What is the capacitor doing in that circuit? That tells you what is important. Very unlikely that ESR is relevant for a 2.2uF cap. So probably near enough identical performance, apart from electrolytic leakage. However, if the cap forms a significant part of the LF rolloff of the circuit then use the film cap; if not use the electrolytic - but consider raising its value by a factor of 10 to be on the safe side.dotneck35 said:
Like electrolytics, polyester is slightly nonlinear. Not too big a problem in most circuits - some guitar amps even choose them deliberately for their 'tone'. Unbiased polyester will give you a little 3rd order distortion; add a DC bias and you get 2nd instead. Even film caps are not perfect. Part of the design process is choosing the right tool for the job.
ESR is not really a 'quality' factor. It is a parasitic - just like stray capacitance of a resistor.
Last edited:
There is an equation in the datasheet, 42Hz cutoff is a tad high, the caps should be bigger.
Sorry guys but you are analyzing a "non-design"
That is an "Internet design" , probably cooked by an amateur, definitely is not and never was a Commercial design, nor ever put in production ... the telltale sign being that it requires +/-9V supplies.
99% of real world pedals are designed to run from a single 9V battery, because that is the de facto standard and only lately (very very few) are designed to be fed 9VAC or 9VAC, simply to be able to get symmetrical 9 to 15V supplies.
I very much anybody has ever built that "4QD pedal" , except maybe by some beginner who found in on the Net and got lured to it.
An old classic design would have used single +9V supply; a modern one would never ever have used 741 Op Amps, this one looks like cut and paste from an old design (think early 70`s, go figure) coupled to textbook approved symmetric supplies and some cookbook LM324 oscillator.
In that context the unterminated 10uF electrolytic is the least of the anomalies.
Only the earliest phaser, Jimi Hendrix approved UniVibe used CDS cells and a current hungry filamet bulb, which required it to be AC fed , no battery would last more than a couple hours there.
They did not even use Op Amps, it´s a late 60s design:
Of course, this is a commercial/professional design and output electrolytics are properly referenced to ground
Later Phasers (early 70´s) , introduced 741 Op Amps and, most important, FET modulation, which made them possible to be fed by regular 9V batteries, this is the Industry Standard MXR Phase 90:
Notice that all caps are properly terminated, unit is fed from 9V, and most important, unlike the "brewed up" 4QD schematic, it includes a switch to turn effect ON-OFF .
In this case, bypassing it; Univibe stops modulating, some means has to be included to achieve that in the rel world, effects are not used all the time on the penalty of becoming **boring** quickly
So in a nutshell, you were analyzing an amateur, incomplete design which was never commercially made
That is an "Internet design" , probably cooked by an amateur, definitely is not and never was a Commercial design, nor ever put in production ... the telltale sign being that it requires +/-9V supplies.
99% of real world pedals are designed to run from a single 9V battery, because that is the de facto standard and only lately (very very few) are designed to be fed 9VAC or 9VAC, simply to be able to get symmetrical 9 to 15V supplies.
I very much anybody has ever built that "4QD pedal" , except maybe by some beginner who found in on the Net and got lured to it.
An old classic design would have used single +9V supply; a modern one would never ever have used 741 Op Amps, this one looks like cut and paste from an old design (think early 70`s, go figure) coupled to textbook approved symmetric supplies and some cookbook LM324 oscillator.
In that context the unterminated 10uF electrolytic is the least of the anomalies.
Only the earliest phaser, Jimi Hendrix approved UniVibe used CDS cells and a current hungry filamet bulb, which required it to be AC fed , no battery would last more than a couple hours there.
They did not even use Op Amps, it´s a late 60s design:
Of course, this is a commercial/professional design and output electrolytics are properly referenced to ground
Later Phasers (early 70´s) , introduced 741 Op Amps and, most important, FET modulation, which made them possible to be fed by regular 9V batteries, this is the Industry Standard MXR Phase 90:
Notice that all caps are properly terminated, unit is fed from 9V, and most important, unlike the "brewed up" 4QD schematic, it includes a switch to turn effect ON-OFF .
