RIAA Network Transfer Function

[BRSHiFi]

Does anyone know how to find the component values for the attached RIAA network?

I am building a tube differential phono stage and will verify it spice.

 

[godfrey]

Where's the input and output?
You also need to take into account the source and load impedances.

 

[BRSHiFi]

More specifically, the circuit is like this. The network is between the two differential outputs.

 

[KSTR]

Look here, this seems to be the equivalent topology to what you have in mind :
phonostage

General information :
Cut and Thrust: RIAA LP Equalization | Stereophile.com

And in LTspice, a reference can be simply constructed with an AC voltage source of magnitude 10 connected to an "e"-source (VCVS) with the transfer function given in direct Laplace notation, using the known time constants :
laplace (1 + 318u*s) / (1 + 3180u*s) / (1+ 75u*s)
Comparison with the Stereophile article shows 100% precise accuracy (compare the detailed numbers given in the text) as expected right away, but always better double check

 

[AndrewT]

I think it is Harm that gives a method to determine the component values.
There is also a method given by one of our audio celebs, Bax or JLH or ??

 

[KSTR]

Calculation, for a starting point, is quite straightforward. RIAA EQ is a one decade wide bass shelf (from 50Hz to 500Hz) with a pole on top (at 2.1kHz). Problem is the somewhat ill-defined plate impedance of the tubes for the actual circuit which influences all three time constants, being part of the source impedance the network works upon.

 

[SY]

I've attached the topology you want. There's an RIAA calculator here: KAB Electro Acoustics http://www.kabusa.com which you can use as follows.

Connect the two resistors marked R! in my drawing to the 12AX7 plates. For the R1 value in the calculator, use 2 times R1 in the drawing plus two times the source impedance of the driving stage (the formula for the source resistance of a diff amp may be found in Morgan Jones's "Valve Amplifiers" as well as any other standard tube text. The other components can be directly plugged into the KAB calculator.

 

[BRSHiFi]

Tried it in simulation. I receive what resembles an RIAA curve, but the voltage differential is supposed to be reasonably flat, as I am running inverse RIAA networks on the inputs, along with the sine wave generator to simulate a cartridge.

The plate impedance for a 12AX7 is around 70k, in parallel with 220k, and double those quantities is about 560k. I put the values in the calculator and came up with the values in the schematic below. I wondered if I calculated it properly?

 

[SY]

You really do want to incorporate the build-out resistors- that will lower distortion, especially at high frequencies, and reduce the dependence of the frequency response on tube aging and drift. There is something of a noise penalty there, unfortunately. There are also ways around that, but with added complication.

Let's pick 100k as the buildout resistor. The equivalent series resistance is then 200k + 2x R_L||R_p. The second term is about 53k, so the total equivalent resistance is 306k. Plugging that into the KAB kalculator, R2 is 44k5, C1 = 7n15, C2 = 2n45.

 

[BRSHiFi]

What are the build-out resistors? Where do they go?

 

[SY]

See my post 7. Attach the ends of each R1 to one of the diff amp plates. Take the signal to the next stage (also balanced) from the caps on the right (though you could optionally direct couple).

 

[BRSHiFi]

Now it is doing this.

 

[SY]

Pick off the differential signal from either side of C6. Then substitute the correct values for the RIAA components.

 

[BRSHiFi]

The calculated values and 100k build-out resistor flattens out the response better, but those values are not standard. I tried it with other build-out values and the calculator, but the others produce WAY too much deviation.

Is there some other variable somewhere we are missing?

 

[BRSHiFi]

I switched the resistors to 62k ohm and re-ran the calculations, as 62k produces values for the RIAA network that are closest to real life component values.

I tried it with a few resistors and none of the combinations produce a flat response. Are we missing something?

 

[SY]

The values may not be standard, but you can build them up out of standard values using paralleling. You can also diddle the values of the buildout resistors in the calculator to try to steer the values of the other components.

 

[BRSHiFi]

Flat Enough?

Do you think this RIAA response is flat enough? The highest gain is a little under 33mV and the 20K is about 30.5 mV.

Or can it be better? This is with all standard values.

 

[SY]

That's about +/-0.3dB, which isn't fantastic, but isn't bad, either. If you put small resistor in series with the shunt cap, you might be able to bring up the top octave.

Keep in mind that the 12AX7 will still be the biggest variable- the assumed plate resistance may easily change by 10-20% tube to tube or as a tube ages. See what effect that has on the RIAA curve. That's one reason I usually use low plate resistance tubes in that spot, but that brings in other complications (especially stability).

 

[jackinnj]

The values may not be standard, but you can build them up out of standard values using paralleling.

Useful to have one of those 3x 365pF variable tuning caps from an old a.m. radio.

 

[BRSHiFi]

That upper octave is stubborn! I tried a few resistors in series with the shunt cap and it seems to have added roll-off sooner, and made the problem worse. The lower values did not help at all.