Passive 6dB EQ

[ruffrecords]

I managed to persuade LTspice to plot a series of graphs for the bass boost/cut. The attached pic shows 2dB steps and I added one further plot with the lower resistor set to zero so you can see the hi pass filter effect this EQ can produce.

Looking at the boost, the mid point seems to be around 200Hz. The cut frequency mid point  looks like be about 80Hz . The bottom yellow trace is the HPF position. This has a -3dB point at about 110Hz.

Cheers

Ian

 

[ruffrecords]

And a similar set of curves for the high boost/cut:

Cheers

Ian

 

[ruffrecords]

And the mid, in this case centred on 800Hz.

Cheers

Ian

 

[joaquins]

Funny thing about the mids, I did mine with an analog computer and a digital measurement instrument, something like an inductor, a cap, a pot and smaart running on my computer... It showed a noticeable shift in freq, I don't remember which way, nor why and this is not the moment to analyse it.

The other thing is about the Q, how to define Q could be a problem for instance, since it's defined for a filter with it's -3dB points, in a 6dB bell it will be quite deformed, in a 2dB bell it won't exist. You could use the Q of the LC filter used to create it but you don't count the degradation by the resistance of the control. Constant Q or proportional Q? etc. for all kind of known problems. Also symmetrical or mirrored Q is quite a discussion, which shows the difference from boost to cut.

...

The only way to avoid that is to have separate switches and inductors for boost and cut so that the Q remains the same. This is what I did in the REDD EQ but it is much more complex and needs two pole switches which are expensive. A possible solution would be to have separate boost and cut controls but that doubles up the number of switches and inductors
...

Ian

I don't know why two poles switches are required, I think it might be a way with single poles doble LC, at least meaning some simplification/aproximation of the circuit. With doble poles and a single LC circuit if you set different impedance networks in series, one for boost one for cut and you use the same LC for one network or the other maybe you can get away with it... I should put that in paper, too late in the night for that right now.

Regards,

JS

 

[ruffrecords]

I think the only sensible way to define the Q is at maximum boost/cut. Strctly speaking it is wL/R and in any passive EQ R is a variable.

I cannot see a way of using a single pole switch and two sets of LC, one for boost and one for cut, to keep Q the same. In the REDD EQ where I wanted Q to be the same I used a 2 pole switch. If you can think of a way to achieve this with a single pole switch that would be great.

Here is a link to the REDD EQ schematic:

https://drive.google.com/file/d/0B_n67A1hN3qtMVFMOXVzdWdDSXc/view?usp=sharing

Cheers

Ian

 

[ruffrecords]

I have been doing some more work on the 10dB mastering version. One problem with all passive EQs is that they need to be fed from a low source impedance. If this is not the case then the curves become distorted. The effect is to reduce the size of the boost peaks and increase the depth of the cut peaks. In the REDD EQ design, I partially overcame this problem by including a series resistor in the input to the EQ and deliberately designing the EQ curves to be correct when this resistance was equal to 2500 ohms.  The actual values of resistor could be adjusted to take into account the winding resistance of any 10K:10K input transformer plus the referred impedance of the driver.

I plan to do the same thing for this EQ. By making the series resistor adjustable it means the EQ can be set to compensate for the actual driving impedance. If two channels of EQ are used in a stereo set up it means they should be able to be more closely  matched to each other.

The initial design will be for a 600 ohm input. Here the nominal value of the series resistor will be 150 ohms. This should be sufficient to compensate for the actual output impedance of most DtoAs and 600:600 transformers There are other reasons for doing the 600 ohm version design first.

1. Inductance values are smaller which means stray capacitance is lower (this has occasionally been a problem in some of my 10K EQs causing boost  frequency inaccuracies  at high boost frequencies).

2. Capacitor values are larger which also means that stray capacitance , both inside inductors and from switch wiring, is a much smaller influence on actual boost/cut frequency.

3. Stray hum pick up should be lower especially is the power supply is in the same enclosure as the EQ. However, if possible, I plan to make provision for the use of screened inductors.

