathematical perspective as pertaining to a pure RC network. DA reflects absorbing something, specifically electrons. These are not absorbed instantly, nor are they released instantly, leaving time related migration caused by electric fields or otherwise drifting around randomly. This can be looked upon from an individual electron perspective being thereafter backed off from there to the behaviour in a macro perspective. Fields from one conductor to the opposing side can be considered instantaneous acting upon both surface and buried electrons. There are two groups of electrons, the unburied surface electrons, being those predictive of RC and those of buried electrons either reentering the zero impedance electrodes or exiting electrons into the dielectric material. The difficulty is in the bulking of both being equivalent in lacking the memory of electrons resulting form DA having relevance. My question is to the correlation between artifacts created by DA not being bulked in and dismissed with conventional RC.
Quote
No Bonsai... seems simply intuitive to me...
I have had experience with ultra low level signals that a teflon interconnect cable can't be bent, otherwise the measurement becomes totally overwhelmed by tribo-electric mechanisms.
Just so you know; I will not be among those laughing. The effect of Vref bypassing is measurable.
Anyone with experience with RF circuits knows that capacitors have a self-resonant frequency and if you operate the capacitor past that frequency it doesn't act as a capacitor. Rather, it acts predominantly as an inductor. Similarly, it is common knowledge that by having multiple capacitors in parallel, say, 10 uF || 100 nF || 1 nF will result in a circuit that has three resonance points and relatively low impedance over a wider frequency range. So I'm not sure what you're trying to say with your blurb above.
Audio and RF will start to overlap once you enter the high-performance realm. Nothing new here either. And by "high-performance" I mean high measured performance. Life is fun past -120 dBc THD. 🙂 At least I think it is. And, no. That does not mean that THD is the only thing I care about.
High perceived performance has more to do with marketing, packaging, and pricing. Olive & Toole demonstrated this while at Harman-Kardon. Since you're clearly a fan of Sean Olive, how about digging out some of those papers? They were all published by AES so they're pretty affordable to access and if you're not willing to pay $125/year for an AES membership you can find many of the results in Olive's blog.
I find it laughable that you keep quoting that 1938 text as if science ended at that time.
Tom
The point Sean Olive was trying to make is basically the same point Earl Geddes has tried to make. If people want to ignore that and quote THD+N numbers, I don't think its so funny personally. Its more like sad that grown up engineers can't learn to stop quoting meaningless numbers.
If you're giving the levels of separate harmonics, then it's no longer the amalgam single-figure measurement known as THD (Total Harmonic Distortion), it's the levels of separate harmonics.
Yes, and how many here are reporting single THD or THD+N numbers? Unless you've noticed FFT spectrum displaying separate harmonics receives similar scoffing and even the same silly quote of Sean Olive.