Chasing 1% accuracy is not easy. The external effects from the input load R/C errors can change the accuracy significantly even with a low Z source. 100 pF can produce a 1% error in a 600 Ohm network at 20 KHz. Verifying the inverse network is also not as easy as it looks.
If you are trying for .1 dB you really need a system accuracy that is much higher (5X or more typically). Relative accuracy (in to out loss) is much easier than absolute accuracy since all you are looking for is insertion loss and frequency accuracy. Still for highest accuracy I would use a very stable oscillator, and a precision attenuator, preferably with a buffer amp on its output to isolate the load from the attenuator.
If you are trying for .1 dB you really need a system accuracy that is much higher (5X or more typically). Relative accuracy (in to out loss) is much easier than absolute accuracy since all you are looking for is insertion loss and frequency accuracy. Still for highest accuracy I would use a very stable oscillator, and a precision attenuator, preferably with a buffer amp on its output to isolate the load from the attenuator.
Hi,
As 1% cap's are not good enough on tolerance I build my "reference" iRIAA Network (Lipshitz) impedance scaled and using 10pcs 1% capacitors in parallel with 0.1% Resistors...
I did have the chance recently to test one on the AP, it was well within +/-0.1dB.
That is how I did my RIAA Measurements back then, in to out for each frequency, plus squarewave on the 'scope, squarewave on the 'scope is actually extremely telling...
Ciao T
As 1% cap's are not good enough on tolerance I build my "reference" iRIAA Network (Lipshitz) impedance scaled and using 10pcs 1% capacitors in parallel with 0.1% Resistors...
I did have the chance recently to test one on the AP, it was well within +/-0.1dB.
That is how I did my RIAA Measurements back then, in to out for each frequency, plus squarewave on the 'scope, squarewave on the 'scope is actually extremely telling...
Ciao T
I agree. My dad objected when I just turned the bass (woofers and subwoofer) down 0.1dB. He said it made the too lean and bright, and I hadn't told him what I did, or that I did anything. He complained and asked me if I had changed something.
I preferred the leaner sound at the time, but decided I agreed with him that a little more bass was preferable.
@ John @ John,
this is indeed an area where the usefulness of measurement is of vanishing relevance in loudspeaker development, that is once you have created the xover curves that measure as best as possible within the design constraints. Differences in energy provided to the high end that do not show up as significant deviations from an optimal flat FR, can still lead to very audible qualitative differences. Naturalness can be destroyed by differences that do no longer show up in measurements as significant.
Note: of course they show up in measurements as such. When you give more energy to the high end, that part of the curve will go up. The problem is that the high end is always wiggly, so where do you put the valley's, and where the peaks? Different settings may still be within the design bandwith, let's say +/- 1.5 dB, so where do you go to: the weapon of last resort, your ears.
vac
this is indeed an area where the usefulness of measurement is of vanishing relevance in loudspeaker development, that is once you have created the xover curves that measure as best as possible within the design constraints. Differences in energy provided to the high end that do not show up as significant deviations from an optimal flat FR, can still lead to very audible qualitative differences. Naturalness can be destroyed by differences that do no longer show up in measurements as significant.
Note: of course they show up in measurements as such. When you give more energy to the high end, that part of the curve will go up. The problem is that the high end is always wiggly, so where do you put the valley's, and where the peaks? Different settings may still be within the design bandwith, let's say +/- 1.5 dB, so where do you go to: the weapon of last resort, your ears.
vac
This is getting absurd. 0.1 dB on the bass, without a direct reference to compare to.
Let's get back to some sort of practical reality.
In the old days, (my youth) we did not necessarily have very, very accurate RIAA curves, even though, (thank goodness) the RIAA curve had been chosen as a reasonable compromise.
However, we did not have really accurate and extended data, easily on hand, to make sure that the RIAA that we designed was dead-on accurate.
This came in the 1970's, with Holman, and then the very, very accurate Jung-Lipshitz inverse RIAA box published in 'The Audio Amateur'. This is what I have used for decades with some success. This made it relatively easy, because you don't have to worry about 'scale errors' from typical analog equipment, and you don't have to convert dB to linear for digital readout. Saves a LOT of time and aggravation.
Now, today, with everybody and his brother owning AP analyzers with built in inverse RIAA, and with SPICE, you folks have it easy. Not like it was in my day, 'when we had to walk 5 mi to school in the snow, barefoot.'
'-)
Now what did this change do for me?
Well, same reviewer, same designer, AUDIO Magazine, two different phono stages
AUDIO, 1976 The JC-2, Bascom King, reviewer: Approximately +/- .5dB (or worse) across 20-20KHz. An almost embarrassing deviation.
To:
AUDIO, Dec. 1990, The Vendetta Research SCP-2B, Bascom King, Approx. 0.1dB, or better, having BK state: "This is the lowest RIAA equalization error I have ever seen" p.118.
This difference is having an accurate reference to compare to, and NOT leaving it to others to assure me that what they use is OK.
Now, why quibble here about making a reasonably accurate RIAA curve? It is easier to do and verify now, not like the old days when the Levinson JC-2 was designed.
To not do so, is just plain sloppy or for economy, and it is probably OK with the cheaper stuff to be off, even 1dB, because look at the associated cartridges, etc. that will be used with it.
Now, when you make a 'world class' phonostage, you should conform as closely as you can to the RIAA as to not let the ABX'ers of the world accuse you of 'cheating' when your product sounds better than a cheaper product, BECAUSE your RIAA is off by .5dB or so. AND they will do so, you bet! '-)
Let's get back to some sort of practical reality.
