Error in D. Self-Riaa?

[Conrad Hoffman]

IMO, Hagerman has some good stuff on inverse networks here. One thing I've found is that measuring RIAA accuracy directly using signals and meters if fraught with difficulty. The dynamic range required is large and if the meter switches ranges you lose accuracy. The best way would be with a ratio transformer, but not many have those hanging about in the lab. Thus, a carefully measured inverse network is best.

[Bonsai]

What about a sound card followed by an attenuator network to generate the inverse curve? I've done (and do) embedded programming, but nothing on the PC but I think this would be an absolute cinch with VB - generate and measure all out of one card. Maybe there's already some software out there, and if not, we should suggest it to RightMark.

[h_a]

If one hooks the circuit up to a pc, one can always do a frequency sweep and compare that to a calculated RIAA response (created in Excel with the measured gain at 1kHz). The comparison is not real-time, but that's usually not necessary as fine-tuning the RIAA is not a matter of changing single component values (as Lipshitz discusses at length ;-) ).

Thanks for the link to inverse RIAAs, might come handy though.

I don't know that I can neccessarily change it

Well, if you set R3 to zero, it has to be zero all the time. I don't see why a value for R3 is calculated at all. If you look e.g. at Table 3a, you see that R3 and R0 always show up together. Setting R3 to zero, only R0 remains in the equations, so there remains in fact no relation one could use to calculate R3.

[Conrad Hoffman]

Bonsai, that seems like a reasonable approach, assuming the s/n ratio is good enough and the attenuator is chosen wisely. I bring up the issue because people get interested in fractional dB errors, and the measurement isn't trivial given the combination of small signal levels, dynamic range and frequency range.

[scott wurcer]

Originally Posted by Bonsai

What about a sound card followed by an attenuator network to generate the inverse curve? I've done (and do) embedded programming, but nothing on the PC but I think this would be an absolute cinch with VB - generate and measure all out of one card. Maybe there's already some software out there, and if not, we should suggest it to RightMark.

I use an inverse RIAA 1/10th octave multitone signal and a sound card which you can average as long as you want. SoX will do it all for free.

[Bonsai]

Thanks Scott - I will take a look. Who/what is SoX?

[h_a]

Sure, that's certainly most practical!

I guess Scott means Sound eXchange, a powerfull command-line tool

SoX - Sound eXchange | HomePage

[Hennie]

The open loop transfer function of Self's circuit will influence the component values required in the feedback network. In the case of an op amp the open loop gain is high enough for it to be ignored, but not here.

You can try a single dominant pole approximation of the open loop transfer function with the actual LF gain and the actual dominant pole, and calculate the required feedback values accordingly. I did that many years ago, but without taking the open loop gain into account I just could not get an accurate RIAA curve, especially at the HF end of the RIAA CURVE IIRC. By taking the open loop transfer function into account, the required target transfer function was met satisfactorily. It might have required an extra component in the feedback network, I can't remember exactly

It is perhaps easier to simulate in Spice until you get the target curve right.

[Conrad Hoffman]

While we're on the topic of simple phono sections, here's the circuit used in the old Ace Audio preamp from around '74. It sounded quite good, though I've no idea how accurate it is. They also had a line stage, but eliminated it for the "zero distortion" version. From my archives.

[gk7]

I was not aware that sox can do an inverse RIAA. The manpage says "riaa Apply RIAA vinyl playback equalisation." How would you invert that ?