Passive correction EQ for full range speaker.

Dear All!
This is first time I designed acoustic system for monitoring purposes.
Speaker is Dayton rs100-8, 4".  It can handle 30watt rms max, and 60 watts at peak.
Also it has very good frequency response curve (dunno how close real speaker is to that curve).

Here is calculation of frequency response in cabinet at zero degrees:


So, to achieve better results, I decided to involve passive filters to flatten FR.


Filters from left to right - Notch filter around 3,3khz, baffle step correction, Notch filter around 17khz.

So the problem I faced - which power rating components to choose in this circuit. Maximum power intended to use here - 20watts.
Does it mean that resistors must have 20watt rating (or better more), and inductors to be 2.5A plus?
I really don't want to have burning and smelling monitors
I would appreciate any constructive criticism. For example I'm not sure if impedance characteristics are good, also High end of FR  has that spikes, and not drops smoothly. May it cause unwanted harmonics or something else? (I understand it is above hearing range )

Resistors in parallel resonant networks can be few watts, they are active only at resonant frequencies, I don't believe you will listen pure tone 3kHz at 20W few hours, don't you? Inductors should be rated for maximum current.
Resistor in BFC is active in almost whole frequency range, an it would be about half of input nominal power.
Bear in mind that the impedance of speaker isn't steady 8 ohms in whole frequency range, and at 17kHz it is for sure much higher.
P.S.
I don't think this speaker would work better with all of these networks. For example, passive BFC would significantly lower damping factor of the system. IMHO, it is much better to make a control/EQ network at line level if it is really necessary.
   

moamps, thanks a lot for explanation and suggestions!
I will choose 50% of max power to all resistors, and for inductors I will add extra 20% of max current.

I don't think this speaker would work better with all of these networks. For example, passive BFC would significantly lower damping factor of the system. IMHO, it is much better to make a control / EQ network at line level if it is really necessary.


I thought about that, too. Some info says low damping factor is not problematic (according to Nelson Pass), in other sources it is problematic, so I'm not sure here ... The main problem - I want to get most linear FR as possible. Without filters the low frequencies will be weak relatively to high end (13dbSpl of difference is not low value indeed). Using active correction in preamp stage will add extra noise, so I'm getting tired to listening it after 3-4 hours. But your suggestion will be best choice, if I will manage to build ultra low noise active filters. Digital correction is the best, but latency will raise, what makes it useless during tracking.

Mikaeel said:

Some info says low damping factor is not problematic (according to Nelson Pass), in other sources it is problematic, so I'm not sure here ...

Click to expand...

Don't believe everything NP says. Damping factor matters, but not in the usual way it's presented. DF for amplifiers is just a notion of how the amp may react to high current demands. DF for speakers is decisive in the LF response and how  it responds to transients, but the DF of a speaker is quite low, usually about 1.3 to 1.5. That's because the DC resistance is close to the nominal impedance. Now if you add 4 ohms to a 8 Ohms speaker, the DF gets significantly changed. Actually, the LF response may be a little better, but not the transient response.

The main problem - I want to get most linear FR as possible. Without filters the low frequencies will be weak relatively to high end (13dbSpl of difference is not low value indeed).

Click to expand...

Have you calculated the filter with the actual speaker impedance, or with an 8ohm resistance? There's a lot of interaction between a passive filter and a speaker, that makes it very hard to predict accurately.

Using active correction in preamp stage will add extra noise, so I'm getting tired to listening it after 3-4 hours.

Click to expand...

Fallacy. An active filter that does this type of EQ will have a noise level better than -95dBu, so that's a dynamic range of 115dB. better than any 16-bit system.

abbey road d enfer, thanks for reply!

abbey road d enfer said:

Don't believe everything NP says. Damping factor matters, but not in the usual way it's presented. Actually, the LF response may be a little better, but not the transient response.

Click to expand...



I definitely will try all possible options, with measurements using specialized microphone.
Thanks for explanation, it helps me to understand it more.
Low frequencies boosted here by resonating properties of the cabinet, unfortunately this speaker has drop by -6db per octave starting from 90hz. This characteristics what shown on graph, are most linear as possible.

abbey road d enfer said:

Have you calculated the filter with the actual speaker impedance, or with an 8ohm resistance? There's a lot of interaction between a passive filter and a speaker, that makes it very hard to predict accurately.

