In the test I've described there are sockets for a quick swap and two couples of elcos: one brand new and one 'burned-in'.
The brand new couple IS the control.
If you swap the couples back and forth and differences are consistent the test is valid, IMHO.
Well, it's true, I'm off topic but no more than people arguing about poor people that fools themselves with burn-in.
This is my last post in this thread about my test.
Could you please define a controlled test and tell me, exactly, where my test faults?
Really, I want to understand.
Thanks in advance
A controlled test is one where the only stimulus is the one being tested for. In the case of auditory experiments, it is important that the test not be biased by preconceptions, unconscious bias, non-aural cues, or experimenter expectation. In your proposed "test," you are perfectly aware of what capacitor is in place when you are listening, so auditory-only stimulus has just flown out the window. Given some other things you said, you also should read up on the basic philosophy of experimentation (the ubiquitous Karl Popper).
It's not trivial to do controlled tests, and the results are often uncomfortable for the person who invests emotions into his beliefs.
It's not trivial to do controlled tests, and the results are often uncomfortable for the person who invests emotions into his beliefs.
Hi Stuart,
Making sure that there is only one variable that may change can be exceedingly difficult. This is where most "controlled testing" fails, all the possible variables have not been accounted for. These types of tests require strict adherence to the testing procedure (you did put it on paper, didn't you?). Running tests on "controls" helps to verify that the test is valid as well. If you can't verify your setup, any data you end up with is very questionable. Running the test many times over some time period with your control also helps you define the limits of accuracy, and will tend to show if there is anything else at work. Here is where the "number crunchers" come in handy. Once you plot the trends with your test numbers, along with temperature, humidity and atmospheric pressure and time of day (with the date also), you tend to get a better understanding of what things may impact the test.
Anything else is a casual test without controls. The cutoff for defining controlled or uncontrolled is the existence of just one extra uncontrolled variable, making the test uncontrolled.
People who do run tests learn how to remove themselves from the procedure. The very worst thing you can do is go into a test series ('cause it's not going to be just one test!) is to prefer one outcome over another. As soon as you favor one result, you will subconsciously try to affect the outcome...
-Chris
The most fundamental requirement for any controlled test. Well said Sir.
Making sure that there is only one variable that may change can be exceedingly difficult. This is where most "controlled testing" fails, all the possible variables have not been accounted for. These types of tests require strict adherence to the testing procedure (you did put it on paper, didn't you?). Running tests on "controls" helps to verify that the test is valid as well. If you can't verify your setup, any data you end up with is very questionable. Running the test many times over some time period with your control also helps you define the limits of accuracy, and will tend to show if there is anything else at work. Here is where the "number crunchers" come in handy. Once you plot the trends with your test numbers, along with temperature, humidity and atmospheric pressure and time of day (with the date also), you tend to get a better understanding of what things may impact the test.
Anything else is a casual test without controls. The cutoff for defining controlled or uncontrolled is the existence of just one extra uncontrolled variable, making the test uncontrolled.
No better way to say that!
People who do run tests learn how to remove themselves from the procedure. The very worst thing you can do is go into a test series ('cause it's not going to be just one test!) is to prefer one outcome over another. As soon as you favor one result, you will subconsciously try to affect the outcome...
-Chris
Sorry to break my promise but I've some more questions about it.
If I've undertood correctly the main problem with my test is that I know what caps are in while I'm listening, right?
So for sure it is not a blind test but it could be easily transformed into one if someone else swap the caps, insn't it?
Some sort of control is still present thanks to the comparison with the 'brand new' couple, not enough to say it's a 'controlled test' but it could give some hints if further investigation is needed or not.
But not everytime a controlled test is necessary, if I simply remove a coupling cap, so opening the circuit, there's no signal... the comparison with the cap in is quite obvious
Ah, when I've made my caps selection for my MyRef I've done the blind test with two different panels... and same results
If I've undertood correctly the main problem with my test is that I know what caps are in while I'm listening, right?
So for sure it is not a blind test but it could be easily transformed into one if someone else swap the caps, insn't it?
Some sort of control is still present thanks to the comparison with the 'brand new' couple, not enough to say it's a 'controlled test' but it could give some hints if further investigation is needed or not.
But not everytime a controlled test is necessary, if I simply remove a coupling cap, so opening the circuit, there's no signal... the comparison with the cap in is quite obvious
Ah, when I've made my caps selection for my MyRef I've done the blind test with two different panels... and same results
first problem is you KNOW the caps have been swapped. Need to get rid of that knowledge.
Next problem is you know the test is occurring and what the potential outcomes are. Need to get rid of that too.
Next problem is you know the base set-up. Need to get rid of that.
Next problem is we have no idea if this is just you, or a population that can detect (or not) the differences. Need more people (subjects).
Starting to see the picture?
Next problem is you know the test is occurring and what the potential outcomes are. Need to get rid of that too.
Next problem is you know the base set-up. Need to get rid of that.
Next problem is we have no idea if this is just you, or a population that can detect (or not) the differences. Need more people (subjects).
Starting to see the picture?
Hi Dario,
Every time you switch the capacitors, you change the temperature of the parts with your hands. Then there is compression on the case as you install them and a time lag in between. This is followed by a relaxing of the case, changing the dimensions of the capacitors themselves.
What you suggest doing has some value, but it will always be an informal test. There are more variables than you realize with this. Listening tests are always difficult, you really should be doing an ABX test.
Having said that, at least you are only changing one part position at a time. Be aware though, different voltages across the capacitor, different current levels and even positioning in the chassis will affect capacitors differently. So you may well find a capacitor that you like in that test situation, and yet another may easily out perform it once it's mounted inside the amplifier. It's enough to make you scream sometimes, isn't it?
Always try the "contestants" in the actual circuit you are considering the use they will be put into. You could use computer type jumpers and headers for the switching, leaving those capacitors untouched throughout the test.
-Chris
Every time you switch the capacitors, you change the temperature of the parts with your hands. Then there is compression on the case as you install them and a time lag in between. This is followed by a relaxing of the case, changing the dimensions of the capacitors themselves.
What you suggest doing has some value, but it will always be an informal test. There are more variables than you realize with this. Listening tests are always difficult, you really should be doing an ABX test.
Having said that, at least you are only changing one part position at a time. Be aware though, different voltages across the capacitor, different current levels and even positioning in the chassis will affect capacitors differently. So you may well find a capacitor that you like in that test situation, and yet another may easily out perform it once it's mounted inside the amplifier. It's enough to make you scream sometimes, isn't it?
Always try the "contestants" in the actual circuit you are considering the use they will be put into. You could use computer type jumpers and headers for the switching, leaving those capacitors untouched throughout the test.
-Chris
Hi Chris,
I agree completely.
I was proposing a simple test with a cap that to my ears sounds obviously different if new.
Now I know that it's not a 'controlled test' but I still think it is useful and could instill doubts about caps burn-in... if tried
It took me three months of tests to select the caps of my MyRef...
It is what I've done
Thanks
I agree completely.
I was proposing a simple test with a cap that to my ears sounds obviously different if new.
Now I know that it's not a 'controlled test' but I still think it is useful and could instill doubts about caps burn-in... if tried
It took me three months of tests to select the caps of my MyRef...
It is what I've done
Thanks
Are we talking burn in or wear out? If you are hearing a difference why do you think the cap is getting better. Is it becuase you like the way it sounds? To pick components by ear is fine if you are doing it for you, but please dont tell me they are "better" components on this basis alone. I just said this on another thread; Blind ABX testing tells us there really is a difference but not which is better, only which is more popular.
- Status
- Not open for further replies.