Does a high-pass filter lower the needed power for a speaker?

[samsa]

Another thing to consider, when applying a (passive) high-pass filter is the effect on the frequency response of the speakers.

Case in point: I have a pair of KEF LS50s, accompanied by a subwoofer. Since the low frequencies are handled by the sub, I thought it would be a great idea to pass the speaker inputs through a high pass filter, instead of driving them full range.

So I put together a 2nd order high pass filter with f_c = 80Hz.

With a purely resistive load, this would have done what I wanted: flat response above ~100Hz or so, dropping by 12dB/octave, below 80 Hz.

Alas, the impedance of the LS50s varies wildly, both in magnitude and phase, below 300 Hz. So, when put in series with this filter, instead of the smooth response curve I was expecting, the response below a few hundred Hz varied wildly. This was verifiable with in-room measurements: the frequency response had peaks and troughs which were smoothed out when the filter was removed.

Moral: without knowing the impedance characteristic of your speakers, it is hard (impossible) to design a (passive) high pass filter which will do what you want.

 

[Head_Unit]

Moral: without knowing the impedance characteristic of your speakers, it is hard (impossible) to design a (passive) high pass filter which will do what you want.

Yup!!! As posted elsewhere, I had to redesign a vaguely similar product for the giant company I worked for, nameless to protect the guilty. I quickly realized the same thing-it just doesn't work. Theoretically you could compensate the impedance of the subwoofer and the satellites...but boy would it be EXPENSIVE!! And even if you "knew" the impedance curve of the speaker, it will actually vary somewhat with the weather and aging and the music signal.
--> Passive subwoofer filters (for the sub OR satellites) do not really work.

 

[Head_Unit]

...low passed signal and high passed signals have voltage swings higher than the original!...

What happens if that square wave is lowpass filtered at say 22 kHz? Do you a way to do that? I'm curious if those spikes would reduce.

Ugh, those spikes remind of when in engineering school lab, I decided to use a large transformer I had on-hand to make a DC-DC converter. Fed by an SCR. Which chops off the AC wave like a guilloutine. Which instantaneous change induces a HUGE current spike which slams every breaker to the lab...oopsie...

 

[RayDunzl]

What happens if that square wave is lowpass filtered at say 22 kHz?

80Hz square at 44.1k
80Hz square at 44.1k low passed at 22kHz
the difference between the original and the low passed version (scale changed to show detail)

 

[RayDunzl]

Ugh, those spikes remind of when in engineering school lab, I decided to use a large transformer I had on-hand to make a DC-DC converter. Fed by an SCR. Which chops off the AC wave like a guilloutine. Which instantaneous change induces a HUGE current spike which slams every breaker to the lab...oopsie...

In an early computer class I instructed the university's IBM370 to take my 0.25 seconds of CPU time and print periods just to see how many it would print on the lineprinter. My early speed test.

Soon there was a BAM BAM BAM BAM BAM BAM BAM BAM BAM coming from the 132 column printer that usually went ZIP ZAP BLAT RIP ZZT as the character drum turned and the hammers hit various letters, whereas I had it hammering out a whole row at a time.

It was pretty loud, got everyone's attention.

Something like 580,000 periods eventually printed (that's 4,393 BAMs). At (maybe) 1400 lines per minute that's more than three minutes. Seemed like forever.

The guy at the output window said "Don't do that again."

 

[samsa]

Yup!!! As posted elsewhere, I had to redesign a vaguely similar product for the giant company I worked for, nameless to protect the guilty. I quickly realized the same thing-it just doesn't work. Theoretically you could compensate the impedance of the subwoofer and the satellites...but boy would it be EXPENSIVE!! And even if you "knew" the impedance curve of the speaker, it will actually vary somewhat with the weather and aging and the music signal.
--> Passive subwoofer filters (for the sub OR satellites) do not really work.

Just for fun, let me post a graph. What I'm plotting is the ratio of the voltage at the speaker terminals over the voltage output of the amplifier, versus frequency.

The blue curve is the "expected" result, for a purely resistive load. What a nice-looking high-pass filter! In orange is a set of points obtained by reading off John Atkinson's published measurements of the LS50's impedance.

Gruesome, eh?

This is just the theory. The measured results weren't quite as bad, but they were still pretty awful.

 

[SpaceMonkey]

So, does it make sense to have a high-pass filter in pre-amp at around 40-45Hz if speakers can do around 50Hz at -3db?

 

[NTK]

What happens if that square wave is lowpass filtered at say 22 kHz? Do you a way to do that? I'm curious if those spikes would reduce. ...

