I believe insufficient excess gain could explain some sound differences - when low 6-10, maybe grudgingly, 20 dB is called "low feedback" - then there isn't enough loop gain for the closed loop gain to be determined to <0.1 dB by the feedback parts
so "low feedback" advocates may well be listening at different levels even with the same feedback parts
not to mention the added distortion, and "Harmonic Multiplication" that come with low feedback
so "low feedback" advocates may well be listening at different levels even with the same feedback parts
not to mention the added distortion, and "Harmonic Multiplication" that come with low feedback
The amps often have different gain, different output impedance (up to ohms for some valve amplifiers) - how would you set as same level as possible in another way than with sound level meter? 0.1% FB resistor matching is not helpful here.
On the other hand, I do not share this 0.1% obsession. Just trying to point it may be useless. But, the story that we used sound level meter for level matching is true.
On the other hand, I do not share this 0.1% obsession. Just trying to point it may be useless. But, the story that we used sound level meter for level matching is true.
If two different amps put the same voltage at the speaker terminals, the output will be the same. It's a LOT easier to get 0.1dB repeatability/reproducibility with an electrical signal than an acoustic one. Not easy, but easier.
If you want to make a valid listening comparison between two boxes of gain or two different ICs, you have to match the levels accurately, whether that's your "obsession" or not. Otherwise, your "testing" is just make-believe.
Level matching speakers is not possible with 0.1dB accuracy since even very good speakers have +/- 1.5dB non linearity at best, have different low end roll off, not to mention the distortions introduced by the measuring environment.
Averaging SPL in an A-weighting might be best, since that follows Fletcher Munson, but even that it is iffy. For example, take two speakers that are identical with the exception of the low roll off frequency, with one at -3dB at 60, and one at 40 Hz. Even if at these frequencies the impact on the average will be some 25dB less than at 1Khz, there is a lot of acoustic energy between 40 and 60Hz. In short, if these two speakers are SPL matched using the A-weighting, the one with the higher roll off will play considerably louder at all frequencies above 60 Hz than the speaker with the 40Hz roll off.
This will most likely lead to a general preference for the 60 Hz roll off speaker, except for source material with a focuss on the below 60Hz. And even then, the 40Hz speaker will sound muffled as compared to the 60Hz one.
Therefore, probably the best way to set levels is to high pass both speaker first so that the low end does not influence the measured SPL as used for level setting.
Averaging SPL in an A-weighting might be best, since that follows Fletcher Munson, but even that it is iffy. For example, take two speakers that are identical with the exception of the low roll off frequency, with one at -3dB at 60, and one at 40 Hz. Even if at these frequencies the impact on the average will be some 25dB less than at 1Khz, there is a lot of acoustic energy between 40 and 60Hz. In short, if these two speakers are SPL matched using the A-weighting, the one with the higher roll off will play considerably louder at all frequencies above 60 Hz than the speaker with the 40Hz roll off.
This will most likely lead to a general preference for the 60 Hz roll off speaker, except for source material with a focuss on the below 60Hz. And even then, the 40Hz speaker will sound muffled as compared to the 60Hz one.
Therefore, probably the best way to set levels is to high pass both speaker first so that the low end does not influence the measured SPL as used for level setting.
Please be more specific about what you think is nonsense so that we can have an intelligent discussion. Periodic signals readily demonstrate all kinds of things, but the connection with my earlier post escapes me.
Why surprised? the load end has non linear components. The caps and inductors have a frequency dependent resistance, the voice coils have frequency and velocity dependent eddy losses...
Why is it everybody just assumes that 1% resistors have a spread of 2%?
Many resistor types are either trimmed up to value, or simply binned by measurement. In either case, the distribution is NOT a normal gaussian distribution.
If the manu has a .5%, 1%, 2% product, then 2% resistors will NEVER have values under 1%, and 1% will never have a value under .5%.
Hasn't anybody done a measurement/distribution look-see at what they are buying?
It would be even better if you found a graph which used the correct equation for skin depth. That table is off by more than a factor of 3. Skinning is not as hard as the table says.
The earth's magnetic field is about .5 gauss. Once powered, you can simply rotate the probe in room air to get a full gauss spread.
Speaker magnets can run a gap field up to about 1.5 tesla, or 15,000 gauss. Little neo magnets can run up to a tesla or 2 easily at their surface.
jn
"If the manu has a .5%, 1%, 2% product, then 2% resistors will NEVER have values under 1%, and 1% will never have a value under .5%."
Absolutely correct, but having worked in a test room where semi parts were sorted, it did happen sometimes that not enough grade "C" parts were in stoc
k to meet an order, and grade "B" parts were marked up as C's to meet the shipment.
But, yes, in general, the lower grade parts' values had a normal distribution with a great wide hole in the middle!
Absolutely correct, but having worked in a test room where semi parts were sorted, it did happen sometimes that not enough grade "C" parts were in stoc
k to meet an order, and grade "B" parts were marked up as C's to meet the shipment.
But, yes, in general, the lower grade parts' values had a normal distribution with a great wide hole in the middle!
Yes, which is why I made that little chiding comment before about not being able to calculate without knowing the distribution. When you're doing large scale electronics production, it's vital to know these things, and one should understand the suppliers' quality systems intimately.
A simple two way speaker is designed with a chosen reference axis. A crossover is chosen for flat phase response across crossover region. Levels of woofer and tweeter are set to get flat frequency response across crossover region. At crossover point objective is having both equal phase and equal amplitude. This is readily checked by reversing polarity of one driver and analyzing the depth of the resultant notch:
An externally hosted image should be here but it was not working when we last tested it.
Here the notch is about -50dB. If it is taken that phase and timing errors are zero, then depth of notch is due to level mismatch of woofer and tweeter. Doing the math, I get level mismatch of 0.03dB.
With good manufacturing, driver to driver variation can be quite small. Differences are readily adjusted for with equalization of frequency response and gain.
Hi Barleywater, thanks for your explanation. We are speaking about two different things.
I was talking about level matching between two different loudspeakers, which can not be done at the .1 dB level. You are talking about level matching between two drivers in the same loudspeaker.
BTW, I have measured numerous speakers and never seen a -50dB notch in real measurements. Is your picture based on a sim, or is it a real speaker?
I was talking about level matching between two different loudspeakers, which can not be done at the .1 dB level. You are talking about level matching between two drivers in the same loudspeaker.
BTW, I have measured numerous speakers and never seen a -50dB notch in real measurements. Is your picture based on a sim, or is it a real speaker?
Perhaps I am being dense here but if you are only looking at two different power amplifiers to compare the sound why would you not use a single set of speakers? Switch from one to the other or two pairs of the same model of speakers at least. Using two even similar types of speakers you could never take that out of the equation, even a slight rise of less than 1db difference is always going to give a preference of one speaker over the other?
This is real measured result.
The same alignment process may be applied to any driver or set of drivers, thus two very well matched speakers may be constructed.
For the above demonstration the crossover frequency may be taken as 930Hz. If it is also taken that the driver levels are perfectly matched at 930Hz, then error is due to phase/timing error. In this case the error is about 0.5 microseconds, or phase angle difference of about 0.17 degrees.
Real result is combination of the two, plus contamination with reflected signals.
- Status
- Not open for further replies.