In the old days, (my youth) we did not necessarily have very, very accurate RIAA curves, even though, (thank goodness) the RIAA curve had been chosen as a reasonable compromise.
However, we did not have really accurate and extended data, easily on hand, to make sure that the RIAA that we designed was dead-on accurate.
This came in the 1970's, with Holman, and then the very, very accurate Jung-Lipshitz inverse RIAA box published in 'The Audio Amateur'. This is what I have used for decades with some success. This made it relatively easy, because you don't have to worry about 'scale errors' from typical analog equipment, and you don't have to convert dB to linear for digital readout. Saves a LOT of time and aggravation.
Now, today, with everybody and his brother owning AP analyzers with built in inverse RIAA, and with SPICE, you folks have it easy. Not like it was in my day, 'when we had to walk 5 mi to school in the snow, barefoot.'
'-)
Now what did this change do for me?
Well, same reviewer, same designer, AUDIO Magazine, two different phono stages
AUDIO, 1976 The JC-2, Bascom King, reviewer: Approximately +/- .5dB (or worse) across 20-20KHz. An almost embarrassing deviation.
To:
AUDIO, Dec. 1990, The Vendetta Research SCP-2B, Bascom King, Approx. 0.1dB, or better, having BK state: "This is the lowest RIAA equalization error I have ever seen" p.118.
This difference is having an accurate reference to compare to, and NOT leaving it to others to assure me that what they use is OK.
Now, why quibble here about making a reasonably accurate RIAA curve? It is easier to do and verify now, not like the old days when the Levinson JC-2 was designed.
To not do so, is just plain sloppy or for economy, and it is probably OK with the cheaper stuff to be off, even 1dB, because look at the associated cartridges, etc. that will be used with it.
Now, when you make a 'world class' phonostage, you should conform as closely as you can to the RIAA as to not let the ABX'ers of the world accuse you of 'cheating' when your product sounds better than a cheaper product, BECAUSE your RIAA is off by .5dB or so. AND they will do so, you bet! '-)
Last edited:
JC, did you suddenly change your feathers? I don't go around telling people to live in ignorance because they don't think differences in cables/resistors are audible.
This is what I agree with:
"It depends on how wide a bandwidth is covered, John."
"A difference of just 0.25dB covering the tweeter's passband therefore appeared to be significant.
John Atkinson
Editor, Stereophile"
absurd? The crossover frequency of the woofer to the mids is 300Hz. This difference was easily audible, and the same as tilting the frequency response up or down.
Someone designs a speaker with tilted up frequency response and they could show up on a message board saying, "Help, my brain is melting" - it happened.
A flat response and people say, it sounds outstanding with these 5 recordings and okay and somewhat bright with some other stuff and unlistenable to some stuff.
A tilted down frequency response and you hear that most stuff sounds good, except for some bad stuff.
And well, some like a flatter response than others. Audio is subjective ...
This is what I agree with:
"It depends on how wide a bandwidth is covered, John."
"A difference of just 0.25dB covering the tweeter's passband therefore appeared to be significant.
John Atkinson
Editor, Stereophile"
john curl said:
absurd? The crossover frequency of the woofer to the mids is 300Hz. This difference was easily audible, and the same as tilting the frequency response up or down.
Someone designs a speaker with tilted up frequency response and they could show up on a message board saying, "Help, my brain is melting" - it happened.
A flat response and people say, it sounds outstanding with these 5 recordings and okay and somewhat bright with some other stuff and unlistenable to some stuff.
A tilted down frequency response and you hear that most stuff sounds good, except for some bad stuff.
And well, some like a flatter response than others. Audio is subjective ...
PMA said:
It's hard to say since I haven't had much to compare it to. I haven't played it through the highs or mids. When I first hooked it up, I though it negatively altered the sound of the lower midrange frequencies, but I'm not positive - was a very small difference.
I like the flexibility of the DCX, and in the end it sounds much better than a crossover I had built. I was using a first order high pass (R-C) with an active 4th order LR lowpass at 55-60 hz, and I just couldn't get that to sound right, even changing slopes and stuff.
With our subs and our room, it sounded best crossover at 85Hz, 4th order Butterworth.
Yes, most complaints I've heard about the DCX are with the high frequency performance. Though, one speaker designer complained about it for bass performance as well.
I hope it is not too bad, since I'll be using it to optimize a 3 way active crossover for a 3 way dipole speaker I've built soon. I've heard many different takes on it, some positive. I don't really expect high end sound, just hope it's good enough to optimize a crossover and build an analog equivalent.
I hope it is not too bad, since I'll be using it to optimize a 3 way active crossover for a 3 way dipole speaker I've built soon. I've heard many different takes on it, some positive. I don't really expect high end sound, just hope it's good enough to optimize a crossover and build an analog equivalent.
Thanks Sy, glad to hear it. I've considered doing something about the gain structure like you mention. Seems like it needs more gain before the input in our system, and some gain attenuation at the output, to get the full dynamic range.
I plan on doing a discrete JFET crossover, and I think the DCX will work a little better with the dipole I'm doing due to it's lower sensitivity. So, hope I can get by with it.
I've heard about Jan's modification. Do know where I can find that? Is it on his website? (Edit: Yes, I see it is on his website.)
I plan on doing a discrete JFET crossover, and I think the DCX will work a little better with the dipole I'm doing due to it's lower sensitivity. So, hope I can get by with it.
I've heard about Jan's modification. Do know where I can find that? Is it on his website? (Edit: Yes, I see it is on his website.)
Last edited:
- Status
- Not open for further replies.