[The Phantom]

In an old notebook, I found an inverse RIAA network which I'm sure I got
from some magazine years ago. I decided to analyze it and see if I could
derive a set of design equations which would make it easy to calculate the
component values for various time constants and impedance levels. The
circuit as originally published has 4 time constants; the usual 3
associated with RIAA equalization and a 4th corresponding to a corner
frequency of about 2 MHz. The circuit and originally published component
values are as shown:


1600 pF

C2 ||
.---------------||-----------------|
| || |
| |
| 68.4k 4650 pF |
| ___ || |
In -----o-------|___|---------||-----------o--- Out
| || C1 |
| R3 R2 |
| ___ |
'--------------|___|---------------|
|
470k |
|
|
.-.
| |
R1 | | 47
'-'
|
|
===
GND

R1 is a low value of 47 ohms because it was intended that the output would
be fed into the phono input of a preamp.

The 4 time constants are as designated below, with the values for RIAA
equalization given. Using these values (and 47 ohms for R1) with the
formulas will give the component values in the originally published
circuit. But the time constants can be selected arbitrarily as long as
their values are strictly decreasing as you go from T1 to T4.

To design for 4 time constants and arbitrary impedance, select the time
constants and choose a value for R1. Then calculate the values of the
other 4 components with the formulas given.

T1 = 3180 uS
T2 = 318 uS
T3 = 75 uS
T4 = .07492 uS

R1 (T1 T3 - T2 T4)
R2 = ------------------
T2 T4

R1 (T1 T3 - T2 T4) (T1 T2 T3 + T1 T3 T4 - T1 T2 T4 - T2 T3 T4)^2
R3 = ----------------------------------------------------------------
T1 T2 T3 T4 (T1 - T2) (T2 - T3) (T1 - T4) (T3 - T4)

T1 T2 T3 T4 (T1 - T2) (T2 - T3) (T1 - T4) (T3 - T4)
C1 = ----------------------------------------------------------------
R1 (T1 T3 - T2 T4)^2 (T1 T2 T3 + T1 T3 T4 - T1 T2 T4 - T2 T3 T4)

T1 T2 T3 T4
C2 = ----------------------------------------------
R1 (T1 T2 T3 + T1 T3 T4 - T1 T2 T4 - T2 T3 T4)

While I was working on this it occurred to me that some phono inputs will
have an impedance of the same order as R1. If this is so, the value of R1
used in the formulas should be the effective value of the actual R1 in
parallel with the input resistance of the phono input to which the output
of the network is connected. Or, one could adjust R1 so that its value in
parallel with the resistance of the phono input gives the desired value.

This means that a particular network may not give the same result with
different preamps if those preamps have different input resistances. For
the very best accuracy, the network must be designed to take into account
the loading effect of a particular preamp. I suppose the network could be
put in a small project box with a battery powered unity gain buffer to
provide a very low output impedance. And, of course, the network must be
driven from a low impedance source, preferably less than R1/10.


Rather than leave the T4 time constant at 2 MHz, it might be good to set it
to 3.18 uS as used with Neumann cutters.

When this is done and the component values calculated (with R1 still 47
ohms) the results are as follows:

R1 = 47
R2 = 11037.90566 ohms
R3 = 1567.50513 ohms
C1 = .196929 uF
C2 = .0699975 uF

When building a network like this, the capacitors are usually the
components that are hardest to find as standard values. Notice that in the
example just calculated, C1 could be a .15 uF and a .047 uF in parallel,
and C2 could be a .05 uF and a .02 uF in parallel. And, of course, if a
very accurate network is desired, the components should measured and
trimmed or selected to be as close as possible to the exact calculated
values. Components that are stable and have a low temperature coefficient
should be used.

The original published version has nearly a 10000:1 reduction in signal
amplitude at the low frequency end of the audio spectrum. The just
calculated set of values doesn't attenuate the low end of the spectrum as
much as the original design. To get a lower output one could tap down on
R1. For example, R1 could be a 43 ohm resistor in series with a 4 ohm
resistor, with the 4 ohm resistor closest to ground. The output could then
be taken across the 4 ohm resistor; then the preamp loading would have less
effect on the network performance. Or even more extreme ratios could be
used to get even more attenuation.

This is one of those "all in one" networks that have been under discussion.
The time constants aren't just simple products of resistor and capacitor
values, such as R2 C2. That's why the formulas given above are so
complicated. But this doesn't mean that the complete network won't have
the desired corner frequencies. It will be just as accurate as a network
consisting of separate RC branches separated by buffer amplifiers. You
just have to use the correct theoretical component values as calculated by
the formulas.

I've posted some better looking, non-ascii versions of the formulas and
schematics on alt.binaries.schematics.electronic.

[Chris Hornbeck]

On Thu, 17 May 2007 22:46:15 -0700, The Phantom
wrote:

R1 = 47

Typo? Gonna take a really big generator to get out of the
mud with that kind of attenuation.

All good fortune,

Chris Hornbeck
"A little note on SUCCESS:

At age 1................you don't give a ****.
At age 4, success is ....not peeing in your pants.
At age 12, success is .....having friends.
At age 16, success is .......having a drivers license.
At age 20, success is .........having sex.
At age 35, success is ...........having money.
At age 50, success is ...........having money.
At age 60, success is .........having sex.
At age 70, success is .......having a drivers license.
At age 75, success is .....having friends.
At age 80, success is ....not peeing in your pants.
At age 100,.............you don't give a ****."

