There is more than one good reason for making accurate RIAA networks. One is BECAUSE WE CAN. It is just good engineering. Second, is the embarrassment when we get reviewed by 'Stereophile' for example, and we have irregular measured frequency response, this has happened to me, (many years ago), and third, we have a reference so that we can accurately compare different phono stages. This should be enough, and it is not as difficult as it looks.
Just wait, they'll ask you to prove it. Then they'll prove you wrong via "hands waiving" with references to high school physics.
I'm still a doubter. Color me blue, I don't understand a iota of your explanation.
See
http://www.physics.sc.edu/kunchur/p...isalignment-of-acoustic-signals---Kunchur.pdf
http://www.physics.sc.edu/kunchur/papers/Temporal-resolution-by-bandwidth-restriction--Kunchur.pdf
http://www.physics.sc.edu/kunchur/p...n-and-bandwidth-of-human-hearing--Kunchur.pdf
Website:
Information for prospective students
These discussions often founder on what it's possible to hear. This might be helpful.
Last edited:
Keith Johnson has maintained that part of the "better" sound from vinyl comes from the secondary motions in the stylus that cause a solo instrument to "bloom" in space.
There is somewhere some academic stuff around the sensitivity to small response variations over large bands. I have always prioritized fixing the known identified issues like frequency response and distortion above .1% before getting sidetracked on esoteric stuff that is harder to defend.
These small errors in response can be fixed, however using 1% parts won't do it. Not until the whole network has been checked as a network. A cap that is +1% and a resistor that is -1% may well put you out of range. further that 1% resistor was 1% at factory test. Add all the environment and age effects and it may well be out of tolerance. And film caps closer than 2% are very difficult to make, the yields are low (you are buying all the rejects). If you are using eq in a feedback network the forward gain will affect the response. for 1% you need a lot of feedback (open loop gain) to get in tolerance. For a 40 dB at 1 KHz RIAA that would be a minimum OL gain at 20 Hz of 100 dB and 120 dB will get you closer. For a moving coil 66 dB gain thats 140 dB or so. 140 dB of open loop gain is not a trivial thing to manage in terms of stability.
There is somewhere some academic stuff around the sensitivity to small response variations over large bands. I have always prioritized fixing the known identified issues like frequency response and distortion above .1% before getting sidetracked on esoteric stuff that is harder to defend.
These small errors in response can be fixed, however using 1% parts won't do it. Not until the whole network has been checked as a network. A cap that is +1% and a resistor that is -1% may well put you out of range. further that 1% resistor was 1% at factory test. Add all the environment and age effects and it may well be out of tolerance. And film caps closer than 2% are very difficult to make, the yields are low (you are buying all the rejects). If you are using eq in a feedback network the forward gain will affect the response. for 1% you need a lot of feedback (open loop gain) to get in tolerance. For a 40 dB at 1 KHz RIAA that would be a minimum OL gain at 20 Hz of 100 dB and 120 dB will get you closer. For a moving coil 66 dB gain thats 140 dB or so. 140 dB of open loop gain is not a trivial thing to manage in terms of stability.
Last edited:
Thanks, very interesting, saved for future reference. Not sure though how this applies to the audibility of 0.1dB errors in the RIAA response.
Demian, where do you get all this crazy stuff? I get 1% caps, all the time. I select from there. Once they set up the machine, they all pretty much make it through, except for Dick Marsh's specials, made by Rel. also. I can work to 1/100 of a dB if I need to, but why bother? I think that how we handle out of band garbage is more important
I'm not sure it does, but earlier, you said
Kunchur's subjects can discriminate a 3mm change in position in the forward back axis position (@5 microsecond) of the transducers at a distance of @4 meters. In another experiment he got about the same results using a low pass filter. This is equivalent of moving the listener off axis. This might get him pretty close to your 1/8".
These were experimental conditions and real life is different but it's good to know what a psychoacoustic baseline might be
Frank,
There are two different issues here.
One is RIAA channel matching. As I said, I am not sure if 0.1dB (mis)matching is audible (I'd say rather not, but hell, what do I know?).
The other one is matching the RIAA characteristic within 0.1dB. While this is easy to measure as well, I'm pretty confident this can not be identified in a blind test. Perhaps someone from the vinyl industry could jump in and tell us what's the vinyl processing RIAA tolerance, from master tape to cutting.
This is an obviously important difference in practice. Assume you have 10 100nF caps of 2% precision. You can be pretty sure you'll find a pair matched to better than 0.5% (in fact you'll end up with several). But if you add the condition to have them with 0.5% precision, unless you are lucky, you may need more than 10 pcs to sort from. At least that's my experience.
But then again, there's technically not much trouble to reach both 0.1 dB target, so why not?
Again, it is not the ABSOLUTE frequency response of the RIAA network that creates the problem. It is the COMPARATIVE frequency response of the RIAA network to the RIAA standard that makes a level playing field for comparison between designs.
Once, long ago, I adopted an RIAA network for the JC-1 from the Levinson LMP-2, that was designed by Dick Burwin (here it comes, the usual name dropping: One of the founders of Analog Devices, and who, in 1966, developed a hybrid op amp with 100V/us slew rate for ADI, using a topology NOW familiar to all of us). I figured that he knew what he was doing, so I used his network. Guess what, it was slightly off, and when measured in 'Audio' magazine, this was noted. When I asked Dick Burwin about it, he said the he liked the sound of that EQ and that was enough for him at the time. I fixed it (mostly) with a simple resistor change, but I learned my lession, to "Trust, but verify".
Once, long ago, I adopted an RIAA network for the JC-1 from the Levinson LMP-2, that was designed by Dick Burwin (here it comes, the usual name dropping: One of the founders of Analog Devices, and who, in 1966, developed a hybrid op amp with 100V/us slew rate for ADI, using a topology NOW familiar to all of us). I figured that he knew what he was doing, so I used his network. Guess what, it was slightly off, and when measured in 'Audio' magazine, this was noted. When I asked Dick Burwin about it, he said the he liked the sound of that EQ and that was enough for him at the time. I fixed it (mostly) with a simple resistor change, but I learned my lession, to "Trust, but verify".
I must be dense today, but what do you mean here by "the ABSOLUTE frequency response of the RIAA network"?
I have the experience that the EQ on records is just about, and in most cases textbook flat restoring Riaa anti filter tends to be subjectively bright. Maybe its the MC carts too, the whole chain is based on loose industrial tech strictness anyway. Did you discover such a pattern with your record collections, or mine are mostly a bit off?
Don't know, but as long as the RIAA standard is in an absolute frequency scale, I guess it's the same as the "relative" to RIAA curve.
That's what I would expect. And that's why I think matching the RIAA standard to 0.1dB is useless. Matching L/R channels to 0.1dB is another kettle of fish, although I am still not convinced it's really required. Other than for avoiding to dine and wine some Stereophile guys, against a good review. It's measurable, so it can potentially be dangerous for the business!
What about the gross mismatch between channels of the cartridge itself? A lot of cartridge manufacturers guarentee only an error between channels of less than 1.5dB and that's 15 times the value that's claimed to be heared in equalization.
It's a mechanical system with large tolerances, after all.
Have fun, Hannes
It's a mechanical system with large tolerances, after all.
Have fun, Hannes