UberCaps...
I agree with [DF96], namely, that far far too much effort went into modeling the remarkably poor performance Y4U dielectric ceramic disc cap. It seem clear (once you get used to the scale-changes in the work) that polyester, mylar, polyethylene, polypropylene, ... they're all quite acceptable.
It always amuses me to hear [LOL] people say how replacing some quite nice capacitors with "much upgraded and totally superior" ones ... just transformed their self-described piece-of-junk into the million dollar baby. That they can't understand how the "other manufacturer" can even be in the business. Etc.
I'm beginning to think that there's a whole line of Pro-Audio capacitor trade that quite simply employs the ultimate "fool 'em with the numbers" trick: State the values as -0.01%/+100%. The 0.01% tickles the irrational-precision knob. The +100% tickles the 'totally over-engineered' knob. The actual, factual effect would be generally increasing capacitance (lowering all those poles and zeros) which will actually affect low end nicely, if the replaced cap was designed too close to the 20Hz aural cutoff. Let's see ... a name ... Scarlet ÜberCäp brand. Hot [scarlet] hot [uber]. Make the things a jet-black with 1/3 (polarized end) in brilliant red. Gold plate the leads. Wrap them individually in tissue paper, and ship in little glass cigar tubes. Tubes in boxes, cellophane wrapped. Mylar hologram authenticity seals over the tops of the cigar-tubes. Individual testing sheets with each item.
Yes - these definitely would sell. Would need an awesome website though, with plenty of studious claptrap giving a strong (baseless, but authentic sounding) science-worded justification for everything, including the low-sonic-signature of both the black and red plastic used to enclose the capacitors. Both chosen to best (least) affect the sonic signature based on the polarity of action. Special editions ("unpolarized") that would have red on both ends, black band in the middle. Signature editions having teeny-tiny paper SKU tags on short strings, attached to the positive lead. Suggestion to leave attached in the amplifier. Lifetime offer to exchange any caps that fail in normal service.
I mean goats... if "Hemingway" can market and sell $5,500 power cables, $12,000 speaker cables and $8,000 interconnects... I tell you - there are PLENTY of fools just itching to be parted from their abundant money.
GoatGuy
I agree with [DF96], namely, that far far too much effort went into modeling the remarkably poor performance Y4U dielectric ceramic disc cap. It seem clear (once you get used to the scale-changes in the work) that polyester, mylar, polyethylene, polypropylene, ... they're all quite acceptable.
It always amuses me to hear [LOL] people say how replacing some quite nice capacitors with "much upgraded and totally superior" ones ... just transformed their self-described piece-of-junk into the million dollar baby. That they can't understand how the "other manufacturer" can even be in the business. Etc.
I'm beginning to think that there's a whole line of Pro-Audio capacitor trade that quite simply employs the ultimate "fool 'em with the numbers" trick: State the values as -0.01%/+100%. The 0.01% tickles the irrational-precision knob. The +100% tickles the 'totally over-engineered' knob. The actual, factual effect would be generally increasing capacitance (lowering all those poles and zeros) which will actually affect low end nicely, if the replaced cap was designed too close to the 20Hz aural cutoff. Let's see ... a name ... Scarlet ÜberCäp brand. Hot [scarlet] hot [uber]. Make the things a jet-black with 1/3 (polarized end) in brilliant red. Gold plate the leads. Wrap them individually in tissue paper, and ship in little glass cigar tubes. Tubes in boxes, cellophane wrapped. Mylar hologram authenticity seals over the tops of the cigar-tubes. Individual testing sheets with each item.
Yes - these definitely would sell. Would need an awesome website though, with plenty of studious claptrap giving a strong (baseless, but authentic sounding) science-worded justification for everything, including the low-sonic-signature of both the black and red plastic used to enclose the capacitors. Both chosen to best (least) affect the sonic signature based on the polarity of action. Special editions ("unpolarized") that would have red on both ends, black band in the middle. Signature editions having teeny-tiny paper SKU tags on short strings, attached to the positive lead. Suggestion to leave attached in the amplifier. Lifetime offer to exchange any caps that fail in normal service.
