User avatar
Wayne Kirkwood
Posts: 41
Joined: Tue Jul 21, 2015 12:43 pm
Location: Dallas, Texas
Contact:

A Second-Order 12dB/Octave Elliptic Equalizer

Post: # 47958Unread post Wayne Kirkwood
Fri Sep 29, 2017 5:39 pm

A Second-Order 12dB/Octave Elliptic Equalizer

Hi Trolls - it's been awhile.

Around 2006 I introduced the Precision MS Matrix and a few years later the Stereo Width Controller. The question about how to implement Elliptic EQ with both of them comes up often.

Until recently the answer has been to insert a high pass filter in Side with the MS and a low pass filter in Side when using the Width controller. This approach works fine when the filter is single-order 6 dB/octave. Both the MS and LR±S Width approaches yield the same results that are functionally identical to the Neumann EE-70/EE-77, the ADT and others.

The problem with any 6 dB/octave Elliptic EQ is that it messes with midrange stereo imaging.


An EEQ filter has two responses: One is the lateral to vertical transition which is a high pass response. The second response, which is Left-Right crosstalk, has a low pass filter shape. A low vertical crossover point reduces midrange separation to an already unacceptable degree and the problem only becomes worse as the EEQ frequency is increased. An EEQ must be used carefully and only when necessary.

Around 2014 I began to explore higher-order filters for Elliptic EQ and discovered that one of the responses, regardless of the method used, is a subtracted "derived" response similar to a loudspeaker crossover. 6 dB/octave single-order filters have symmetrical slopes. Multi-order derived filters have one slope equal to the filter's. The other subtracted derived slope is always 6 dB/octave and mildly peaks. The table below shows various EEQ methods and their slopes.

Image
Elliptical Equalizer Comparison. Neumann EE-70/EE-77, VAB-84, LR±S Width, MS Mid Side, MS Mid Side with Correction Allpass.

For an example of response peaking see: http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=828&start=6

In November 2016 I discovered that by adding a corrective allpass filter to the Mid path to compensate for the phase response in Side I could achieve a fully-symmetrical second-order 12dB/octave EEQ without response peaking. The "advanced" EEQ's characteristics are the last table entry. The advantages are two-fold:

1) The second-order EEQ frequency can be set lower because the steeper filter controls vertical modulation more effectively. (See: http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=828&start=23)
2) The steeper 12dB/octave crosstalk slope, combined with a potentially-lower vertical crossover, has a far less impact on midrange stereo imaging.

After being asked by a couple of customers to make a board I did.

This is the completed Elliptic Equalizer PC board.

The EEQ board has an optional on-board MS codec or it can be jumpered to receive MS inputs directly as an insert to the Precision MS Matrix. In the later configuration only the 4 dual op amps are used. The rest of the "stuff" is balanced I/O and MS.

Image
Second-order Elliptic Equalizer First Build

The EEQ can be switched for second-order or first order vertical/crosstalk responses of 12/12dB (corrected) 6/6dB (EE-77 emulation) or 12/6dB brickwall-friendly*. (*See: http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=828&start=32)

12 dB/octave Crosstalk Curves for 75, 150, 300 Hz

Left-only input.
Top trace is left out.
Bottom traces are the right channel output with the EEQ switched from 75 to 150 and 300 Hz.

Image
Second-order Elliptic Equalizer Crosstalk Curves for 75, 150, 300 Hz

Comparison of 6 dB vs. 12 dB Per Octave Lateral Versus Vertical Response

Left-only input. EEQ set at 300 Hz.
Top trace is lateral (mono) response.
Bottom traces are the vertical output with the EEQ plotted at 6 and 12 dB per octave.

Image
Second-order Elliptic Equalizer Lateral vs Vertical at 300 Hz with 6 or 12 dB per octave slope

Comparison of 6 dB vs. 12 dB Per Octave Crosstalk Response

Left-only input. EE set at 300 Hz.
Top trace is left out.
Bottom traces are the right channel crosstalk with the EEQ plotted at 6 dB and 12 dB per octave

Image
Second-order Elliptic Equalizer Crosstalk Comparison at 300 Hz with 6 or 12 dB per octave slope

The final curve (above) shows the vast reduction in crosstalk of the 12dB/octave slope.