To provide constructors with the maximum flexibility, there will be no switches or pots on the PCB. 3 pin Molex connectors will be provided for connection to pots/stepped switches and 16 way headers for connection of frequency selection switches. Since there will be four bands, that means there will be four 3 way Molex connector and four 16 way headers on the PCB. Input and output will be on two pin Molex connectors. At this stage I do not plan to include and EQ bypass as I expect in most mastering situation a relay bypass  of the complete unit will be present. As I did on the universal EQ, I plan to make provision for a a pair of capacitors per frequency so people can more easily customise the frequencies to their own needs.

Next step is to calculate all the RLC values, select some inductors and start the PCB layout.

Cheers

Ian

 

[ruffrecords]

Just been thinking again about the frequency ranges for the two mid bands. I looked at the Pultec MEQ5. This has a low mid band that extends from 200Hz to 1KHz and a high mid that spans 1.5KHz to 5K. This ties in quite well with the earlier suggestion of centring the low mid on 300Hz  and spanning from 75Hz to 1.2KKz and the high mid band being centred on 3KHz and spanning from 750Hz to 12KHz. So I am going to work out some LC values based on a 12 way switch for each of the two bands.

Cheers

Ian

 

[PRR]

> problem with all passive EQs is that they need to be fed from a low source impedance.

Not strictly true. For any loZ->hiZ circuit, there is a dual in hiZ->loZ form.

However as your load is already hi Z, and hiZ->loZ is unusual thus mind-bending, I agree you want to design from low Z (or any medium-Z compromise you can find).

 

[ruffrecords]

I have  been thinking about the actual frequencies for the lo mid and hi mid bands. We have pretty much decided the low mid will be centred at 300Hz and span two octaves above and two below (75Hz to 1200Hz) and the high mid will be centred on 3KHz and similarly span two octaves either side (750Hz to 12KHz). The next question is how to divide these ranges into a set of frequencies. The Helios seems to use a ratio of close to root 2 (e.g. 1K, 1K4, 2K, 2K8) but that would only give use 9 frequencies and given we will be using 12 way switches it seems a pity not to have at least 11 frequencies in each band. So, to keep it musical it tried an interval of a perfect fourth which gives the following eleven  frequencies (rounded to the nearest ten Hz) for the low mid:

75, 100, 130, 180, 240, 300, 400, 530, 700, 940, 1200

And for the hi mid we get:

750, 1K, 1K3, 1K8, 2K4, 3K, 4K, 5K3, 7K, 9K4, 12K

Does that seem reasonable/useful?

Cheers

Ian

 

[ruffrecords]

I have also been looking at the interaction between the four bands and it does not seem to be as bad as I first thought. Here is a set of sweeps of all four bands. Note that the boost./cut is the same for each band for the sweep.

Cheers

Ian

 

[ruffrecords]

And here is a close interaction between the low mid and high mid bands when set to 1200Hz and 1300Hz respectively.

Cheers

Ian

 

[ruffrecords]

I was looking at the NYD passive equaliser schematic the other day and it occurred to me there might be a way to configure the 6dB EQ so that it uses a single ladder network  rather than one per band. NYD's EQ seems to work in a similar way but the ladder is not complete and it needs  a specific load to work. If we build a full ladder spanning the whole range of boost and cut we can just use a rotary switch to connect the frequency selective network between the appropriate two points on the ladder.

In implementation terms I envisage one PCB containing the ladder and a number of IDC connectors wired to the ladder, in and out connectors and provision for an EQ in/out switch. Each band would then have its own PCB containing, say, an 11 way switch for EQ level (-10 to +10 in 2dB steps),  another switch to select the frequency and the RLC components of the EQ itself. There would be one PCB for each band up to a maximum of four.

For a really serious mastering EQ you might even go to a 23 way switch for -11dB to +11dB in 1 dB steps. If you want to be very sophisticated and ensure the Q is the same for cut as for boost you could use a 2 pole switch, one for boost and the other for cut. What I like about this is that one basic topology suits a range of applications and pockets.

There will still be interaction between bands but that is true of almost any passive EQ. Next step is to simulate this to make sure it actually works; then I will draw up a schematic and post it here for comment.

