ESL resonance measurement

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I built a pair of large ESL hybrids and they sound fantastic, except there seems to be a rather harsh peaking in the audio around 3kHz or so. It's only perceptable on certain audio tracks and can be quite pronounced on poorly engineered recordings that are abnormally bright or harsh to begin with.

I can construct a notch filter to reduce this, but I've been unable to accurately measure a peak response. Is there a preferred method for determining this, ie measured acoustic output through a mic, voltage peaking at the terminals with a scope/DVM, etc?
 
3KHz is fairly high for an ESL resonance. Most diaphragm resonances are in the range of 70-120Hz.

What technology is your panel? Is it a perforated metal stator, a wire stator, TIG rod, etc? Did you use foam tape as the assembly mechanism and spacer material? What physical size is your panel? Is it one stator or segmented?

An un-damped metal panel could have a resonance in the 500Hz to 1KHz range, so could a vibrating TIG rod.

For that high of a frequency, I would more suspect the lobing effects from panel width vs audio wavelength.
 
I built a pair of large ESL hybrids and they sound fantastic, except there seems to be a rather harsh peaking in the audio around 3kHz or so. It's only perceptable on certain audio tracks and can be quite pronounced on poorly engineered recordings that are abnormally bright or harsh to begin with.

I can construct a notch filter to reduce this, but I've been unable to accurately measure a peak response. Is there a preferred method for determining this, ie measured acoustic output through a mic, voltage peaking at the terminals with a scope/DVM, etc?

I had a similar peak centered around 2.7Khz with my homebuilt hybrid ESL's too-- a certain harshness that was more pronounced in some recordings. I also had a spare set of panels lying around made with different hole perforations and I noted that the harshness was much more pronounced on the 50% open perf panels as compared to the 40% open panels.

In my case I use a digital crossover and I also have a Behringer DEQ2496 digital EQ with RTA & microphone. The peak was obvious on the RTA--- not a sharp spike but rather a curve with its peak centered around 2.7kHz. I'm not sure what portion of the harshness (if any) was attributable to diaphragm resonance or if it was all just an interaction between the transformers' (tandem 230v/6v toroids) stay inductance and the panels' capacitance.

In any case, I was able to tame the peak and compensate the low frequency dipole roll off using my EQ and my speakers sound wonderful.

It's nice to have the mic and RTA but I found that I like the sound better when I EQ the speakers by ear (not necessarily flat), as opposed to setting a flat response curve per the RTA. Before I had the digital setup I used an analog crossover and graphic EQ and that setup worked pretty well too for tuning out that harsh peak.

It's amazing how obnoxious even a slight bit of lower treble harshness can be.

Good luck with that!
 
3KHz is fairly high for an ESL resonance. Most diaphragm resonances are in the range of 70-120Hz.

What technology is your panel? Is it a perforated metal stator, a wire stator, TIG rod, etc? Did you use foam tape as the assembly mechanism and spacer material? What physical size is your panel? Is it one stator or segmented?

An un-damped metal panel could have a resonance in the 500Hz to 1KHz range, so could a vibrating TIG rod.

For that high of a frequency, I would more suspect the lobing effects from panel width vs audio wavelength.

The panels are 18 GA powder coated steel, 18" x 48". Hole size is .127" on staggered .188" centers with 41% open area. The panel is separated into seven identical 16.5" x 5.8" windows with .75" solid (no perf) bands between windows. The spacer material is 0.062" 3M VHB tape. The diaphram is one piece (18" x 48") 0.6mil Hostaphan, graphite coated.

The step up is a Plitron PAT4134-ES (75:1) terminated on the panel side with 68K resistive load; this seemed to tame it somewhat without changing the sound significantly and transforms to a ~12 Ohm resistance on the input side.

I don't know if it is mechanical resonance or an electrical peak, but it is very noticable and quite annoying, rendering some recordings unlistenable.

Charlie-

I bi-amp with a DIY line level analog xover at 300 Hz. I could try an electronic variable notch until I find a suitable "sweet spot" as you suggest, then apply it in the xover. I was hoping there was a reliable way to measure the anomaly so I could at least start with something close.
 
I don't know if it is mechanical resonance or an electrical peak, but it is very noticable and quite annoying, rendering some recordings listenable.

Do you have capability to make near field and far field measurements?
I think you will probably find that near field measurement will not show any peaking.
This would clear any concerns with the transformer and mechanical/diaphragm resonance.

My guess is that you are hearing the bump(usually 3dB to 5dB) in the response about 1-2 octaves above where your finite line source is transitioning to a point source. This is similar in nature to the diffraction peak seen on boxed dynamic speakers where they transition from half space to full space. Unfortunately unlike the diffraction peak, this bump in your finite line source is a moving target that shifts up and down in frequency based on your listening distance. Floor to ceiling line sources are desirable because they do not have this problem.
Details on how the transition frequency shifts with listening distance can be found here:
http://www.diyaudio.com/forums/plan...r-esl-simulator-esl_seg_ui-2.html#post2913884

I bi-amp with a DIY line level analog xover at 300 Hz. I could try an electronic variable notch until I find a suitable "sweet spot" as you suggest, then apply it in the xover. I was hoping there was a reliable way to measure the anomaly so I could at least start with something close.

