In Reply to: Call for comments on tubes vs. transistors posted by theaudiohobby on July 15, 2006 at 21:59:29:
First, I note the date of this non-peer reviewed article: 1980...
I begin by part one.
The tantalum capacitor test shown is irrealistic, and does not address the cause of this distortion, which is due to bad design, as shown below.
A tantalum capacitor (furthermore from the '70s: military types T101, T110 for example) is an electrolytic capacitor, which means the tantalum anode is surrounded by an electrolytic solution (which means ions can move in. When subjected to a negative voltage above tantalum vs cathode material's electrolytic voltage (about 2 volts), it frees H+ ions and builds up a tantalum sesquoxyde layer, which is an insulator, stops the electrochemical reaction, and then acts as the dielectrics layer in a capacitor.
Once built, it stays as it is, does not change its width, and so, the resulting capacitor is rather stable. But, if you apply an invert voltage in the range of the electrochemical potential, (2 volts, as I said above), the dielectric layer begins to react and to give back metallic tantalum. At the end of the process, the dielectric layer will soon disappear, and the capacitor will not be a capacitor anymore, but a resistor of a few ohms.
To dissolve back the sequioxyde layer to metal takes in the range of a second.
So, now, if you create across the capacitor a AC voltage at 35HZ (like done in the article)and around 2V, you're not likely to dissolve back the whole dielectrics layer on negative side of AC: too short. Furthermore, you heal the layer back on each positive wave.
So, the mechanism of distorsion is as follows:One should note that during phases (2) and (3), the relationship current/voltage is far from linear.
- On negative values of signal under the electrochemical potential, the component acts as a good cap with its specified capacitance.
- Above the electrochemical potential, since the dielectric layer is eroding, capacitance goes higher and sustained voltage goes down. This mechanism evolves in time, but soon...
- a positive voltage occurs, which can first rebuild the dielectric layer (so a time and voltage-varying current flows through the component), then
- the capacitor is rebuilt back to its pristine state.
So, the distortion shown in the article can be easily explained without resorting to dielectric absorbtion or soakage or even quantum dots (oh, not even yet invented at the time!).
So, why do I criticize since I admit these distorsions?
Because they can be easily avoided as long as you know how to design analog electronics.
- A DC bias which keeps the voltage across the cap always above the electrochemical potential is one answer (but you add the DC bias noise to the audio signal, so caveat).They show in the article some results which sounds correct.
- Putting 2 such caps can alleviate the problem, since at least negative and positive waves will behave the same way, so symetry will fade away even harmonics distortion components . With a real-world load value, it can be an acceptable solution. (comments, J Curl, since it is said it's your idea? Thanks)
- Where are those caps? On an audio input/output, it is just bad design: since you don't know which equipment your design will be connected too, it's obvious practice to make it able to accept common oddities of equipment, among those a strong DC bias. Elsewhere, you are able to modelize the voltage your cap will be subjected to, and then choose the right technology, tantalum or anything else that would fit better.
- Another answer goes deeper, and is a direct critic on the experimental conditions of the test.
He uses a 600 ohm resistor as load behind the cap. This is way unrealistic: no high-end audio equipment uses 600 ohms as loads since the '30s, but instead at least 20,000 ohms. So, look at his schematics: you have a voltage divider with the cap and the load resistance. With the value given, you get an impedance of 670 ohms for the cap at 35 Hz. So about 2.7 volts peak at the caps pins. Enough to get the nasty effects I talked about.
But if you use realistic values: 47uF for the cap, and 15,000ohms for the resistor, you get an impedance of 100ohms for the cap, and then a few dozens of millivolts only across it. With such a voltage, no electrochemical reaction can and will occur.
Then with those realistic values, the distorsion is gone. Not decreased, just gone... No more matter to write an article, sad isn't it?
Well, you could say "hey, this 600ohm value is still realistic on some inputs".
Don't forget that at 35 Hz with the proposed schematics, you're yet rolling off... furthermore, you will find those 600 ohm only on prewar proaudio equipment. But in this case, you can get up to 40 or so volts peak-peak, which would destroy the cap. So, nobody never did it this way!
So, this schematic is consumer-like for voltage level, and pro-like for load impedance. Wouldn't it be in fact just an ad-hoc demonstrator? That's what I think.
Otherwise stated, this article has no meaning above that of apprentice analog designers. Topic of the course: How Not To Connect An Electrochemical Capacitor"
In this post, I only addressed part I of the first document. I still have to flame part II. Large matter to feed the fire!
Then, I will have to address relevance of critics on the base/collector and gate/drain capacitance vs that of tubes. On Part III or IV, depends on my time.
I will quote Dawson (still not yet decided whether he has an acute sense of humor, or was really serious...): "However don't expect a tutorial unless you're willing to pay for it".
I think I'll be paid if the level of technicity of this forum increases above hunting quacks (furthermore from fossilized ducks)
This post is made possible by the generous support of people like you and our sponsors:
Follow Ups
- OK, let us analyse these archeological remains of "Golden Age". Part I - Jacques 09:59:01 07/16/06 (1)
- Re: OK, let us analyse these archeological remains of "Golden Age". Part I - morricab 02:15:16 07/17/06 (0)