Cheers

Ian

 

[ruffrecords]

I just finished simulating and I am pleased to report it works extremely well. I wish I had thought of this a long time ago. It makes stepped EQs so much simpler. No more wiring dozens of resistors across rotary switch contacts. I will draw up the schematic and then start work on PCBs. I have also worked out the ladder values for the 600 ohm version which only uses E24 series resistors yet gives an accuracy of no worse than 0.1dB not including inductor resistance. I also created a simple spread sheet to calculate the L and C values given the frequency and required Q. For the time being I am going to assume that a Q of 2 will be about right.

Cheers

Ian

 

[kambo]

cant wait for the schematic 8)

is it 600r in - 600r out design ?

 

[abbey road d enfer]

I think the only sensible way to define the Q is at maximum boost/cut. Strctly speaking it is wL/R and in any passive EQ R is a variable.

There's no "sensible" way to define Q in an equalizer. Q is valid only for function transfer with asymptotes at minus infinity. The transfer function of an equalizer is a fraction in which the numerator and denominator have a different damping factor.
The idea of defining the selectivity of an EQ by using a supposed Q is ludicrous. It has been and still is the cause of many errors and misconceptions. The only value in Q is that the higher it is, the narrower is the BW.  But Q has no subjective meaning, when an indication of BW (relative or absolute) has significant meaning.

 

[ruffrecords]

I think the only sensible way to define the Q is at maximum boost/cut. Strctly speaking it is wL/R and in any passive EQ R is a variable.

There's no "sensible" way to define Q in an equalizer. Q is valid only for function transfer with asymptotes at minus infinity. The transfer function of an equalizer is a fraction in which the numerator and denominator have a different damping factor.
The idea of defining the selectivity of an EQ by using a supposed Q is ludicrous. It has been and still is the cause of many errors and misconceptions. The only value in Q is that the higher it is, the narrower is the BW.  But Q has no subjective meaning, when an indication of BW (relative or absolute) has significant meaning.

So how would you differentiate quantitatively between two EQs that produce a bell shaped boost of the same amount at the same frequency but whose bode plots are different, other than by publishing the bode plots?

Cheers

Ian

 

[ruffrecords]

cant wait for the schematic 8)

is it 600r in - 600r out design ?

Your wish.....

The first sheet is just the ladder, the input, the output and the EQ in/out switch.

https://drive.google.com/file/d/0B_n67A1hN3qtZmM2NGJqanA5akU/view?usp=sharing

The second sheet shows the arrangement for each of the four bands. At the moment bands 1 and 2 are assumed to be shelving for bass and treble (but they could be bell shaped) and the low and high mids (bands 3 and 4) are Bell shaped.

https://drive.google.com/file/d/0B_n67A1hN3qtbjlvaU1FazlDVUE/view?usp=sharing

I have shown the ladder resistor values for the 600 ohm version. It is unbalanced so you will need a 600:600 transformer at the input. As it is passive it needs to be followed by a gain make up amplifier which can include a balanced output. I have not shown any values for the EQ components yet.

Cheers

Ian

 

[kambo]

hey Ian, what exactly is the difference between ur design and NYD design ?
kind a puzzled...
the only thing i noticed is recalculated resistor ladder  :

 

[ruffrecords]

hey Ian, what exactly is the difference between ur design and NYD design ?
kind a puzzled...
the only thing i noticed is recalculated resistor ladder  :

If you look closely at the NYD schematic, you will see the top half of the ladder is not always connected to the bottom half.

Cheers

Ian

 

[abbey road d enfer]

So how would you differentiate quantitatively between two EQs that produce a bell shaped boost of the same amount at the same frequency but whose bode plots are different, other than by publishing the bode plots?

I would either specify a BW, but not in the usual academic way (the -3dB points), which just doesn't work, for the same reasons as Q. Although the AES failed to agree on a common definition, I still think describing width in octaves is the most significant way. My definition is based on the points where 1/4th of the max boost/cut is achieved, i.e. +/+3dB for 12dB max.
It has meaning for users. It doesn't have a direct mathematic reference to any calculation, but I know that EQ's designed from a hypothetic Q value end up being awfully wrong in terms of results.
This way of defining the turnover frequencies has the added advantage is that it works also for shelving EQ's or higher order EQ's, where the notion of Q or of a single Q would be definitely wrong.