I mean goats... if "Hemingway" can market and sell $5,500 power cables, $12,000 speaker cables and $8,000 interconnects... I tell you - there are PLENTY of fools just itching to be parted from their abundant money.
GoatGuy
Great discussion, I'm really pleased to see all the objective, quantifiable results being brought into the discussion, though a some of it is over my head it is certainly helping my understanding.
However, I would like to add one simple caveat (not that I think anyone has lost sight of this), that sometimes the electrical defects of a part add an effect that sounds pleasing. For example, isn't that why some people prefer tubes to solid state, and why most of us are in this part of the forum.
I did some blind A-B testing of 4 different input coupling caps and a jumper (no cap) in a simple chip amp. We used 2 film caps and 2 bipolar electrolytics.
The speakers we used weren't great, nor was the source (pc sound card), but it was what I had at the time, all caps had been run on a dummy load prior to testing. There was at most a 30 second gap between tests, one of us would install the caps, the other would switch between them (so unfortunately not truly double blind), we simply said wrote down whether we like a or b more, or whether there was no difference. Each combination was tried multiple times, and 3 different source materials were used for each combination. The results myself and one other tester came up with were as follows: polyester cap ended up being the reliably least liked; polypro cap, jumper, and 1 of the electrolytics were basically indistinguishable, with a slight, but easily random, preference for jumper and polypro over electrolytic; other electrolytic was the most liked. After combining the listening results with the record of which cap had been which, we then repeated with the most liked and jumper to try and describe what was different. We came to the conclusion the electrolytic cap was attenuating the high end slightly, as well as some other difference neither of use could adequately describe. Again when doing this test we both knew we were listening to either an electrolytic or jumper, but neither of us knew which at the moment
Basically the cap was correcting a perceived defect in either the source, amp, speakers, or room. Likely it was the speaker crossover, as they were not great speakers. I suspect the distortion discribed above might have been the ineffable something we both noticed.
I now have better speakers, so I may need to revisit the input coupling in that amp and try to polypro again. Additionally I used the same brand/type of bipolar caps in the woofer section of those better speakers, I may try switching them out with the more "neutral" BPs, unless someone wants to buy me some 100uf film caps
Have any of the more audio appropriate linearity and distortion tests you guys are talking about been done with oil and paper? I had always assumed that PIO's created a distortion or some other effect, and that was why some people liked them, I was surprised to see that didn't appear to be the case (Science!).
A further question, I think this got discussed already but I can't find it in the 14 pages now. When designing an input coupling, is it better electrically (less effect) to build it around a smaller or larger input coupling value? Lets say .1uf vs 2uf for example, while attaining the same -3db point. Does the time constant have an impact? Or are properties of the specific cap, like ESR, more relevant?
Thanks, sorry for the basic question that may have already been answered!
However, I would like to add one simple caveat (not that I think anyone has lost sight of this), that sometimes the electrical defects of a part add an effect that sounds pleasing. For example, isn't that why some people prefer tubes to solid state, and why most of us are in this part of the forum.
I did some blind A-B testing of 4 different input coupling caps and a jumper (no cap) in a simple chip amp. We used 2 film caps and 2 bipolar electrolytics.
The speakers we used weren't great, nor was the source (pc sound card), but it was what I had at the time, all caps had been run on a dummy load prior to testing. There was at most a 30 second gap between tests, one of us would install the caps, the other would switch between them (so unfortunately not truly double blind), we simply said wrote down whether we like a or b more, or whether there was no difference. Each combination was tried multiple times, and 3 different source materials were used for each combination. The results myself and one other tester came up with were as follows: polyester cap ended up being the reliably least liked; polypro cap, jumper, and 1 of the electrolytics were basically indistinguishable, with a slight, but easily random, preference for jumper and polypro over electrolytic; other electrolytic was the most liked. After combining the listening results with the record of which cap had been which, we then repeated with the most liked and jumper to try and describe what was different. We came to the conclusion the electrolytic cap was attenuating the high end slightly, as well as some other difference neither of use could adequately describe. Again when doing this test we both knew we were listening to either an electrolytic or jumper, but neither of us knew which at the moment
Basically the cap was correcting a perceived defect in either the source, amp, speakers, or room. Likely it was the speaker crossover, as they were not great speakers. I suspect the distortion discribed above might have been the ineffable something we both noticed.
