Parallel High Pass and Phase change

Going through an incredibly busy patch at work and I haven't had any bench time to sit down and do the work / sim this question,  I'm floating here but in the knowledge that I have work to do.

I'd like to blend a high passed signal with the original unprocessed signal in the analog domain (audio frequencies only).  Obviously the phase change induced by the high pass will be an issue for the summation.  As I see it I have two options

1) "Correct" the phase of the high passed signal (not seeing an easy way to do this)
2) Match the phase response of the previously unprocessed signal to the high passed using an all pass filter

Ideally my high pass would be 6 dB/Oct but this only induces a 90 degree phase shift which is hard to match with an all pass which will give 180 degrees.  That may limit me to a 2nd order high pass.

Any obvious strategies that I'm missing?

Is a linear phase high pass too noticeable??? Interesting....I've never considered this problem.....

Wonder if it's the same as parallel mixing a transformer bandwidth limited unit with one that isn't....

scott2000 said:

Is a linear phase high pass too noticeable??? Interesting....I've never considered this problem.....

Click to expand...

In digital yes, absolutely IMO, it's a mess.  But my question was about analog/

Wonder if it's the same as parallel mixing a transformer bandwidth limited unit with one that isn't....

Click to expand...

Mixing any two correlated signals with different phase relationships will result in frequency response deviations. it may or may not be an issue in practice.

Hi Ruairi,

I am not sure why you want to do this. Superposition tells us mixing these two signal together simply changes the HPF into a shelving cut.

Cheers

Ian

Hard to do

In stead of blending with original, could you blend with (original-highpassed), i.e. With the "missing stuff"?

That way you blend between the hpf'ed and the original
This is probably the only way that recombines even remotely well, like for multibandlimiters...

Jakob E.

I guess I have the same question as Ian. Why? Since you said "at work" I'm assuming it's in a mastering context and not a situation where you have no choice.

Proper engineer responses all round 

In digital I am using a plug in that blends a linear phase hi passed signal back in with the dry signal, it creates something of a high shelf effect.  The trick is that the split / high passed signal is being compressed, and the side chain is fed from the split / hi passed signal.  This creates a nice effect where the hi shelf ducks when there is too much high end energy.

The result is that quiet and delicate sections get  little high end lift for detail but brash / louder section don't get too hot.

I'd like to emulate that in the analog world for various reasons and obviously a linear phase high pass is impossible in analog so I am left with my original problem.  I think a THAT RMS style VCA timing setup would really suit this task.

ruairioflaherty said:

Proper engineer responses all round 

In digital I am using a plug in that blends a linear phase hi passed signal back in with the dry signal, it creates something of a high shelf effect.  The trick is that the split / high passed signal is being compressed, and the side chain is fed from the split / hi passed signal.  This creates a nice effect where the hi shelf ducks when there is too much high end energy.

The result is that quiet and delicate sections get  little high end lift for detail but brash / louder section don't get too hot.

I'd like to emulate that in the analog world for various reasons and obviously a linear phase high pass is impossible in analog so I am left with my original problem.  I think a THAT RMS style VCA timing setup would really suit this task.

Click to expand...

That is what we used to use Dolby A for.

Cheers

Ian

scott2000 said:

Is a linear phase high pass too noticeable??? Interesting....I've never considered this problem.....

Click to expand...

Scott,

I should have been clearer earlier.  For typical high pass duties between 10 and 50 Hz as used in mastering a Linear Phase EQ is a disaster, a big smeary mess.

In fact, I'm in the minority in that I can't really stand linear phase EQ in general and I've tried all the best ones including the Weiss hardware.  The pre-echo is audible, unnatural to my ear.  I could hear that before I even understood the concepts.

My use in this case is specific to the high end which is just about acceptable IMO but I'd love to come up with an analog version.

I've linked this here before but this video illustrates the pre-ringing issue pretty well, exaggerated cases but you will hear the same effects in normal everyday use.

https://www.youtube.com/watch?v=efKabAQQsPQ

Wow....a different thought process here in my mind when "mixing" a high-passed signal with "program".  I think this is wandering waaaay off the path for this thread, but WTF...

It begins in the mid/late 1970's:

http://brianroth.com/articles/roth-aphex-review.pdf

Early 1980's, I built a Gizmo dubbed the "AirFX" in a SCAMP rack module at the studio where I was working.

It was a two pole Butterworth high-pass with selectable corner frequencies, and was patched into the desk as a send/receive device...like a reverb....like the original Aphex.

It was popular in my studio as a "brightener".....when used in small dosages.

Two pole filter vs. the normal single pole EQ response on the top end.

But.....    Nevermind!  lol

Bri

  Why not a dynamic EQ? So you don't have the recombination issue. With a bit more effort in the dynamic processor sidechain you can archive a similar curve, limiting the minimum gain or maximum attenuation.

  Take the sontec for instance, add the dynamic processing at the output of the high shelving filter, all you was thinking for it plus a limit for the gain range. A variable DC reference and a diode would make that adjustable.

  Also, in mastering, processing are pretty subtle in general, so changes are small compared to other situations. That helps to predict the circuit a little better as it's operating in a small range, or make some approximations thanks to that. I mean, you are probably making this for few Db range, so the crossing freq moves a little but if you would been doing this for several dB the crossing freq would jump quite a lot making the phase shift even harder to fight. I think taking this into consideration can simplify the problem a bit.

JS

That is what we used to use Dolby A for

Click to expand...

. 

I worked at a studio in the 70’s and an engineer wired a Dolby M 16/M8 rack with switches to have  the low bands  decode but the high band not decode (of the cat 22 cards)  to do the smooth high frequency compression trick as used on acoustic guitars on super tramp albums and Elton vocals and such.  All these mini toggle switches.  In believable amount of wiring for 24 channels. 

