How to extend the high pass filter below 20hz in DCX2496

The DCX2496 only allows for corner frequencies down to 20hz, but many folks could use a high pass feature below that point. This is easily accomplished using a combination of a 20hz high pass filter and a negative gain high shelf filter.

Here is how it works illustrated in WinISD (divide the frequency scale by 10).

The green line is the target curve--a 2nd order high pass filter at 10hz.

Step 1: The red line is a 2nd order high pass filter at 20hz. Enter this in the DCX.

Step 2: The yellow line is a 12db/octave high shelf filter with negative gain of -12db also at 20hz. Enter this in the DCX.

Step 3: You're done. When these two filters are combined, they produce the target curve.

If you need a different high pass point, just adjust how many db of gain are in the shelf filter. For example, a setting of -6db should provide a high pass filter at around 14hz.



Step 1 detail.



Step 2 detail.


Edited by LTD02 - 3/4/13 at 1:59am

"I suspect LTD02 meant to use positive Low Shelf Filter because that's what needed to boost output below 20Hz to counteract early roll-off from the 20Hz HPF, just like LT (yellow) curve in his simulation."

the *low* shelf filter is set by the low corner. that is the trick.

in this case, it would need to be 10hz, but corner settings below 20hz are not available.

the *high* shelf filter is set by the high corner.

in this case, it would need to be 20hz, and that setting is possible.

that is why a high shelf filter with negative gain is used instead of a low shelf filter with positive gain.
Edited by LTD02 - 3/4/13 at 4:21pm

"LTD, thanks, I was never entirely clear on how to do this either."

np. i have heard of folks using various methods to shape a signal below 20hz using the dcx, i just never saw the high pass application detailed explicity.

"It might be just a function of the inherent LF rolloff of the electronics, but do we even know for sure that the boost persists below 20 Hz?"

i asked about that previously and a few people confirmed that the filters do continue to work below 20hz.

there are people fudging parametric eq below 20hz with multiple filters above 20hz (i believe ricci has done this) and it would be odd if filters such as simple high pass filters stopped working below 20hz.
Edited by LTD02 - 3/4/13 at 4:22pm

here are the two filters:

(again, divide the frequency scale by 10)

1) low pass filter 12db/octave, +12db positive gain, corner set to 10hz (blue line).

2) high pass filter 12db/octave, -12db negative gain, corner set to 20hz. (yellow line).

in this case, 100db is the arbitrary "zero point". the low shelf with positive gain filter increases up to 112db while the high shelf with negative gain drops to 88db.

the revelation was that these two filters are exactly the same *shape* and shaping the signal is what we are trying to do. the level of the signal can easily be changed if necessary with gain.

since filter #2 only requires a 20hz corner setting, it is possible to use it. filter #1 requires a 10hz corner, so it is not possible. the difference is because the dcx uses the lower corner of a low pass filter to set the frequency/locate the filter, while the high pass filter uses the upper corner.

"I am officially confused."
"I do not understand how combining 1 and 3 would."

zheka, just for the sake of experimentation, change the gain to +/-12db so it will be easier to see the corners. think about what is going on in post #10, example 1. the signal is flat across all frequencies but reduced in magnitude from about 60hz up (in your example). below that point, the level begins to rise (just as it would with a low shelf with positive gain) up to whatever plateau you have set (6db, 12db or whatever). so you are *shaping* the signal in the same way.

if you use the low pass filter with positive gain and set it at 20hz, most all of the "gain" will have already occured by the time you get down to 20hz.

if you use the high pass filter with negative gain and set it at 20hz, most all of the "gain" gets stuffed under 20hz. that is the trick.

"This is as close as I could get, using this method on that DSP. Close enough..."

bassthathz, it appears that itech dsp uses the front corner for the low pass shelf, so yours won't require using negative gain on the high shelf.

just for the sake of experimentation, try these changes.

1. change the high pass frequency from 275hz to 200hz.
2. on eq filter1 change the gain from 17.8 to 12.0.

the result should be a net 100hz high pass filter, 12db/oct slope, which has a -3db point at 100hz.

of course, for the 10hz filter, just divide all the frequencies by 10.

....

edit: after eyeballing it a little bit more, it appears the itech dsp uses the mid-point frequency to locate the shelf filter. let me see if i can cook up something that works with that.

bassthathz,

if i am eyeballing it correctly, the itech dsp locates the shelf based on the midpoint of the slope. this is different from the dcx.

based on that, the best that you can do, in theory, is a 14hz high pass filter using the following configuration.

1. 12db/oct low pass shelf with +12db gain set at 20hz (the yellow line).
2. 12db/oct high pass filter set at 280hz (the red line).
3. the net is equivalent to a 12db/oct high pass at 14hz (the green line).



however, as you point out, it doesn't have to be perfect. so one method that gets you within 1.5db of a theoretically perfect 10hz high pass is:

1a. 12db/oct low pass shelf with +12db gain set at 20hz
1b. parametric eq, centered at 20hz, q=1.00, gain= -3.0db (the yellow line).
2. 12db/oct high pass filter set at 20hz (the red line).
3. the net is equivalent to a 12db/oct high pass at about 9hz with a smidge of boost before rolloff (the green line).

with model, we are using a parametric eq to beat down the front end of the shelf filter. it ends up being pretty close.


Edited by LTD02 - 3/4/13 at 5:17pm

"Are there any other downsides to this approach?"

the downside is that headroom/signal to noise ratio are lowered by the amount of gain used in the shelf filter. with a 10hz highpass setup, that is 12db. for the 14hz highpass setup, it is 6db. as looney mentioned, it is a "workaround" (i like that term), not an ideal solution. an ideal highpass 10hz would take up no headroom nor would it reduce the signal to noise ratio at all.

"Interesting. I wish there was a way to model this in WinISD"

of course there is or i wouldn't have been able to do it. :-)

the easiest way to get a feel for how to create a shelf filter in winisd is to:

first we need a flat line to model on.
1. plop some driver in a sealed enclosure, it doesn't matter which driver or which enclosure.
2. set the inductance to zero. this will flatten out the top end.
3. put on a linkwitz transform with the lower end set to 1hz.

now you should have a flat line to work with.

to model a shelf filter, add a second linkwitz transform, but this time set f0 to 200hz and fp to 100hz.
ensure that the q0 and qp are both about 0.707

bang, now you should have the shelf.

to create a high shelf, just swapt he 100 and 200hz frequency values.

other shelf filters can be created by changing the f0 and fp.

once you have the shelf filter modeled appropriately using the flat line model, you can just add that to whatever driver/enclosure model you are working with to see the net effect.

also, i'm not sure if headroom is actually lost. since the combination of boosting and cutting will never exceed the level of the original signal, i'm thinking that no headroom will be lost.

not sure about that one though.

i don't know.