KLH-5 Measurements

[Carlspeak]

Below are two response tests done on a recently restored KLH-5 speaker. The cabs were stuffed pretty well with blocks of pink FG. Measured Fc (box resonance w/woofer) was in the 40-45 hz range between the pair I restored. This is quite similar to the AR3a target Fc. Box Qtc was in the 0.55 to 0.7 range - pretty well damped for an AS speaker of that era. The bass response on this pair will be quite 'tight'.

However, the 45 deg. off axis test shows a pretty severe drop beyond 5 kHz.

I recall reading somewhere that the model 5 was created to compete with the AR3a. It surely doesn't with regard to off axis response.

The on-axis test shows a 2-3 dB drop in resp. from 4-12 kHz when the grille is on compared to no grille. Strangely though, at 45 deg. off axis the grill didn't affect response nearly as much.

[speaker dave]

However, the 45 deg. off axis test shows a pretty severe drop beyond 5 kHz.

I recall reading somewhere that the model 5 was created to compete with the AR3a. It surely doesn't with regard to off axis response.

Interesting curves, Carl. Thanks for putting them up. The KLH 5 doesn't have the off axis response that a 3a had but it is considerably flatter on axis. In fact the 2 crossover points are undetectable.

David S.

[Steve F]

Interesting curves, Carl. Thanks for putting them up. The KLH 5 doesn't have the off axis response that a 3a had but it is considerably flatter on axis. In fact the 2 crossover points are undetectable.

David S.

Perfectly in keeping with Henry Kloss' (and later at Advent and BA, Andy Kotsatos' (nee Petite's)) design philosophy that favored on-axis, octave-to-octave balance over the Allison/Villchur far-field power/wide dispersion philosophy. The KLH-5/6/17 sounded the way they did on purpose, while the 4x/2ax/3a sounded like they did on purpose.

We won't go over that argument again, as we've beat that horse to a fare-thee-well, I'd say. It's just good to see that Carl's measurements are accurately indicative of the speakers' design intent, which gives all of us an additional valued source for FR data (Carl). Thanks, Carl.

Steve F.

[Carlspeak]

Perfectly in keeping with Henry Kloss' (and later at Advent and BA, Andy Kotsatos' (nee Petite's)) design philosophy that favored on-axis, octave-to-octave balance over the Allison/Villchur far-field power/wide dispersion philosophy. The KLH-5/6/17 sounded the way they did on purpose, while the 4x/2ax/3a sounded like they did on purpose.

We won't go over that argument again, as we've beat that horse to a fare-thee-well, I'd say. It's just good to see that Carl's measurements are accurately indicative of the speakers' design intent, which gives all of us an additional valued source for FR data (Carl). Thanks, Carl.

Steve F.

Thanks Steve and David.

I just rec'd two more to restore and will take and publish some more measurements.

[tysontom]

Below are two response tests done on a recently restored KLH-5 speaker. The cabs were stuffed pretty well with blocks of pink FG. Measured Fc (box resonance w/woofer) was in the 40-45 hz range between the pair I restored. This is quite similar to the AR3a target Fc. Box Qtc was in the 0.55 to 0.7 range - pretty well damped for an AS speaker of that era. The bass response on this pair will be quite 'tight'.

However, the 45 deg. off axis test shows a pretty severe drop beyond 5 kHz.

I recall reading somewhere that the model 5 was created to compete with the AR3a. It surely doesn't with regard to off axis response.

The on-axis test shows a 2-3 dB drop in resp. from 4-12 kHz when the grille is on compared to no grille. Strangely though, at 45 deg. off axis the grill didn't affect response nearly as much.

Carl,

Great job on your measurements!

I agree with Steve F that what you show in your measurements is in keeping with Henry Kloss' philosophy (and his other advocates at Advent and Boston Acoustics) of favoring on-axis smoothness and balance. All the more interesting is that other than initial measurements, most adjustments to the sound of KLH speakers were done by "ear" rather than by lab measurements. The rapid fall-off of the tweeter (in your measurements) above about 15 kHz seems a bit unusual, as the tweeter is shown at least -10dB on axis at 20kHz. The 45-degrees-off-axis droop is severe as well; I didn't this it was off this much.

I have a few questions:

(1) could you describe the conditions under which you made your measurements; i.e., 1-meter, 2-meter, ground-plane, in-room-gaited, anechoic (assuming outdoor quasi-anechoic in lieu of an actual anechoic chamber), etc;

(2) did you also make far-field measurements and

(3) what is the technical definition of your bass-response description "tight?"