Hey Brian I remember seeing you article in the magazine .  We had one of those modules.  Anything to get topend into the mix was used those days.  Now I’m constantly trying smooth topend from attacking my ears.  Also curt ‘the Afex inventor,  said the phase shift from the inductive coil  of the mag cartridge was important to the sound of the hi frequency sound that Afex  provided.  Ruairy is trying  to have no phase shift.  I think you got away with lots of things on tape due to lagging phase response of hi freq.  It would turn to mud if you weren’t constantly getting your high frequently right.  That said the eq curves of tape with all the processing was Anything but coherent but did have a sweet quality to my ears.

this video illustrates the pre-ringing issue pretty well,

Click to expand...

.

That really helps to understand that problem.    Good explanation.  Heard that but just changed plugins thinking the plugin sucked.  It did for the application

joaquins said:

  Why not a dynamic EQ? So you don't have the recombination issue. With a bit more effort in the dynamic processor sidechain you can archive a similar curve, limiting the minimum gain or maximum attenuation.

Click to expand...

Hey Joaquins,

I've tried all of the software dynamic EQs over the years but it doesn't sound the same as this process to my ear, maybe in hardware it would be more effective or sweeter.

This parallel compressed linear phase high pass is definitely not the typical approach for me, I'm a minimalist in general with processing but it's bringing something new to the table.

You are correct that the amounts used are very small, everything in mastering is but the cumulative effect is worthwhile.

ruairioflaherty said:

Obviously the phase change induced by the high pass will be an issue for the summation.

Click to expand...

You've been overthinking this. A 1st-order HP signal is always in the same quadrant as the original signal, so there's no summation issue. As Ian pointed, was used in all the Dolby noise-reduction units.

ruairioflaherty said:

Hey Joaquins,

I've tried all of the software dynamic EQs over the years but it doesn't sound the same as this process to my ear, maybe in hardware it would be more effective or sweeter.

This parallel compressed linear phase high pass is definitely not the typical approach for me, I'm a minimalist in general with processing but it's bringing something new to the table.

You are correct that the amounts used are very small, everything in mastering is but the cumulative effect is worthwhile.

Click to expand...

  Maybe you don't have the same controls on the dynamic EQs you tried and the compressor you are using on the parallel filter. When you do parallel compression (full band or filtered) it results in a maximum gain reduction, when it get's to the same level as the parallel uncompressed signal. The Brainworx dynamic EQ does provide all the necesary controls do do something like it but the curve might be too rough, when you are aproaching the original level (high parallel compression) the curve smooth out softly, if this does it rough, compressing linearly till it hits the maximum gain, it might sound different.
  Building this in hardware or programming your own dynamic EQ would give you the control over all of this details and you might get what you are looking for.

  While 90º doesn't bring summation problems it does have a phase shift and that could be pretty noticeable when you shift in freq the phase shift (no pun intended). I don't know how you could avoid this effect but I see how this could be a problem, it might still be there with a dynamic EQ.

JS

abbey road d enfer said:

You've been overthinking this. A 1st-order HP signal is always in the same quadrant as the original signal, so there's no summation issue. As Ian pointed, was used in all the Dolby noise-reduction units.

Click to expand...

That's the problem with knowing a little bit!  A beginner would have just done it and liked the sound, a pro would know the answer and there I was stuck in the middle

Thanks you

joaquins said:

  Maybe you don't have the same controls on the dynamic EQs you tried and the compressor you are using on the parallel filter. When you do parallel compression (full band or filtered) it results in a maximum gain reduction, when it get's to the same level as the parallel uncompressed signal. The Brainworx dynamic EQ does provide all the necesary controls do do something like it but the curve might be too rough, when you are aproaching the original level (high parallel compression) the curve smooth out softly, if this does it rough, compressing linearly till it hits the maximum gain, it might sound different.
  Building this in hardware or programming your own dynamic EQ would give you the control over all of this details and you might get what you are looking for.

  While 90º doesn't bring summation problems it does have a phase shift and that could be pretty noticeable when you shift in freq the phase shift (no pun intended). I don't know how you could avoid this effect but I see how this could be a problem, it might still be there with a dynamic EQ.

JS

Click to expand...

I have the Brainworx dynamic EQ, and the Fabfilter and bunch of others but they are not giving the same effect as this parallel compressed high pass. On the few masters I've used it on the parallel high passed signal is being blended in at -20dB or so, it's subtle but gives a nice consistent sheen that is then compressed with fast attack/medium release if the source gets too loud or bright.  Apparently its a common trick for Weiss DS-1 users but I don't get along with that hardware or the new plug in.

Perhaps it's the linear phase nature of the high pass but even emulating that in Fabfilter does not give the same results.  I stopped using the Brainwork a long time ago, good controls but I didn't like the sound.

I'll do some more experimentation in the next few days.

Brian Roth said:

Two pole filter vs. the normal single pole EQ response on the top end.

Click to expand...

Exactly! So the end result is that the HF lift is preceded with a slight cut (due to second quadrant phase-shift).
It's a known trick; when you want to make a track stand out, you enhance its characteristics frequencies, but you also need to "clean up" the surroundings, in particular the octaves just below that would otherwise have a masking effect.

Early 1980's, I built a Gizmo dubbed the "AirFX" in a SCAMP rack module at the studio where I was working.

Click to expand...

I also had a product out in the early 80's that was just doing that; I was a little ashamed of selling for big money (3U rackmount unit with 2 big SIFAM VU-meters and a PCB the size of a credit card) something that could be duplicated with a half-decent parametric EQ and a little savvy.