You can audibly compare the crosstalk differences here: http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=828&start=45

This is a block diagram of the EEQ

Image
Elliptic Equalizer Block Diagram

The schematics

Image
A Second-Order 12 dB/Octave Elliptic Equalizer with Allpass Delay Correction in Mid.

Larger image: http://www.proaudiodesignforum.com/images/png/Elliptic_EQ_Sheet_1.png

Circuit Description

IC1 and IC2 provide balanced inputs for the MS encoder formed by IC3 and IC4.
J1 and J2 bypass the MS encoder to accept external single-ended inputs which are already MS encoded.
The external MS configuration is useful when the EE is incorporated into an existing MS PC board at insert point F: http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=7&t=262#p2918
Using an existing MS board provides the advantage of having the existing MS inserts ahead of the EE and one less MS encode/decode.
The EE is bypassed "hardwire" by RY1 and RY2.
For total bypass the EEQ should be used with an insert switcher: http://www.ka-electronics.com/kaelectronics/Insert_Switcher/Insert_Switcher.htm

Side High Pass Filter

Op amps A, B and C provide a single-order or second order high pass filter for the Side channel.
RY3 switches the filter from 12 dB/octave to 6 dB/octave.

Op amp A is a unity-gain voltage follower to buffer the filter input.
Maintaining precise unity gain within the filter passband is important to maintaining separation.
Very small gain errors cause very large changes in crosstalk above the EE frequency.

R1 provides bias for op amp A.
R2, and similar resistors R4, R6 and R14 limit inverting input current for NPN op amps such as the NE5532 or LME49720 that have back-to-back input protection diodes.
With some op amp types these may be reduced to 0Ω and with others having them in place can also provide some distortion cancellation.
With the NE5532, NJM2114 or LME40720 they are essential to prevent latch-up.

Op amp B is the first stage high pass filter for the Side channel.
For second-order EE filtering the HP network is provided by C3 and R3 plus R10 and R11.
When the EE is switched to single-order, the parallel combination of C1||C2 in series with C3 shift the cutoff frequency.

Op amp C and associated components provide a second, cascaded first order stage.

Side Filter Tuning

The filter frequencies can be tuned with either on-board relays or an external 4 pole switch.
In most cases three EE frequencies selected by relays is sufficient.
Four pole switches can be very costly: A 24 position 4 pole Goldpoint/Elma V24C, unwired, is about $191 US. http://goldpt.com/prices.html
Twelve position switches are similarly-priced. Six positions drops the cost to about $122 US.

(When Modulation and Preview is required relays should be used to avoid an 8 pole switch.)

Rather than use simpler single op amp second-order filters that would have unequal resistor values on each switch deck, a decision was made to use simpler cascaded single-order filters with all four poles of the switch having identical resistor values.

Two simple 6 dB/octave filters are cascaded to form a 12 dB/octave filter.

The EE frequency of the second-order filter is 1.56 times the cutoff of each single-order section.
Each stage of the 75 Hz filter will have an Fc that is 48 Hz. The 150 Hz Fc is 96 Hz; 300 Hz Fc is 192 Hz.

R3 and R5 limit the maximum frequency.
RY4 and RY5 switch the filter frequency by adding R10-R13 to step the frequency from 75 Hz to 150 and 300 Hz.

When stepped switches are used R10-13and RY4-7 are not installed.
Headers S1A-S1D provide a switch tie point.
DC bias current must be maintained through the off-board resistor string.
The switch common should ground the unused resistors to change tuning.

When the Side filter is set to 6 dB/octave the parallel combination of C1 and C2 are switched in by RY3.
When C1 and C2 in parallel are in series with C3, the effective capacitance is 0.14 uF.
The shifts the cutoff frequency to the actual 75, 150 or 300 Hz.

Mid Allpass Correction Filter

An allpass filter is used to match the phase response of the Side filter to achieve true second-order response of the crosstalk curve and eliminate uncorrected response peaking. (See earlier posts for various response curves.)

Op amp E is a unity gain voltage follower similar to op amp A that is used to buffer the allpass filter input.

Op amps F and G are allpass filter sections tuned identically to the second-order Side filter.
To match the phase response of the Side filter C8+C9 (or C10+C11) must be twice the value of C3 and C4.