I now have better speakers, so I may need to revisit the input coupling in that amp and try to polypro again. Additionally I used the same brand/type of bipolar caps in the woofer section of those better speakers, I may try switching them out with the more "neutral" BPs, unless someone wants to buy me some 100uf film caps
Have any of the more audio appropriate linearity and distortion tests you guys are talking about been done with oil and paper? I had always assumed that PIO's created a distortion or some other effect, and that was why some people liked them, I was surprised to see that didn't appear to be the case (Science!).
A further question, I think this got discussed already but I can't find it in the 14 pages now. When designing an input coupling, is it better electrically (less effect) to build it around a smaller or larger input coupling value? Lets say .1uf vs 2uf for example, while attaining the same -3db point. Does the time constant have an impact? Or are properties of the specific cap, like ESR, more relevant?
Thanks, sorry for the basic question that may have already been answered!
Last edited:
Just as an afterword... it strikes me that the people who somehow get convinced to have ubercables ... don't consider the whole system: plugging a $5,000 power cable with rhodium plated silver contacts and grainless silver wire ... into one's wall socket (which has a pretty-bog-standard bronze set of contacts, and maybe 12ga wire leading to the nest-of-snakes junction box...) is like putting a silk top-hat on a pig. Even looking inside the various amplifying/routing devices, looking right AT the socket-to-PCboard wires, are any claimed to be grainless silver? Is that same level of $8,000 1.5 meter long interconnect being done INSIDE the pair of connected devices? Hardly. Tutu on an elephant. ... in the living room ...
Arrgghhhhhhh!!!!!
GoatGuy
Arrgghhhhhhh!!!!!
GoatGuy
At least the guy had the balls to say that the "bridge residual measurement" wasn't necessarily the way to measure nonlinearity. And, moreover, that the "air capacitor" (which by definition shouldn't have any easily measured residual) was gummed up by 60Hz pickup, but, it too looked kind of like the others at the "good end".
I remain pretty much convinced there is a market for Scarlet UberCaps.
GoatGuy:
I've long suspected that a big part of the "I changed the Xs to brand Y and now this is amazing!!!" effect is (aside from expectation bias) the result of the parts tolerance. +/- 20% is a common value for capacitors, and from my own experience and testing other people have done it is often highly optimistic. I've seen +/- 50% variations in caps I've ordered that claimed to be 10%. Although this is just using the capacitance setting on my DMM, and may not be the most accurate, it seems reliable with most of the parts i've tested. For reference I've found Nichicon, Panasonic, and United/Nippon Chemicon to be pretty good for values, Kemet, Sprague, Elna, and Epcos to be a little less discerning in their QC for tolerance. Resistors I find are much more likely to be close to their stated value, I don't think I've actually tested a 10% R that was more than 5% out. Maybe I'm just lucky with resistors and unluck with caps though.
I personally test and match values both for the design spec, and between channels. Unless i'm using 1% or 2% resistors I generally order 1.5 to 2x the number of component required and hand pick the best. If i'm using tight tolerance resistors I usually order 1 extra, sometimes 2. Maybe I'm just anal over component values though. I generally order 1-2 extra caps unless they are expensive ($5 is an expensive cap to me) such as 500v electrolytic or large value film.
Many commercial audio products, even high end ones, use loose tolerance parts and it's highly unlikely they check each or even any component during construction. A beginner DIYer likely isn't going to check values or even have the equipment to do so.