In this case, bypassing it; Univibe stops modulating, some means has to be included to achieve that in the rel world, effects are not used all the time on the penalty of becoming **boring** quickly
So in a nutshell, you were analyzing an amateur, incomplete design which was never commercially made
Well, the earliest articles about capacitor sound (early 1980's) concentrated on dielectric absorption, later articles concentrated on capacitor non-linear distortion and your posts are the first I see that concentrate on ESR.
Regarding dielectric absorption, those articles showed that the step response of a first-order RC highpass (that is, DC blocking cap-load resistor) with dielectric absorption differs from the step response of an RC highpass with an ideal capacitor. They implied that this would damage the sound, but in fact, the difference in step response with and without dielectric absorption is much smaller than the difference in step response between an ideal RC highpass and just a plain wire with no filter at all.
This left me wondering: if the small difference between the highpass filters with and without dielectric absorption were big enough to damage the sound, why would capacitive coupling be acceptable at all? In fact, the output voltage of the RC highpass is the input voltage minus the voltage across the capacitor. Shouldn't we then concentrate on just minimizing the voltage across the capacitor (that is, use large RC products) rather than worrying about its similarity to the voltage drop across an ideal capacitor?
The articles about distortion showed that capacitors can distort, but that the distortion is generally so small that it is difficult to measure except on the very worst capacitor types, class-2 ceramic and tantalum. If it is the barely measurable distortion that matters, then why use high-performance capacitors in open-loop valve amplifiers with 1 % distortion?
Regarding ESR, I guess it makes no difference in a DC blocking capacitor as long as it is very much smaller than the load impedance. If you disagree, please explain why.
Of course we also have the subjective experience of many audiophiles. As they never do controlled double-blind tests, you never know if they actually hear a difference or hear their own bias.
Still, when a few euros extra don't matter, I would go for polypropylene or polystyrene capacitors. They have low dielectic absorption, low distortion, low ESR and good acceptance from audiophiles. I doubt whether this makes any audible difference, even when taking into account that there are usually many DC blocking capacitors in the total signal path, but it is better to be safe than sorry.
DA sometimes frightens people because it can change a waveform very slightly. As it can be modelled by a linear system (a CR ladder) it is most unlikely to create distortion. All that is actually happening is that it creates a very mild LF filter action: important for long term sample and hold application, quite unimportant for audio. However, there are claims that DA is correlated with poor sound so if true this can only mean that there is a correlation between dielectrics with poor DA and dielectrics with some other unspecified problem - unproven at this stage.
ESR matters to a small extent in PSU caps, and almost nowhere else.
Dielectric linearity matters in filter caps (including those coupling caps which set the LF rolloff) but not elsewhere, and almost all typical film caps are fairly linear anyway. Those concerned should avoid polyester, as it is about the least linear of the common cheap dielectrics. Curiously, some audiophiles prefer polyester!
As a general rule, people who worry a lot about caps (some even to the point of avoiding them) usually end up with worse systems because they are insufficiently concerned about more importan things.
ESR matters to a small extent in PSU caps, and almost nowhere else.
Dielectric linearity matters in filter caps (including those coupling caps which set the LF rolloff) but not elsewhere, and almost all typical film caps are fairly linear anyway. Those concerned should avoid polyester, as it is about the least linear of the common cheap dielectrics. Curiously, some audiophiles prefer polyester!
As a general rule, people who worry a lot about caps (some even to the point of avoiding them) usually end up with worse systems because they are insufficiently concerned about more importan things.
http://www.ti.com/lit/ug/slou250/slou250.pdf
TI uses very costly low impedance high quality electrolytic capacitors (typically designed for switching PSU IMHO), it don't seems to be a "cheap choice" and it don't seems to be a "audio dedicated" product.