Here are the waveforms after (and before) the HP filter. I filtered it two ways. Blue is linear phase sinc filter. What you see are the spikes from the classical Gibbs phenomenon. (You can also see the time delay caused by the length of sinc waveform I used in the convolution.) Orange is 4th order Butterworth (minimum phase), and it produced slightly higher peaks than the linear phase sinc filter.

Again, this example shows the worst case. Square waves have a crest factor of 0 dB (i.e. the peak voltage equals to the RMS voltage). Anything you do to them can only increase the crest factor (i.e. make the peak voltage higher than the RMS).

 

[Destination: Moon]

Another thing to consider, when applying a (passive) high-pass filter is the effect on the frequency response of the speakers.

Case in point: I have a pair of KEF LS50s, accompanied by a subwoofer. Since the low frequencies are handled by the sub, I thought it would be a great idea to pass the speaker inputs through a high pass filter, instead of driving them full range.

So I put together a 2nd order high pass filter with f_c = 80Hz.
View attachment 67263

With a purely resistive load, this would have done what I wanted: flat response above ~100Hz or so, dropping by 12dB/octave, below 80 Hz.

Alas, the impedance of the LS50s varies wildly, both in magnitude and phase, below 300 Hz. So, when put in series with this filter, instead of the smooth response curve I was expecting, the response below a few hundred Hz varied wildly. This was verifiable with in-room measurements: the frequency response had peaks and troughs which were smoothed out when the filter was removed.

Moral: without knowing the impedance characteristic of your speakers, it is hard (impossible) to design a (passive) high pass filter which will do what you want.

Forgive me if this is a stupid question but does this argue for installing the high pass filtering on the line level signal to off load bookshelf speakers when using a sub; or just passing the full range signal along to the bookshelf and let them roll off naturally?

 

[pjug]

Another thing to consider, when applying a (passive) high-pass filter is the effect on the frequency response of the speakers.

Case in point: I have a pair of KEF LS50s, accompanied by a subwoofer. Since the low frequencies are handled by the sub, I thought it would be a great idea to pass the speaker inputs through a high pass filter, instead of driving them full range.

So I put together a 2nd order high pass filter with f_c = 80Hz.
View attachment 67263

With a purely resistive load, this would have done what I wanted: flat response above ~100Hz or so, dropping by 12dB/octave, below 80 Hz.

Alas, the impedance of the LS50s varies wildly, both in magnitude and phase, below 300 Hz. So, when put in series with this filter, instead of the smooth response curve I was expecting, the response below a few hundred Hz varied wildly. This was verifiable with in-room measurements: the frequency response had peaks and troughs which were smoothed out when the filter was removed.

Moral: without knowing the impedance characteristic of your speakers, it is hard (impossible) to design a (passive) high pass filter which will do what you want.

VirtuixCAD has a tracing tool. So if you have the speaker measurements (FR and impedance) then you can do traces and import the speaker and see what the filter will do. I tried this and it seemed pretty accurate.
Here is what I get modeling the LS50 and the 80Hz 4-ohm filter from PartsExpress. Check that I didn't mess this up, but not good. Look at the impedance dip. I was warned that these high pass filters can blow up amplifiers!

 

[Destination: Moon]

VirtuixCAD has a tracing tool. So if you have the speaker measurements (FR and impedance) then you can do traces and import the speaker and see what the filter will do. I tried this and it seemed pretty accurate.
Here is what I get modeling the LS50 and the 80Hz 4-ohm filter from PartsExpress. Check that I didn't mess this up, but not good. Look at the impedance dip. I was warned that these high pass filters can blow up amplifiers!
View attachment 120123View attachment 120124

I'm not certain what that graph represents? Is that a prediction model of high level, high pass filtering?
If so, it doesn't really address my question? I do appreciate the reply

 

[pjug]

I'm not certain what that graph represents? Is that a prediction model of high level, high pass filtering?
If so, it doesn't really address my question? I do appreciate the reply

The graph is modeling of the partsexpress LC filter in line with the LS50 speaker, in response to what samsa was discussing. So another going at his example of why it is better to do this pre speaker amp, if possible. Rolling off naturally can work too if your mains woofers have low distortion.

 

[Destination: Moon]

The graph is modeling of the partsexpress LC filter in line with the LS50 speaker, in response to what samsa was discussing. So another going at his example of why it is better to do this pre speaker amp, if possible. Rolling off naturally can work too if your mains woofers have low distortion.

Thanks, I'm not really sure about my speakers actual FR. They're a few years old. To my ears I really like them.
I have a new sub coming and trying to decide how to wire them up. The SVS only has pass through outputs without any filtering. My speakers claim to roll of at 36 hz I think, which seems awfully low for a modest sized bookshelf.
Definitive Tech Studio Monitor 55s

 

[Head_Unit]

I was warned that these high pass filters can blow up amplifiers!

