While building phono amplifiers is difficult enough, getting rid of humm and hiss is even more. And after that, checking your design against the reproduction curve is a pain without expensive tools. After all, the change in gain over the frequency band of 2020kHz makes it difficult to accurately measure the performance of your amp without having to adjust your scope every once in a while to read the gain.
Wouldn't it be nice if we could check our design so that for each frequency we ideally should have the same output level (and a lower/higher level would indicate trouble).
For that, it would be possible to build just a filter which could be installed before the input of your phono stage and would output according to the RIAA reproduction curve just as a record would.
Your input would then be a linelevel (=CD) input. So you can test the phono amp with recorded sounds of same recording level or a tone generator.
Figure 6.2 contains the formula for plotting the inverse RIAA function for the enhanced RIAA curve (remember from page 1, the one with the 3.18 uSec breakpoint)
When building an inverse RIAA filter it is very important to keep the following things in mind:
The latter will mean that there will be less signal left at the output, but whether we use a CD (2Volts) or a signal generator as a source, we probably will end with enough signal to feed the MM input stage of a phono amp (2 mVolt). The simplest way (and therefore maybe the best) is to build a voltage divider, with the lowest resistance being a fixed resistor value for the output impedance, and the upper part being the resulting impedance of the filter. The filter might have similar setup as used in the active feedback loop figure 3.4 on page 3. (See also the reference of Hagerman for a similar description).
The calculations for the filter are in itself not difficult, but it is hard to solve all values without help of a spreadsheet model.
Hmm, still working on it, it would be so nice if we could select a value for
some components and distill the remaining values from there. But for the moment
some trialerror with the spreadsheet is necessary to get the values right and
close to standard resistor and capacitor values.
Some SPICE simulations will be done to check the outcome of the calculations.
I made the spreadsheet such that I will end up with standard E96 values where
possible. After all, I do not mind to parallel capacitors but making resistor
values up from several standard values is not a good idea (each component adds
noise).
Fortunately, we have a spreadsheet at hand, and since the filter resembles the
active filter of page 3 in a way, we
do not have to work too hard to have a spreadsheet aiding us in some sensible
values for making your own inverse RIAA filter.
Note: Since the output of the function generator on the input of this filter adds to the resistor value R3, it is advised to choose the correct value of R3 by substracting the output impedance of the function generator (typically 50 Ohms) from the ideal/computed value. in principle, the same is true for the output impedance, but 600 Ohms is a reasonable value in the audio world so I expect this to be workable.
The following table defines a few sets of component values that match the formulas above, assuming a source impedance at the input of 50 Ohms.
Set#  R3  R1  C1  R2  C2  R_out 
My filter 1  2490  680k  4.7nF  54k9  1.36nF  604 
My filter 2  1540  464k  6.8nF  37k4  2nF  604 
Lipshitz  **  883k3  3.6nF  75k  1nF  604 
Lipshitz with t4  3600  910k  3.5nF  75k  1nF  604 
Hageman published  1910  511k 
6.2nF 
42k2  1.8nF 
536+60 
Hageman modified  1740  511k  6.2nF  41k2  1.8nF  604 
**: Lipshitz did not use the 4th timeconstant for breakpoint at 50kHz
*** Hageman also did not fully model the 4th timeconstant
Below is a chart wit the plotted frequency behaviour of all filters described above.
Building an inverse Filter as described above is done best using as many standard component values where possible. Based on this, my modified version of the Hagerman filter may be the best compromise.
The complete inverse RIAA filter is so small that it will easily fit in any small case. It is equally possible to build several versions and make them user selectable.
Component  Value  Made by  
C1  6.2nF  1.5nF and 4.7nF parallel Both Polypropylene 1% 

C2  1.8nF  Polypropylene 1%  
R1  511k  Metalfilm 0.6Watt, 1%  
R2  41k2  Metalfilm 0.6Watt, 1%  
R3 
1740 
Metalfilm 0.6Watt, 1%  
R out  604  Metalfilm 0.6Watt, 1%  
Not taken into account are parasitic capacitances and the connecting cables. When modelling in Spice, it's easy to account for these capacitances in a model to see how it impacts the inverse RIAA network.
Building the reverse filter was really easy: Just a small leftover piece of board was enough.
What remains is a small enclosure so I'm able to use it for testing my phono amps...
<< Page 1: Introduction 