When people go to upgrade their first DIY project or their commercial product I would imagine they are far more likely to double check values and match parts, far fewer are going to check the actual value of the part the are replacing as opposed to its indicated value. In other words most of the "upgrade" is actually coming from the circuit working as designed now, not unicorn blood dielectric caps wrapped in pages of the guttenberg bible, dipped in shellac, and all assembled by an old Japanese man that works only by candle light.
I've long suspected that a big part of the "I changed the Xs to brand Y and now this is amazing!!!" effect is (aside from expectation bias) the result of the parts tolerance. +/- 20% is a common value for capacitors, and from my own experience and testing other people have done it is often highly optimistic. I've seen +/- 50% variations in caps I've ordered that claimed to be 10%. Although this is just using the capacitance setting on my DMM, and may not be the most accurate, it seems reliable with most of the parts i've tested. For reference I've found Nichicon, Panasonic, and United/Nippon Chemicon to be pretty good for values, Kemet, Sprague, Elna, and Epcos to be a little less discerning in their QC for tolerance. Resistors I find are much more likely to be close to their stated value, I don't think I've actually tested a 10% R that was more than 5% out. Maybe I'm just lucky with resistors and unluck with caps though.
I personally test and match values both for the design spec, and between channels. Unless i'm using 1% or 2% resistors I generally order 1.5 to 2x the number of component required and hand pick the best. If i'm using tight tolerance resistors I usually order 1 extra, sometimes 2. Maybe I'm just anal over component values though. I generally order 1-2 extra caps unless they are expensive ($5 is an expensive cap to me) such as 500v electrolytic or large value film.
Many commercial audio products, even high end ones, use loose tolerance parts and it's highly unlikely they check each or even any component during construction. A beginner DIYer likely isn't going to check values or even have the equipment to do so.
When people go to upgrade their first DIY project or their commercial product I would imagine they are far more likely to double check values and match parts, far fewer are going to check the actual value of the part the are replacing as opposed to its indicated value. In other words most of the "upgrade" is actually coming from the circuit working as designed now, not unicorn blood dielectric caps wrapped in pages of the guttenberg bible, dipped in shellac, and all assembled by an old Japanese man that works only by candle light.
You, SY, and RDH4 (p. 925) all agree that, for a grounded grid stage, Zin = (rp + RL) / (mu + 1). (RDH4 gives it as an approximation, but likely because it's discussed in an RF section and is ignoring inductance, transit time effects and such). RDH4 then gives "loading across the output circuit" as approximately equal to rp and voltage gain as RL (mu + 1) / (rp + RL). Here gain and input Z are self-consistent and consistent with equal currents in input and output. In this model, as anode loading approaches constant current, input Z continually increases.
But I cannot reconcile this with the current generator equivalent circuit, where voltage gain = gm (rp || RL) + 1 and Zin = (rp || RL) / gm (rp || RL) + 1. This is also consistent with equal currents in input and output, but as anode loading approaches constant current, input Z approaches rp / (mu + 1), roughly 1 / gm.
Both cannot be right, and I'm still leaning toward the gm model. The mu model "feels" to me like it has incorrectly conflated the series intrinsic anode resistance.
Thanks, as always,
Chris
The derivation of grounded grid input Z for the gm model is:
Because input and output currents must be equal (grid doesn't conduct) voltage gain Eout / Ein must equal Zout / Zin, and equals gm (rp || RL) + 1. Zout is simply rp || RL. Substitution gives Zin = Zout / voltage gain = (rp || RL) / gm (rp || RL) + 1.
Thanks,
Chris
Because input and output currents must be equal (grid doesn't conduct) voltage gain Eout / Ein must equal Zout / Zin, and equals gm (rp || RL) + 1. Zout is simply rp || RL. Substitution gives Zin = Zout / voltage gain = (rp || RL) / gm (rp || RL) + 1.
Thanks,
Chris
@Chris: Yes, it goes in reverse of what happens looking in to the cathode, that is, the effective plate resistance becomes rp + (mu + 1)rk, where rk is the source resistance of whatever is driving the cathode.