RY6 and RY7 switch the allpass filter frequency using an identical resistive and relay network to the Side HP filter.

MS Decoder

The filtered outputs, which are in MS format, are decoded by IC8 and IC9.
MS decoded produces 2L and 2R outputs.
Op amps D and H attenuate the MS decoder to provide unity gain operation.

THAT1646 line drivers buffer the decoded L and R outputs.
Note that the 1646 output polarities are redefined to correct the inversion of op amps D and H.

When the MS decoder is not needed, J3 and J4 can be jumpered to bypass it.
When J3/J4 are used to link the filter outputs to an external MS decoder the THAT1646s must not be installed.

Bypass

RY1 bypasses the Side filter.
RY2 bypasses the Mid allpass filter.
RY3 switches the Side filter output from the second-order output to the first filter stage.

Relay Coil Wiring

Image
A Second-Order 12 dB/Octave Elliptic Equalizer with Allpass Delay Correction in Mid. Sheet 2. Relay Coil Wiring.

Larger image: http://www.proaudiodesignforum.com/images/png/Elliptic_EQ_Relay_Sheet_2.png

I've tried to design a lot of flexibility into the EE's relay switching to permit a lot of front panel options and the ability to optionally switch it to 6 dB/octave.

If the Elliptic EQ is only going to be used at 12 dB/octave RY3 does not need to be installed and can be jumpered in the normally-closed position.
C1 and C2 can also be eliminated.

J5 provides a bypass option when the frequency relays are used.

When in the "A" position, connecting any Frequency control line to relay common will automatically engage the EE filter.
This configuration permits a single pole 3 or 4 position switch to act as both frequency select and bypass.
When a three position switch is used a bypass switch for A/B comparison can be in series with the common.
A four position switch could also be used to provide Off-75-150-300 Hz positions.

When J5 is in the "B" position a separate bypass pushbutton can be used to activate the EE.
If an external rotary switch is used to set the EE frequency J5 should be in the "B" position.

The second block of switch connections, labeled "Sum" and "Difference" are optional SPDT switch connections to alter filter slopes to 6 dB/octave.
If these are not used J6 and J7 are installed.

If a stepped frequency switch is used RY4-RY7 are not installed.

Relay Power and Indicators

The EE will have LED outputs to show actuation of RY1, RY2 and RY3.
Resistors R30-R33 provide current limiting.

When the EE is configured to operate at only 12 dB/octave only the "EE In" LED is used.

If switches are installed to permit 6 dB/octave operation the Difference and Sum LED outputs may be useful.

Relay power for the EE is galvanically-isolated on the board from audio ground.
External coil power may be unregulated or regulated.
The on-board 24V regulator has a separate input connection.
Jumper J8 connects the 7824 regulator to the Relay Supply connection.
When the on-board regulator is used the relay supply connections become 24V outputs.

Other Uses

Bass-to-Mono "Focus" for headphone listening.
Vertical warp reduction for vinyl transfer.

Project Status

The first PC board run was four units three of which have already sold. The fourth board is assembled and will be available for sale soon. My intent is to offer both assembled and bare boards. If you're interested in a bare board when they become available PM me.

Best;

Wayne

User avatar
petrolio
Posts: 36
Joined: Sun Dec 29, 2013 5:35 pm

Re: A Second-Order 12dB/Octave Elliptic Equalizer

Post: # 47962Unread post petrolio
Sat Sep 30, 2017 4:52 am

Wow Wayne, great job!!!
Need to study for a while...:)
Thanks a lot!!

Inviato dal mio ALE-L21 utilizzando Tapatalk

User avatar
Wayne Kirkwood
Posts: 41
Joined: Tue Jul 21, 2015 12:43 pm
Location: Dallas, Texas
Contact:

Re: A Second-Order 12dB/Octave Elliptic Equalizer

Post: # 47965Unread post Wayne Kirkwood
Sat Sep 30, 2017 9:48 am

Thanks petrolio!

I noticed that the updated drawing's component designators didn't quite match the circuit descriptions.
I missed the edit window to correct them.

The Mid allpass filter is made up of IC5A/B and IC7A.
The Side HP filter is IC6A/B and IC8A.
The MS decoder is IC9 and IC10.
IC7B and IC8B are -6dB attenuators.

Post Reply