Click to expand...



Emulation software uses both frequency response of speaker and the impedance data for it's calculations.
Red curve on graph is impedance characteristic.

abbey road d enfer said:

Fallacy. An active filter that does this type of EQ will have a noise level better than -95dBu, so that's a dynamic range of 115dB. better than any 16-bit system.

Click to expand...

I agree, only in case if I will manage to design precise filters using op-amps.

Mikaeel said:

Emulation software uses both frequency response of speaker and the impedance data for it's calculations.

Click to expand...

OK, I wasn't certain.

I agree, only in case if I will manage to design precise filters using op-amps.

Click to expand...

I don't really get it. Active filters are always more accurate than passive ones, if only because inductors vary with level and frequency.

abbey road d enfer said:

I don't really get it. Active filters are always more accurate than passive ones, if only because inductors vary with level and frequency.

Click to expand...

Sorry that was a typo.
I meant silent filters with lowest THD possible.
I attached graph to this message, which represents how will look FR curve with active line-level filters.
I need minimum six OpAmp's to achieve such results. It might be with high probability, that I will get correction unit with high self noise. It must be carrefully designed.

Mikaeel said:

Sorry that was a typo.
I meant silent filters with lowest THD possible.

Click to expand...

Silent is relative. Put atarget figure there.

I need minimum six OpAmp's to achieve such results. It might be with high probability, that I will get correction unit with high self noise.

Click to expand...

  6 stages of good audio opamps (5532) results in ca. -98 dBu noise. Is that enough?

abbey road d enfer said:

Silent is relative. Put atarget figure there.

Click to expand...

Well, I expect less than -80db @ 20watt (after poweramp), for regular distance from nearfield monitors. It's around 1 -1.5 meter.

abbey road d enfer said:

6 stages of good audio opamps (5532) results in ca. -98 dBu noise. Is that enough?

Click to expand...

Of course it's enough. But this is selfnoize of OP amps, filters will add harmonics too. Schemotechnic required here, must be smart. Filters must be organized in series, from high frequency to low (to avoid unnecessary amplification of second and third harmonics.
I'm kinda new to opamps, please correct me If I'm wrong.

Mikaeel said:

Well, I expect less than -80db @ 20watt (after poweramp), for regular distance from nearfield monitors. It's around 1 -1.5 meter.

Click to expand...

Since we're talking about active filters, the reference should be the input level the amplifier is operated at. Power, speaker efficiency and distance are irrelevant.
What is the nominal input level of the amp?

Of course it's enough.

Click to expand...

Why "of course? We don't know the nominal input level/sensitivity of the amp....

But this is selfnoize of OP amps, filters will add harmonics too. Schemotechnic required here, must be smart. Filters must be organized in series, from high frequency to low (to avoid unnecessary amplification of second and third harmonics.
I'm kinda new to opamps, please correct me If I'm wrong.

Click to expand...

According to the desired response of the filters, they are very likely to add very little noise; indeed careful implementation and gain-staging is required but it's almost trivial in that case.

abbey road d enfer said:

Since we're talking about active filters, the reference should be the input level the amplifier is operated at.
What is the nominal input level of the amp?
Why "of course? We don't know the nominal input level/sensitivity of the amp....

Click to expand...

I thought we're talking about line-level?
So, intput level of amplifier will be +4 dBu - 1.736Vpk max.

abbey road d enfer said:

Power, speaker efficiency and distance are irrelevant.

Click to expand...


How it is irrelevant, when I measure Noize of whole system in real space?

Mikaeel said:

I thought we're talking about line-level?
So, intput level of amplifier will be +4 dBu - 1.736Vpk max.

Click to expand...

So, how much Signal-to-noise are you expecting?


How it is irrelevant, when I measure Noize of whole system in real space?

Click to expand...

Because noise is a relative notion in most cases, that means you have to target a S/N ratio.
Or do you want to make your system unheard in comparison with ambient noise? Then yes, sensitivity matters. But what is the ambient noise level?

abbey road d enfer said:

But what is the ambient noise level?

