Inverse RIAA/RIAA Encode-Decode Accuracy
Posted: Wed Jul 22, 2015 3:59 pm
I altered the schematic of the inverse RIAA to move the ultrasonic time constant to 3.3 us.
It's effect at 20 kHz turns out to be about 1/2 dB.
If the 100 pF Cfb is added the out-of-band ultrasonic HF rise is tamed further.
With both in circuit the 20 kHz drop is about 0.75 dB.
I originally checked the playback EQ accuracy using the Jung/Lipshitz network through the preamp.
The playback RIAA network I built for the protoboard used "8300 pF" 2.5% caps. (1500||6800 pF, 383K, 38K3, 11K8).
The overall accuracy was ruler flat using the Jung/Lipshitz Inverse RIAA with the RIAA playback network.
The Inverse RIAA network used 10 nF 1% caps and series resistors to get the R exact. (316K +2K etc.)
As a reality check I decided to check the Inverse RIAA feeding the RIAA playback EQ end-to-end.
I used a sine input.
Again they were ruler flat with the exception of the minor 20 kHz droop.
A picture is often worth a thousand words and time domain accuracy - square wave response - can speak volumes about frequency response.
Though we don't record square waves they can tell us a lot about the frequency domain in one quick glance.
If the Inverse RIAA and RIAA curves are precisely matched, then a square wave in should produce a square wave out.
RIAA pre-emphasis is a brutal thing.
The top trace is the input, the bottom trace the output from the Inverse RIAA filter.
To give you an idea of the dynamic range a pre-emphasized square wave generates, the input trace is 200 mV/div, the output trace is 5V/div.
Headroom is precious and to say that Inverse RIAA is a differentiator is an understatement.
Output of an Inverse RIAA network driven by a 1 kHz square wave.
So what happens to waveform fidelity after RIAA de-emphasis?
Can this be put back together accurately?
Yes it can.
Encode-Decode Response of Inverse/Playback RIAA networks passing a 1 kHz square wave.
Top trace is input, bottom trace is after encode-decode.
To provide absolute purist Inverse RIAA requires an op amp with almost infinite bandwidth.
Since we don't have those I'll accept the single pole rolloff and the 3/4 dB 20 kHz penalty.
Others might want a 2 pole pre-filter at 50 kHz.
The hits of the 60's and 70 used LM709s (Neumann SE-66) and LF356s (Neumann SAB-74) in these stages.
We should be able to do just fine today with a 5532, NJM2114 or LME49860.
It's effect at 20 kHz turns out to be about 1/2 dB.
If the 100 pF Cfb is added the out-of-band ultrasonic HF rise is tamed further.
With both in circuit the 20 kHz drop is about 0.75 dB.
I originally checked the playback EQ accuracy using the Jung/Lipshitz network through the preamp.
The playback RIAA network I built for the protoboard used "8300 pF" 2.5% caps. (1500||6800 pF, 383K, 38K3, 11K8).
The overall accuracy was ruler flat using the Jung/Lipshitz Inverse RIAA with the RIAA playback network.
The Inverse RIAA network used 10 nF 1% caps and series resistors to get the R exact. (316K +2K etc.)
As a reality check I decided to check the Inverse RIAA feeding the RIAA playback EQ end-to-end.
I used a sine input.
Again they were ruler flat with the exception of the minor 20 kHz droop.
A picture is often worth a thousand words and time domain accuracy - square wave response - can speak volumes about frequency response.
Though we don't record square waves they can tell us a lot about the frequency domain in one quick glance.
If the Inverse RIAA and RIAA curves are precisely matched, then a square wave in should produce a square wave out.
RIAA pre-emphasis is a brutal thing.
The top trace is the input, the bottom trace the output from the Inverse RIAA filter.
To give you an idea of the dynamic range a pre-emphasized square wave generates, the input trace is 200 mV/div, the output trace is 5V/div.
Headroom is precious and to say that Inverse RIAA is a differentiator is an understatement.
Output of an Inverse RIAA network driven by a 1 kHz square wave.
So what happens to waveform fidelity after RIAA de-emphasis?
Can this be put back together accurately?
Yes it can.
Encode-Decode Response of Inverse/Playback RIAA networks passing a 1 kHz square wave.
Top trace is input, bottom trace is after encode-decode.
To provide absolute purist Inverse RIAA requires an op amp with almost infinite bandwidth.
Since we don't have those I'll accept the single pole rolloff and the 3/4 dB 20 kHz penalty.
Others might want a 2 pole pre-filter at 50 kHz.
The hits of the 60's and 70 used LM709s (Neumann SE-66) and LF356s (Neumann SAB-74) in these stages.
We should be able to do just fine today with a 5532, NJM2114 or LME49860.