Quote:
as anode loading approaches constant current, input Z approaches rp / (mu + 1), roughly 1 / gm
For the gm model, as anode loading approaches constant current, (rp || RL) approaches rp, so Zin = (rp || RL) / gm (rp || RL) + 1 approaches (rp) / (gm)(rp) + 1, or rp / (mu + 1).
Thanks,
Chris
as anode loading approaches constant current, input Z approaches rp / (mu + 1), roughly 1 / gm
For the gm model, as anode loading approaches constant current, (rp || RL) approaches rp, so Zin = (rp || RL) / gm (rp || RL) + 1 approaches (rp) / (gm)(rp) + 1, or rp / (mu + 1).
Thanks,
Chris
Quote:
Originally Posted by DF96
Be careful: this is only true for a zero source impedance.
Yes, it goes in reverse of what happens looking in to the cathode, that is, the effective plate resistance becomes rp + (mu + 1)rk, where rk is the source resistance of whatever is driving the cathode.
One of us is double-dipping, or, as I used to say in the Army, everybody's out of step except me!
I'll keep working on it. Thanks, as always,
Chris
This is true only if the intrinsic anode resistance is in series with the output, but not if it's in parallel. In the gm model, parallel applies, and a constant current load is external to the gm device.
I do think we're nibbling away at something. Thanks,
Chris
In general it is the voltage across the cap that is proportional to the distortion, and that will be the same in either case if you have the same -3dB point, whatever the combination of R and C (assuming caps of equal 'quality').
Just for fun..
Picking Capacitors - Walter G. Jung and Richard Marsh
I guess it was all imagination..
http://web.archive.org/web/20030807...ors.com/picking_capacitors/pickcap.htm#signal
Walt Jung..
http://waltjung.org/PDFs/Picking_Capacitors_1.pdf
Regards
M. Gregg
Picking Capacitors - Walter G. Jung and Richard Marsh
I guess it was all imagination..
http://web.archive.org/web/20030807...ors.com/picking_capacitors/pickcap.htm#signal
Walt Jung..
http://waltjung.org/PDFs/Picking_Capacitors_1.pdf
Regards
M. Gregg
Forget the model. If current can't change then impedance is infinite. Put a perfect CCS in the anode circuit and the cathode current can't change, whatever cathode voltage you set. This is infinite impedance.Chris Hornbeck said:
First rule of engineering: if the maths confuses, just think about the physics instead.
And let's not forget too quickly ... CCS removes the need for a cathode resistor bypass capacitor. Quite a subtle effect that's forgotten in the food fight. The Rk becomes a CVS (since the current flow is "constant"). So, since it is not tasked with having to have a lower Zk < Rk over the design audio band, it therefore needs no Ck to bypass the resistor's Rk. One thing tho' is for sure: such triode amplification sections certainly don't sound the same as the more conventional C bypassed Rk resistor! That constant current sends poor ol' triode deep into negative plate feedback voltage amplification.
BTW - not only can there be no such thing as a perfect CCS, in practice, CCS's have to have a very tight impedance variation depending on current. The curve ain't square. Because of that(!), the CCS can deliver the different voltages that track the Vin source's wiggles. Hence "amplification". Having such a critical function though, and being so near 'infinite impedance/gain', every last quiver of non-linearity and non-uniform voltage tracking turns up on the output signal. For this reason, I've long kept the CCS in the bag of tricks to "talk about", but not implement. It can be quite unforgiving.
GoatGuy
BTW - not only can there be no such thing as a perfect CCS, in practice, CCS's have to have a very tight impedance variation depending on current. The curve ain't square. Because of that(!), the CCS can deliver the different voltages that track the Vin source's wiggles. Hence "amplification". Having such a critical function though, and being so near 'infinite impedance/gain', every last quiver of non-linearity and non-uniform voltage tracking turns up on the output signal. For this reason, I've long kept the CCS in the bag of tricks to "talk about", but not implement. It can be quite unforgiving.
GoatGuy
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.