- anon from the Great and Powerful Internet, with
corrections by Patrick Turner

[The Phantom]

On Fri, 18 May 2007 06:09:04 GMT, Chris Hornbeck
wrote:

On Thu, 17 May 2007 22:46:15 -0700, The Phantom
wrote:

R1 = 47

Typo? Gonna take a really big generator to get out of the
mud with that kind of attenuation.

The idea was to get a millivolt signal out to apply to the phono input of
a preamp.


All good fortune,

Chris Hornbeck
"A little note on SUCCESS:

At age 1................you don't give a ****.
At age 4, success is ....not peeing in your pants.
At age 12, success is .....having friends.
At age 16, success is .......having a drivers license.
At age 20, success is .........having sex.
At age 35, success is ...........having money.
At age 50, success is ...........having money.
At age 60, success is .........having sex.
At age 70, success is .......having a drivers license.
At age 75, success is .....having friends.
At age 80, success is ....not peeing in your pants.
At age 100,.............you don't give a ****."

- anon from the Great and Powerful Internet, with
corrections by Patrick Turner

[Patrick Turner]

Chris Hornbeck wrote:

On Thu, 17 May 2007 22:46:15 -0700, The Phantom
wrote:

R1 = 47

Typo? Gonna take a really big generator to get out of the
mud with that kind of attenuation.

No, the 10,000:1 attenuation at 10Hz is fine, so that
5Vrms from signal gene and all F becomes 0.5 mV at 10Hz,
5mV at 1 kHz, and 50mV at 21khz, which mimics a moving magnet cart
output.
The attemuation could be increased 20dB for MC carts, ie
use 42 ohms and 5 ohms as a divider instead of 47.

Noise is reduced because the shunt R of 47 or 5 ohms is so low.

Patrick Turner.










All good fortune,

Chris Hornbeck
"A little note on SUCCESS:

At age 1................you don't give a ****.
At age 4, success is ....not peeing in your pants.
At age 12, success is .....having friends.
At age 16, success is .......having a drivers license.
At age 20, success is .........having sex.
At age 35, success is ...........having money.
At age 50, success is ...........having money.
At age 60, success is .........having sex.
At age 70, success is .......having a drivers license.
At age 75, success is .....having friends.
At age 80, success is ....not peeing in your pants.
At age 100,.............you don't give a ****."

- anon from the Great and Powerful Internet, with
corrections by Patrick Turner

[robert casey]

No, the 10,000:1 attenuation at 10Hz is fine, so that
5Vrms from signal gene and all F becomes 0.5 mV at 10Hz,
5mV at 1 kHz, and 50mV at 21khz, which mimics a moving magnet cart
output.
The attemuation could be increased 20dB for MC carts, ie
use 42 ohms and 5 ohms as a divider instead of 47.

Noise is reduced because the shunt R of 47 or 5 ohms is so low.

Patrick Turner.

You could almost use a CD player as the "generator", though CD players
produce a little lower output than 5V. So make it 36 ohms and 10 ohms
for an MC cart simulation.

[Chris Hornbeck]

On Fri, 18 May 2007 09:32:56 GMT, Patrick Turner
wrote:

Chris Hornbeck wrote:

R1 = 47

Typo? Gonna take a really big generator to get out of the
mud with that kind of attenuation.

No, the 10,000:1 attenuation at 10Hz is fine, so that
5Vrms from signal gene and all F becomes 0.5 mV at 10Hz,
5mV at 1 kHz, and 50mV at 21khz, which mimics a moving magnet cart
output.
The attemuation could be increased 20dB for MC carts, ie
use 42 ohms and 5 ohms as a divider instead of 47.

Kind of a stretch, but to each their own. The usual
inverse network has 40dB midband attenuation, which
both relaxes generator voltage requirements and
doesn't as easily run up as close to preamp noise.
Ain't no thang, ideally, though.

Noise is reduced because the shunt R of 47 or 5 ohms is so low.

Noise will be dominated by any real preamp, and our
only concern is an adequate S/N at the input for
measurement. Again, it's usually no thang.

Much thanks, as always,

Chris Hornbeck

[Patrick Turner]

Chris Hornbeck wrote:

On Fri, 18 May 2007 09:32:56 GMT, Patrick Turner
wrote:

Chris Hornbeck wrote:

R1 = 47

Typo? Gonna take a really big generator to get out of the
mud with that kind of attenuation.

No, the 10,000:1 attenuation at 10Hz is fine, so that
5Vrms from signal gene and all F becomes 0.5 mV at 10Hz,
5mV at 1 kHz, and 50mV at 21khz, which mimics a moving magnet cart
output.
The attemuation could be increased 20dB for MC carts, ie
use 42 ohms and 5 ohms as a divider instead of 47.

Kind of a stretch, but to each their own. The usual
inverse network has 40dB midband attenuation, which
both relaxes generator voltage requirements and
doesn't as easily run up as close to preamp noise.
Ain't no thang, ideally, though.

Noise is reduced because the shunt R of 47 or 5 ohms is so low.

Noise will be dominated by any real preamp, and our
only concern is an adequate S/N at the input for
measurement. Again, it's usually no thang.

If you place 5 ohms to shunt the input of an amp system
and the signal level is 0.3mV then to get SNR = -66dB, the noise must be
less
than 0.15uV, and you are getting down to easy limits to achieve.
Trying for 0.015uV of noise becomes quite difficult.

Patrick Turner.

Much thanks, as always,

Chris Hornbeck