To determine a good starting point for the notch settings, I would follow CharlieM and measure with an RTA at your listening position.
You mention using a 300Hz analog crossover....did you include any shelving equalization to EQ the dipole roll off?
If not this would further accentuate the bump in the response.
 
Can you explain a bit more about the shelving eq? On the LR website they give an example of a circuit, but no parameters for determining the cutoff point. Is this a fixed response, or is it determined by the driver and or x-over? If it is fixed, then the response after my x-over is almost entirely in the flat region of the eq response anyway. The notch would be needed at a much higher point than the example would indicate, somewhere around 2-3kHz instead of 700Hz as the LR website shows.

I don't have an RTA system at my disposal, but have been looking at a couple of them on-line. Any suggestions?
 
Hi,

read the first post until ~#30 of this thread about frequency correction
Non-segmented panels always exhibit a peaking response in a range of ~1-5kHz, flat panels a bit stronger than curved panels.
This peak may result in soprano voices becoming quite shrill.
On the other hand it gives a impression of speed and livelyness.
It' a small rim between enough and too much.
ESLs are much more forgiving in this respect than dynamic speakers.
A dynamic driver with a similar amplitude lift in that range could pull Your inlays.

jauu
Calvin
 
Can you explain a bit more about the shelving eq? On the LR website they give an example of a circuit, but no parameters for determining the cutoff point. Is this a fixed response, or is it determined by the driver and or x-over? If it is fixed, then the response after my x-over is almost entirely in the flat region of the eq response anyway. The notch would be needed at a much higher point than the example would indicate, somewhere around 2-3kHz instead of 700Hz as the LR website shows.

You are correct that the shelving network probably won’t be affecting the response up at 2-3kHz where your peak is. What I was suggesting is that if no shelving equalizing is used before your crossover, the overall spectral balance might be tilted upward rather than flat. This would make any peak in the 2-3kHz range even more audibly prominent. As Calvin said, just a little bit can make a big difference with ESLs.

In general shelving networks can easily be designed with whatever transition frequencies you need to properly equalize the dipole roll-off down to your crossover frequency. Attached is a design note pulled from the Linkwitz Active Filters page.
Active Filters

A link to an Excel spreadsheet is provided at the top of the Filters page that will help calculating circuit component values for your desired response.
http://www.linkwitzlab.com/Linkwitz Filter Circuits Tool V4.1.xls


I don't have an RTA system at my disposal, but have been looking at a couple of them on-line. Any suggestions?
If you don’t want to drop the $ for a standalone RTA system, I would recommend REW. It is free and is one of the easier measurement software packages to get up and running. If you don’t already have a microphone, the microphone from minidsp is a nice option that works with REW and doesn’t require any power supply are pre-amp…just a free USB port.
UMIK-1 | MiniDSP
REW - Room EQ Wizard Room Acoustics Software
 

Attachments

  • LW_shlv-lpf.gif
    LW_shlv-lpf.gif
    24.2 KB · Views: 250
I think I finally solved the harsh response problem in my ESL panels, but part of the solution is still a mystery to me.

I was using Plitron toroidal step ups with a ratio of 75:1. I had used a pair of Hammond 60W output transformers on a previous design and they worked well (no harshness). I still had a pair of them laying around; with a 4K3 plate resistance it reduced the step up ratio to 32:1 but that only reduced the signal to the panels by ~6dB and I had plenty of head room on the amp. It seemed like the easiest thing to try, so I swapped out the Plitrons for the Hammonds which improved the harshness immensely; Any idea why this would make a difference?

There was still some peaking at ~3kHz, although it was much more tolerable than before. I bought the REW mic/software from miniDSP and did some room measurements. There was definitely a peak between 2.5kHz and 5kHz (May 31 500-5K).

I simulated a 4 tap LC notch filter that should have flattened out the peak response. I prototyped the filter and connected it to the input of the step up xfmr; big difference. The harshness all but disappeared (very bright or harsh recordings still sound a bit shrill, but that's probably how they were recorded). The room measurement (Jun 9 500-5K) bore this out.

I constructed the filters on a piece of copper clad PCB which I routed with a trim router and flush cut bit. I'll continue with listening tests, but so far, it sounds as close to neutral as these speakers ever have.
 

Attachments

  • may 31 500-5k.jpg
    may 31 500-5k.jpg
    43.6 KB · Views: 108
  • jun 9 500-5k.jpg
    jun 9 500-5k.jpg
    42 KB · Views: 109
  • DSCN1020.jpg
    DSCN1020.jpg
    268.9 KB · Views: 100
  • DSCN1021.jpg
    DSCN1021.jpg
    454.3 KB · Views: 109
Hi,

most panels show a slight but broad peak in the 3-4kHz range.
This is the same freq-range of high pitched female voices (sopranos).
These voices tend to shrill anyway.
A speaker having a pronounced amplitude response in this range can easily become unbearable at elevated levels.
A difference of less than 1dB in amplitude response can make an important difference sonically.
ESLs seem somehow less negatively affected by this phenomen than dynamic speakers.
Still though with most panels one has to care for the amplitude response in that freq-range.

jauu
Calvin
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.