We've batted this around before, and it's time to make it happen: The Big SY RC Listening Test. Consider this the official call for help.
Here's the basics: We set up some passive high pass RC filters of moderate impedance (say, 10-20K ohm), with a nominal fc of 20 Hz. Two sets are made, one with universally acclaimed premium components, one with cheap parts like Radio Shack- magnetic leads are especially encouraged. The cheap circuits are trimmed to match the premium ones. The two sets of circuits are sealed up in a box with a premium rotary switch, say 22 position. Position 1 is the cheap parts, position 2 is the premium parts. The rest of the positions are randomized and a key sheet is depositied with a neutral party that we can all trust.
The box is sent to someone who firmly believes he (it's usually a "he") can hear the difference between RS parts with magnetic leads and crummy dielectrics (cheap mylar, for example). My restriction would be that this person is not in the business of making high-end amps featuring premium parts- I'm not making this perfectly cheat-proof and I don't want any temptation to someone whose income stream is at stake.
The test subject can hook this box up in his sytem, listen at his leisure to whatever he wants, switch however he wants, Freedom Hall. At the end of the test period (two weeks enough?), he is to identify for any 10 switch positions which is which. If the person scores 10/10, I'm willing to call that a positive result. 7/10 correct or less is a null result for that person. 8 or 9/10 and the person will be retested with the other 10 positions and a fresh test period. If the score of the second trial is 7/10 or above, I'll call that a positive result. (I may, before the beginning of the test, jigger these numbers up or down one or two to fit standard 95% confidence intervals, but this is approximately correct). The test subject can get all the help he wants from family or friends, as long as we're just talking about listening. No measurement! The test subject must agree to keep the score sheet confidential, so that we can repeat the test with other subjects.
So... here's where I need a collaborator or collaborators. First, to help me with the mechanical fabrication. I'm lousy at that. Second, I need a generous donor of premium parts, once there's some consensus as to what those parts should be. The parts can be eventually returned, if desired. I've already got the switch and I'll be a sport and donate the Radio Shack parts. Third, I'd like some consensus on whether buffering the filter would cloud the test results in people's minds.
Depending on the final mechanical layout, we might even be able to rerandomize between test subjects.
So, who wants to pitch in and help me out?
Here's the basics: We set up some passive high pass RC filters of moderate impedance (say, 10-20K ohm), with a nominal fc of 20 Hz. Two sets are made, one with universally acclaimed premium components, one with cheap parts like Radio Shack- magnetic leads are especially encouraged. The cheap circuits are trimmed to match the premium ones. The two sets of circuits are sealed up in a box with a premium rotary switch, say 22 position. Position 1 is the cheap parts, position 2 is the premium parts. The rest of the positions are randomized and a key sheet is depositied with a neutral party that we can all trust.
The box is sent to someone who firmly believes he (it's usually a "he") can hear the difference between RS parts with magnetic leads and crummy dielectrics (cheap mylar, for example). My restriction would be that this person is not in the business of making high-end amps featuring premium parts- I'm not making this perfectly cheat-proof and I don't want any temptation to someone whose income stream is at stake.
The test subject can hook this box up in his sytem, listen at his leisure to whatever he wants, switch however he wants, Freedom Hall. At the end of the test period (two weeks enough?), he is to identify for any 10 switch positions which is which. If the person scores 10/10, I'm willing to call that a positive result. 7/10 correct or less is a null result for that person. 8 or 9/10 and the person will be retested with the other 10 positions and a fresh test period. If the score of the second trial is 7/10 or above, I'll call that a positive result. (I may, before the beginning of the test, jigger these numbers up or down one or two to fit standard 95% confidence intervals, but this is approximately correct). The test subject can get all the help he wants from family or friends, as long as we're just talking about listening. No measurement! The test subject must agree to keep the score sheet confidential, so that we can repeat the test with other subjects.
So... here's where I need a collaborator or collaborators. First, to help me with the mechanical fabrication. I'm lousy at that. Second, I need a generous donor of premium parts, once there's some consensus as to what those parts should be. The parts can be eventually returned, if desired. I've already got the switch and I'll be a sport and donate the Radio Shack parts. Third, I'd like some consensus on whether buffering the filter would cloud the test results in people's minds.
Depending on the final mechanical layout, we might even be able to rerandomize between test subjects.
So, who wants to pitch in and help me out?
Do you have any components values in mind? I would like to check that setup before we do any work as it may happen that with this application the difference is very small.
It is obviously very simple setup, but I would prefer if the parts were actually doing more than just blocking anything below 20Hz.
For instance a resistor in a feedback loop of a gainclone. This shows very clearly any difference.
It is obviously very simple setup, but I would prefer if the parts were actually doing more than just blocking anything below 20Hz.
For instance a resistor in a feedback loop of a gainclone. This shows very clearly any difference.
