Another Wrench In the Works... Inductive RIAA

[wrenchone]

I have read on several occasions about using inductive RIAA filters, and how some pundits think they sound better than RC filters. One opmap-based feedback design that looked reasonably easy to implement was published in Audio Express a few years ago.

Attached is a schematic for a gain cell using 2 JFETs and feedback-style inductive equalization. The equalization is set up fro the first 50Hz corner.One could increase the gain of the circuit and add the second corner, or put the higher frequency corner in a second stage. The gain cell uses capacitive coupling, but nothing's perfect, and the couopling capacitor value is small enough that one could contemplate using a teflon capacitor.

Anyway, food for thought. I won't be building this one for a while, as I have a big backlog of projects, but I thought I'd sling the concept out and let i fester in public...

[wrenchone]

Attached is a picture of a more complete circuit, though I think a follower might be advisable for interfacing to the cold, cruel world. I want to look a the turnover points a bit more closely, though the 1kHz gain as plotted by PSpice is pretty close to my goal of 40 dB.

[wrenchone]

R15 should be 1.18k to roll off the gain reduction of the second stage at 50 kHz. Ill try rolling that value into the sim at work tomorrow.

[wrenchone]

Schematic for modified circuit here....

[wrenchone]

Here's the response, looking pretty RIAA-ish.

[h_a]

Hi,

some cool stuff you have here ;-)

Still, your inductors are huge! I mean 159mH will be the big block of inductors; that beast will in practice maybe pickup so much noise that it's no longer useable for a phonostage. Further, the inductor is then iron-cored which usually limits bandwidth (probably small problem) and introduces distortion.

Have fun, Hannes

[wrenchone]

The inductor values are actually rather small compared to many other inductive RIAA setups. Wound on gapped ferrite pot cores, the inductors will also be reasonably sized and linear. The 159 mH inductor is ~500t on a 2616 pot core gapped for 630 nHT^2 inductance coefficient (a standard size and inductance coefficient from many suppliers). A 3019 core may be a better choice, requiring fewer turns. Pot cores are to a large extent self-shielding, and cans are offered as accessories that offer even more shielding. After all, pot cores were developed for use in high precision filter circuits where stability, precision, and freedom from noise pickup are all very important. I'm more worried about stray capacitance, though that can be fixed to a certain exent with a multisection bobbin (also a standard offering). The main thing to worry about is factoring the wire resistance into the RIAA filter settings.

I deliberately forced the design to accomodate a small value of gain setting resistor in order to make the inductor values more reasonable. A standard passive RIAA network would need to be low impedance (the standard ones available are 600 ohms and cost hundreds of dollars). At 600 ohms, the inductors are huge, and it requires a matching transformer or other sleight of hand to drive them. This approach here is very similar to one mentioned in a 2003 article in Audio Xpress, but that guy used an opamp, and I have a fetish for discrete JFET designs...

[nicoch58]

you can find some info here http://www.intactaudio.com/forum/vie...078d68807f414b

there is Thread on 1k5 too ,with single coil , not like S&B ,Tango or SilK that are complete module

tango/pultec on pic

[wrenchone]

The local surplus emporium has canned 4229-sized inductors - as a bonus, they have the PC-mount base plates, and weren't varnished, so they're easy to take apart. As a last cherry on top, the bobbins are 2-sectioned, for a lower winding capacitance. These cores are a little larger than I envisioned, but they have everything needed to make at least the low frequency inductor for a L/R-type RIAA filter. Once I get my BA 2009 project out of the way, this might be next up.

[nicoch58]

for what I read for good riaa ,under 0,3db we must use lower RLc , the S&B give up on 10k.....too bad