Click to expand...

Our room is quiet enough. Around 26 -30 dbA.
Using that speakers, I will have approx. 79db of dynamic range. 20 watt - 99.2db @ 1 meter
I control level digitally on computer, so amplification system will work at constant power (Power amp don't has gain pot).  If whole system will have self noize (audio interface + correction eq + power amp) around 30dbSpl or more , then I will hear it for sure, especially on distance of 1 meter. The amp I ihave,  itself  has less than 1% of THD @ 20 watt.
Using  the non expensive active monitors (D class amps inside),  It produces noize around 32-36dbSPL. What makes me sad, after 4-5 hours of work.

Mikaeel said:

Our room is quiet enough. Around 26 -30 dbA.
Using that speakers, I will have approx. 79db of dynamic range. 20 watt - 99.2db @ 1 meter
I control level digitally on computer, so amplification system will work at constant power (Power amp don't has gain pot).  If whole system will have self noize (audio interface + correction eq + power amp) around 30dbSpl or more , then I will hear it for sure, especially on distance of 1 meter. The amp I ihave,  itself  has less than 1% of THD @ 20 watt.
Using  the non expensive active monitors (D class amps inside),  It produces noize around 32-36dbSPL. What makes me sad, after 4-5 hours of work.

Click to expand...

So you need the EQ to produce less than about 80dB below +4 dBu, that's -76dBu.
As I said earlier, it's trivial. You can easily achieve -90 dBu with non-exotic opamps, even with TL07x.

abbey road d enfer said:

As I said earlier, it's trivial. You can easily achieve -90 dBu with non-exotic opamps, even with TL07x.

Click to expand...

Thanks for your suggestions! I choosed NE5534.
Now I transfered FRD and ZMA of full speaker with cabinet influence into LtSpice.
Here below is example of current progress (one shelving filter & one notch in progress of calculations)
What you think, would be necessary to have input and output buffers for impedance matching?

Update:

Here is how I imagine it.
This is first time I calculate something like this. Still I'm not sure, if it's mandatory to have Highpass & Lowpass filters (20khz+cut, under 1hz cut), and if required to have in-out buffers .
Screenshot attached below.

Mikaeel said:

Update:

Here is how I imagine it.
This is first time I calculate something like this. Still I'm not sure, if it's mandatory to have Highpass & Lowpass filters (20khz+cut, under 1hz cut), and if required to have in-out buffers .
Screenshot attached below.

Click to expand...

The filters look all right. However the power amp is not. Powered with only 12V, you can't expect more than a few watts of it.
High and low-pass filters are not mandatory, but you have to have them; if not, you'll probably regret it. 
You don't need an output buffer because the ouptut from the 5534 is good enough to drive almost any power amp.
You need an input buffer because the first filter's response depends on the impedance of the source, though.
Good practice puts the high and low-pass filters at the input so they act as a buffer.

Update!
We finished assembling these reference monitors.
Yesterday I did measurements using Sonarworks microphone.
The frequency response of whole Box+speaker amazed me. But there is strange dependence of non-linearities and distance from speaker.
When it raises, then responce curve goes more nonlinear. I think that happens because we measure it in real room - reflections and standing waves may affect signal, as result we're getting false readings.
Anyway, I think this project is successfull.
I attached graph to this post, which compares Project nonlinearities with real measurements.
Red curve - simulated response. Black - real. !!! Correction EQ not used in this setup !!!
20watt AB class amplifier was used during measurements, focusrite 2i2 usb interface, sonarworks mic.

Mikaeel said:

But there is strange dependence of non-linearities and distance from speaker.
When it raises, then responce curve goes more nonlinear. I think that happens because we measure it in real room - reflections and standing waves may affect signal, as result we're getting false readings.

Click to expand...

When you move the speaker, the response changes because the loudspeakers are directional and also the difference of arrival time between the woofer and the tweeter changes with angulation.

We don't have tweeter. Only one speaker used in the box - rs100-8.
Here is comparison with Tannoy Reveal 502 (attached to this post)
Thanks for the suggestions abbey road d enfer!
I think we will made this project opensource. It might help small studios with low budget.