As most of you will already know, phono equalisation appeared to make mastering of high fidelity records a practical possibility. The groove amplitude on a record is a function of signal level and frequency. So the larger the signal, and/or the lower the frequency, the larger the required groove modulation. This leads to a pair of problems (1) at low frequencies the groove modulation would be excessive and be un-cuttable / un- traceable, while at high frequencies it would be too small, and lost in the surface noise. To get around this problem the signal reaching the cutting head is filtered to reduce low frequency amplitude levels, and increase high frequencies. In the early days each record company had their own filter, so eventually the RIAA stepped in and set a standard pre-emphasis for record mastering. Obviously, to correctly replay records which have been pre-emphasised your amplifier requires a de-emphasis filter which EXACTLY matches the pre-emphasis.
To make the job of building phono stages interesting the RIAA curve was designed to have three corner frequencies in it, but without them being linked easily. Suffice to say that it would also have been much better if E24/E48/E96 values had been around when it was finalised, as it is you end up having to make non-standard values to get an accurate response. Thus our phono stage must apply post replay equalisation across the whole bandwidth, apply about 50dB of gain and introduce minimum distortion, noise or additional phase shift - no wonder cd players look so easy.
So why have we got 3/4 time constants and a funny shaped curve? Noise is the simple answer. Because we have 10 octaves to cover a simple first order slope at say 100Hz would produce HF to loud to cut onto vinyl, thus a wiggle was put in between 500Hz and 2122Hz to limit the maximum variation at the frequency extremes to 20dB. Eventually we ended up with 3 turnover frequencies at 50.05Hz, 500.5Hz and 2122Hz. Years latter a 20Hz corner was introduced to attenuate rumble and arm/cartridge resonance effects, most designs which sound any good don't have 20Hz corners, but 10-15Hz corners instead, thus reducing low frequency phase shift in the 20-40Hz region. In the vast majority of valve systems any additional 20Hz level remaining after the phono stage will be subsequently attenuated by the line stage and power amplifier long before they ever meet the output tranny or loudspeaker.
One would imagine that designing a phono stage with the correct corner frequencies is easy, as they are in a Standard, however many phono stages can't even manage this. Many have the wrong frequencies as there design aims. A good example is the phono stage of a well known manufacturer, (hey, I don't want a liable case) if this is built and analysed, its equalisation is -3.5dB at 50Hz, +0.5dB at 500Hz, -1dB at 3KHz and +0.5dB at 20KHz compared to the RIAA standard. Please note that this is based upon an example I built from a schematic sent out by the manufacturer. It is not unique, I have seen many others whose schematics show that the de-emphasis filter has been designed with different corner frequencies.
If we restrict the discussion to passive equalisation networks, as they sound better and are feedback free implementations, we have two main options.
(1) A composite network which tackles all three corner frequencies, (50.05Hz, 500.5Hz and 2122Hz) in one hit.
(2) A split passive network which tackles two corners in one hit (50.05Hz and 500.5Hz) and the other (2122Hz) separately. They are normally separated by a gain stage.
THE CLASSIC paper on "how to do" a composite RIAA was by Stanley Lipshitz and published in the journal of the AES several years ago, 1979 issue 1/2 I believe. This is a magnificent maths intensive tour-de-force through RIAA network design. With thought and understanding a series of maths equations are extracted to calc all the relevant component values. These can then be set up on a spreadsheet to assist. The most useful thing is to calc and graph the error. Thus if the correct values are not available you can substitute real values and see how they effect the response.
In the common composite RIAA there are virtually no easy ways of calculating the corner frequencies from the component values as they are all in series/parallel combination with each other, resulting in the number of components effectively in circuit varies with frequency. Some things are still reasonably straight forward though. The values of C1 and R2 must give you 318uS for the 500.5Hz turnover point, all the others are due to parallel combinations.
The best way to work out your component values is to use the following:
(a) R1' x C1 = 2187 x 10^-6 sec
(b) R1' x C2 = 750 x 10^-6 sec
(c) R2 x C1 = 318 x 10^-6 sec
(also, C1/C2 = 2.9160 and R1'/R2 = 6.8774.)
And please note that R1' = (R1 x R0)/(R1 + R0).
Therefore, to design such a network, one would:
1. Choose a value for C1.
2. Calculate R1' from equation (a).
3. Calculate C2 from (b).
4. Calculate R2 from (c).
The only difficulty is trying to work out what R1 is, but this can be approximated fairly accurately.