Yes, if you are unlucky you could end up with a kind of short circuit at some frequencies. I've posted elsewhere that when I worked in product planning/engineering for a giant autosound company I inherited one of these passive crossovers. It was supposed to let you run satellites and a bridged subwoofer all off the same amp. Some thought and LEAP simulations showed they just don't really work:
- The complex impedance of the subs and satellites makes a mess instead of a nice filter.
- Woofers have a semi-inductive impedance rise at high frequencies. Put a coil in series, and you *maybe* get just a -3 dB per octave rolloff, which is not enough.

Forgive me if this is a stupid question but does this argue for installing the high pass filtering on the line level signal to off load bookshelf speakers when using a sub; or just passing the full range signal along to the bookshelf and let them roll off naturally?

That depends. If with NO sub you can crank the system to the maximum and the satellites are fine, you can run without the highpass. If like most satellites they strain-or shoot the woofer cones into the next room-you need the highpass filter. And yes it must be BEFORE the power amplifiers.

 

[Destination: Moon]

Yes, if you are unlucky you could end up with a kind of short circuit at some frequencies. I've posted elsewhere that when I worked in product planning/engineering for a giant autosound company I inherited one of these passive crossovers. It was supposed to let you run satellites and a bridged subwoofer all off the same amp. Some thought and LEAP simulations showed they just don't really work:
- The complex impedance of the subs and satellites makes a mess instead of a nice filter.
- Woofers have a semi-inductive impedance rise at high frequencies. Put a coil in series, and you *maybe* get just a -3 dB per octave rolloff, which is not enough.
That depends. If with NO sub you can crank the system to the maximum and the satellites are fine, you can run without the highpass. If like most satellites they strain-or shoot the woofer cones into the next room-you need the highpass filter. And yes it must be BEFORE the power amplifiers.

Are those in line ones from harrison labs acceptable? I was thinking 100 hz hi pass is the right one?

 

[Head_Unit]

Are those in line ones from harrison labs acceptable? I was thinking 100 hz hi pass is the right one?

They are OK, I have some. But the crossover frequency is perhaps -6 dB not -3 dB. At least when I measured some "100 Hz" they were -3 dB at 155 Hz and -6 at 100. This implied to me that inside are two stacked 1st order circuits, EACH -3 dB at 100 Hz.

Since the price is hardly much more I would get the adjustable PFMOD (it says "electronic" but it is still passive) and just not use the lowpass part. Adjustability is very important trying to get crossovers to match in-room. Plus some kind of measurement capability, at least a phone app.
https://www.hlabs.com/products/crossovers/index_files/Page384.htm

An electronic option which someone had once posted sounded clean is the Rolls
https://smile.amazon.com/rolls-Tiny-Stereo-Crossover-SX45/dp/B00102VWJK

 

[Head_Unit]

So, does it make sense to have a high-pass filter in pre-amp at around 40-45Hz if speakers can do around 50Hz at -3db?

- It's not so much how low the satellite speakers' frequency response can go, it's more like at full volume how low can they go without sounding bad.
- The response will also depend on the room.
I guess I would experiment playing at full volume without the sub playing, start at 40 sure and keep going up. When it seems OK, then start at the same frequency with the sub.
- You need a way to flip the polarity (often labeled as "phase" or "0/180"): the correct position should give strong bass and the wrong correction will make a noticeable big cancellation dip.
- You need to play like a pink or brown noise signal, more consistent than music. And measure with at least a phone app.
- If there's no dip either way then both are partly cancelling, and I'd spread the crossover frequencies a bit and check again.
This is why room correction like Audyssey is popular I guess!

 

[Destination: Moon]

- It's not so much how low the satellite speakers' frequency response can go, it's more like at full volume how low can they go without sounding bad.
- The response will also depend on the room.
I guess I would experiment playing at full volume without the sub playing, start at 40 sure and keep going up. When it seems OK, then start at the same frequency with the sub.
- You need a way to flip the polarity (often labeled as "phase" or "0/180"): the correct position should give strong bass and the wrong correction will make a noticeable big cancellation dip.
- You need to play like a pink or brown noise signal, more consistent than music. And measure with at least a phone app.
- If there's no dip either way then both are partly cancelling, and I'd spread the crossover frequencies a bit and check again.
This is why room correction like Audyssey is popular I guess!

Thanks so much HU! One last question that you raise - since I never ever approach full volume, not even close, should I just keep it simple and not use the hi pass on the mains? Or, will losing some of the lower frequencies allow better imaging, cleaner sound etc from the mains at normal listening levels??