They've probably benched a lot of capacitors especially for their peculiar application and find that these PSU caps were the best overall performers, perhaps.
Electrolytic, film, ceramic... perhaps that every technology has its champions !
when utilized in the right context
Last edited:
As TI don't say why they chose that particular electrolytic (in fact they say nothing at all about it) it would be speculation to form a conclusion. However, I have seen application notes (I forget which SS company) which were simply pushing audiophile buttons; whether this is a matter of marketing, or young engineers who don't know better is unclear. 470uF seems a rather low output coupling cap value for 4 ohm speakers: LF rolloff at 85Hz. Clearly hi-fi is not the intended application. Perhaps the aim is to prevent switching stuff which has got through the output filter from cooking the caps? Speculation, I accept.
Perhaps, did they also bench the 1.0µF 16V 10% X5R that is used as input capacitor? Sometimes we jump to the wrong conclusions.
I don't know...
http://www.mariobon.com/Glossario/___Componenti_Passivi/Condensatori_Tantalio.pdf
The amp is designed for a 8 ohm load and 24VCC PSU, this chip datasheet exhibit barely the same the performance measurements as others class D amps, i don't see any rational attack angle.
Is there any application where a BP electrolytic would actually be superior to a polyester film, assuming same size / lead length? Or is polyester just "average for a film", but still always the better choice of the two aside from cost?
why have you chosen an "8" in your formula?There is an equation in the datasheet, 42Hz cutoff is a tad high, the caps should be bigger.
That looks stupid.
Post 1 is clearly (at least to me) not intended to drive a speaker.
The input coupling capacitor is driving 390k and the output coupling capacitor is fed via a 390ohms resistor.
10k to 100k would be a better loading case to assess the suitability of the 10uF output coupling capacitor. Not 8 !
Last edited:
It says 4 ohms in the table on page 3. Even 8 ohms would still be 42Hz rolloff - not too good. I am unclear why you seek to defend a position by offering such poor examples from others who you believe share your position. They use a small cap for output, and a fairly non-linear cap for input.silverprout said:
The bipolar electrolytic is likely to be smaller so it is possible that in some applications it would have smaller stray capacitance to other nearby items. Apart from that the polyester would probably be better. One other issue: the polyester would have lower ESR and smaller HF losses which could mean that it doesn't damp an HF resonance so well. These two issues are why sometimes someone 'upgrading' a circuit by swapping in a film cap can run into problems.zaib4tsu said:
Ok, DF96 it's designed for 4ohm, with a 84Hz rolloff... witch solpe ?
seek to defend a position i'm really trying to find a misconception, but make a evaluation board tolerant to errors by decoupling the ouputs seems not an error to me since it's a lab board.
fairly non-linear... non linear of what ?
In post 27 you seemed to imply that as TI chose an expensive low ESR cap for an audio amp then they must know what they are doing so we should do likewise. Did I misunderstand you? It has been shown that this particular design is somewhat flawed, which suggests that it might be a bad idea to take inspiration from it.
The outputs are not "decoupled". They use a coupling capacitor which is too small for the job, unless restricted bandwidth was intended for some unstated reason. If restricted bandwidth is required, then the output coupling cap is about the worst possible place to do it.
The input seems to use a high-k ceramic cap for coupling, which is possibly a bad idea - unless they intend some input distortion on low frequencies.
The outputs are not "decoupled". They use a coupling capacitor which is too small for the job, unless restricted bandwidth was intended for some unstated reason. If restricted bandwidth is required, then the output coupling cap is about the worst possible place to do it.
The input seems to use a high-k ceramic cap for coupling, which is possibly a bad idea - unless they intend some input distortion on low frequencies.
Last edited:
Ok, i must admit that i've replaced the X5R input caps by two wima MKS2 and suppress the output caps on my design
Last edited by a moderator:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.