You can do that, but it will be a little more complicated. My logic here is that if we run the signal through a "lousy" cap, cheap wire, and a cooking-grade resistor, the degradation should be audible, compared to high-price caps, resistors, and wires if the hypotheses of the designer-parts guys are correct. I'm thinking possibly 10-20K for the resistor and the cap value chosen to give a rolloff that won't affect the critical midbass, midrange, and treble.
If everyone is happy with switching resistors in a Gainclone, we could do it that way, assuming that the extra lead lengths (on the order of 3 or 4 inches won't (in your judgement) cover up any differences. The listening protocol is still the same.
If everyone is happy with switching resistors in a Gainclone, we could do it that way, assuming that the extra lead lengths (on the order of 3 or 4 inches won't (in your judgement) cover up any differences. The listening protocol is still the same.
Thanks, Rodd. You know where to find my address.
Analog_SA, absolutely! The switch I procured for this is what we call in Yiddish a "behayma." That might be close enough to Afrikaans to not need translation
I'll snap a photo and post it this evening (my time). It is at least of the quality of anything I've seen in high-end preamps.
Anyone willing to suggest or donate caps and resistors? (two each)
Analog_SA, absolutely! The switch I procured for this is what we call in Yiddish a "behayma." That might be close enough to Afrikaans to not need translation
I'll snap a photo and post it this evening (my time). It is at least of the quality of anything I've seen in high-end preamps.Anyone willing to suggest or donate caps and resistors? (two each)
Sy,
Interesting idea.
Some input from my experience though, a potential divider is a more sensitive test for a resistor comparison, so an attenuator would be a far better test.
The caps are harder though, electrolytics are unsuitable for this test due the the absence of any DC bias, so that limits us to film caps, which may be impractical.
Film caps with mag / non-mag leads of identical dielectric and similar construction could be a possibility (i.e. polyester, stacked film for both types), these can be had in values up to a few uF in sensible sizes. One needs to ensure there are no significant in-band changes though, with the range of impedance the device is likely to see.
We need to ensure that changes to phase reponse are kept outside of the audio band in order to keep the underlying system performance similar, so as not to mask effects.
Andy.
P.S. I suggest eliminating the switch too - better to have several duplicate circuits and swap connections.
Interesting idea.
Some input from my experience though, a potential divider is a more sensitive test for a resistor comparison, so an attenuator would be a far better test.
The caps are harder though, electrolytics are unsuitable for this test due the the absence of any DC bias, so that limits us to film caps, which may be impractical.
Film caps with mag / non-mag leads of identical dielectric and similar construction could be a possibility (i.e. polyester, stacked film for both types), these can be had in values up to a few uF in sensible sizes. One needs to ensure there are no significant in-band changes though, with the range of impedance the device is likely to see.
We need to ensure that changes to phase reponse are kept outside of the audio band in order to keep the underlying system performance similar, so as not to mask effects.
Andy.
P.S. I suggest eliminating the switch too - better to have several duplicate circuits and swap connections.
By the way, I think misgivings regarding a quality switch are unfounded. I don't give a rat's what your signal source and system are like, before that musical signal gets transduced by your speakers, it goes through a WHOLE lot of switches, plugs, and terminals.
But using a high-quality switch in any comparison test is essential.
But using a high-quality switch in any comparison test is essential.
How about adding a low pass filter as well? A cutoff at 30K would be reasonable. Additional parts should make the differences more evident.
Personally, I'd like to see the "bad" section with what seems to be agreed upon as the worst possible parts - an electrolytic and a wirewound sand resistor.
Personally, I'd like to see the "bad" section with what seems to be agreed upon as the worst possible parts - an electrolytic and a wirewound sand resistor.
Sy,
Shame on you!
It matters not one jot what it's been through before, the switch is a variable that potentially affects the test results. We're not testing switches
It will also mean different wiring paths between circuits.
Andy.
P.S. there's no need for exotic phono's etc to be used, some ineexpensive cheap parts are likely to be far better matched than most multiple pole switches, and a cheap plastic box would suffice. They can be potted solid in resin, for tamper-proofing.
Trimming an electrolytic (back to back, presumably) could be a nightmare. Normally, this would be done with low value caps, but then I'd be guilty of "improving" the electrolytics with a bypass. In theory, I could trim the "good" components to match the 'lytic, but then there would be many more of those nasty solder joints and leads, reducing that fine resistor and capacitor to the sonic equivalent of the Radio Shack parts.
Hey, if there's such a major improvement, clearly obvious to even old, deaf alcoholics (heh, heh) from going between various "good" capacitors, it should be a snap to hear the difference between a crummy steel-leaded mylar cap and something more wondrous.
Well, for that matter, between channels, you use different plugs and jacks.
Frankly, if the effect of crummy mylar caps, steel leads, crappy wiring, and sandcast resistors is below the level of coloration introduced by a single high quality set of switch contacts, that's a pretty useful thing to know, wouldn't you agree?
Frankly, if the effect of crummy mylar caps, steel leads, crappy wiring, and sandcast resistors is below the level of coloration introduced by a single high quality set of switch contacts, that's a pretty useful thing to know, wouldn't you agree?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.