We also lose a lot of signal going through the composite RIAA because of the voltage divider at the heart of it. The series resistor R1 and the grid resistor R0 are obviously a voltage divider reducing overall gain by around 6dB at dc, at 1KHz we will be a further 20dB down. Also as far as the rest of the network is concerned the grid resistor (R0) is in parallel with the series resistor (R1), thus R0+R1 must be a good dc load for the driver valve, R1 alone must be a good ac load for the driver valve, and R1 in parallel with R0 must be the required value for the RIAA equalisation. All this obviously limits the possible range of values, mainly due to the drive requirements of the preceding tube, it also makes the network a difficult ac load as it almost halves in value across the frequency range, possibly introducing still further attenuation at the frequency extremes as gain varies with load in the driver valve.
It does appear from the above tale that the composite RIAA network is hardly worth using, this is not strictly true. Its main advantage is that it only requires two active gain stages to get your phono stage up and running, this having the huge commercial advantage of cost and phase linearity - yes your phono stage will be absolute phase linear to join up nicely with your line stages.
If we move on now to the less common split RIAA networks, these have had a lot of exposure recently in Sound Practices. The major reason not to use them has always been cost - who wants to add another amplifier in a commercial product when the composite RIAA is around? Things that can work out better with split RIAA can be: load matching, gain/overload margins (not really an issue with tubes) and simplicity of design. In a split RIAA network the corner frequencies are easy to spot, and thus easy to recalculate and adjust if required. As they tend to use three active stages you can end up with to much gain or noise if care isn't taken, and the output will be absolute phase reversed, which may or may not be a problem, as you will be able to correct for this in several places if you are sensitive to absolute phase. We need two bits:
The split RIAA method involves having two separate RC filters, each is driven by its own active stage. The network which is usually place first is that of R1, C1 - these give the 2122Hz corner (or 75uS). Typical values used in tube circuits are: R1=100K, C1=750pF; or R1=50K, C1=1n5F etc. As you can see its pretty easy to calculate, and can be achieved using standard values. It is usually placed first because the signal coming off disc has its highest amplitudes at high frequencies, so we wish to avoid high frequency overload in our gain stages so we filter it before it becomes to large. The second reason is that we wish to defer filtering the low frequencies until we have amplified them up to a reasonably healthy level. I have heard it argued that the stages should be the other way around because putting the HF filtering second will help reduce noise - this is possible but just ignores the reasoning above.
In the second stage we take care of the 500.5Hz and 50.05Hz corner frequencies, as these are a decade apart we again have a pretty straightforward set of calculations to perform. C2 and R3 give you the 500.5Hz corner (318uS) and R2 and C2 give you the 50.05Hz corner (3180uS), it is thus obvious that R2=10*R3. A typical value of C2 is 0.033uF with R2 and R3 calculated from there.
The only part of the filter not discussed thus far is the 20Hz corner for rumble filtering. I tend to place this at around 15Hz and normally use the cap coupling the final stage to the volume pot as the filter (ie 0.12uF and 100K//1M0) in a full preamp, or between the first stage and the first filter for a stand alone. All the other coupling and bypass caps are then made large enough to introduce no further low frequency filtering (ie corner frequencies below 2Hz). I find that if there are several corners between 2Hz and 15Hz the bass sound thin and weak, without scale or real timbre, probably due to phase error brought on by the other filters. Note: some people like this sound and say it gives the bass speed and clarity. Probably does if your speakers don't go down to 20 odd Hz, but mine do and to me it sounds wrong.
I hope this has been of some use, I shall now stand back and await the inevitable fallout.
For further helpful information I have now added schematics of two of my most recent phono stage.
The first (ph102.gif) is a composite RIAA stage with 417A input tube and 6922 output tube. This is far and away the best sounding composite RIAA phono stage I have ever built. The 417A offers a very quiet input tube, coupled with a low plate impedance and high transconductance to effectively drive the equalisation network.
NOTE: R1 and R0 in the schematic have 2 values each.....use EITHER 475K with 1M0 OR 359K with 592K.
The second (phono_am.gif) is a three stage split RIAA design using 417A, 6922, 6922. With sufficient care a low noise mc phono stage can be built from this circuit which offers very high quality reproduction.
Copyright Notice and Linking permission:
These HTML documents, and associated graphics, are Copyright 1996, SJS
Electroacoustics. These documents may not be reproduced in any manner without the
permission of the copyright owner. The schematics presented here are Copyright 1996, SJS
Electroacoustics and are offered for construction on a non-profit basis only. Any persons
wishing to profit from these designs, or requiring similar should contact SJS
Electroacoustics for licensing. We invite the audio and music community to link to this
home page, which will be periodically revised.
This document has been revised and updated April 4, 1999.