Nevertheless, your measurements do show to some extent what Steve F describes as a difference in design philosophy between AR's acoustic-power response and KLH's (and Advent's) flatter on-axis response. Thanks for sharing your results.

--Tom Tyson

[RoyC]

Carl,

Where did you have the level controls set for your measurements?

Roy

[Carlspeak]

Carl,

Great job on your measurements!

I agree with Steve F that what you show in your measurements is in keeping with Henry Kloss' philosophy (and his other advocates at Advent and Boston Acoustics) of favoring on-axis smoothness and balance. All the more interesting is that other than initial measurements, most adjustments to the sound of KLH speakers were done by "ear" rather than by lab measurements. The rapid fall-off of the tweeter (in your measurements) above about 15 kHz seems a bit unusual, as the tweeter is shown at least -10dB on axis at 20kHz. The 45-degrees-off-axis droop is severe as well; I didn't this it was off this much.

I have a few questions:

(1) could you describe the conditions under which you made your measurements; i.e., 1-meter, 2-meter, ground-plane, in-room-gaited, anechoic (assuming outdoor quasi-anechoic in lieu of an actual anechoic chamber), etc;

(2) did you also make far-field measurements and

(3) what is the technical definition of your bass-response description "tight?"

Nevertheless, your measurements do show to some extent what Steve F describes as a difference in design philosophy between AR's acoustic-power response and KLH's (and Advent's) flatter on-axis response. Thanks for sharing your results.

--Tom Tyson

Hi Tom:

I measured these speaker the same as I do most - in my living room. The test speaker is elevated off the floor by simply having it sit atop its mate. Measuring distance is 1 M using a special test signal developed by Kirchner elektronik, the maker of ATB PC Pro. Below is an excerpt from their manual that explains the test signal:

"THE TEST SIGNAL

For measuring the PPN (Pseudo-Pink-Noise) signal is used. It is a deterministic noise. Through the function of 4X oversampling it is an analogue signal. One of the considerations during the development of the signal was the adaption of the measurement technology to the later function of the instruments tested. A loud speaker is for the transmission of music and speech. In correspondence with this the capacity of the loudspeaker is defined for the frequency range. In accordance with DIN is the performance at 100% total capacity as follows:

Low frequency from 40Hz – 600Hz 62%

Midrange from 600HZ – 4kHz 30%

Tweeter from 4kHz – 20kHz 8%

Because of it’s steady energy distribution the Pink-Noise signal is adapted to the frequency characteristic of the music. The MLS (Maximum Length Sequence) signal is the exact opposite.

PPN MLS

The pictures (attached .doc file) show the analyser measurements of the PPN and MLS signals. Both signals are a deterministic noise (a frequency mix). The PPN signal shows a pink noise. Practically all terz bars have the same amplitude. The drop in the low frequency range is not to be taken into consideration for measurements. The MLS signal tends to comply with white noise. This is easily audible when testing. The high frequencies have an extremely high amplitude and the low frequencies are contained in the signal with –45dB. In opposition to terz analysis the MLS signal appears after testing through the correlation with a straight frequency characteristic.

The loudspeaker is steered with the MLS signal shown above. If the loudspeaker has for example a linear frequency characteristic then the same signal will be recorded by the microphone used. It is changed into a digital signal within the measuring instrument. This signal as well as the MLS signal are withheld in the computer in digital form. During the correlation both signals are compared. The result ( frequency characteristic ) consists of a picture of the differences. If both signals are the same(in the case of the linear loudspeaker) the result is a straight line. If the loudspeaker does not transmit the low frequency range very well, then the difference between the signal measured and the output signal are pictured as a drop in the frequency characteristic.

That the MLS signal is not suitable for testing is proved by the following :

The loudspeaker is steered with the MLS signal. The energy distribution of the MLS the exact opposite of the music signal ; in the high a lot and in the low range nearly no energy.

The results of this are:

1. The extremely high energy of the high frequencies overload the high tones, so that a part of the frequency characteristic measured consists of dynamic compression and distortion. The result is that one of the music transmission frequency characteristics cannot be measured.

2. Because of the very small signals in the low frequency range an exact and reproducible measurement is nearly impossible. This leads to the need for at least 10 measurements with the MLS signal before an exact result (with adaptive windows) can be expected.

The PPN Signal owns a steady energy distribution. It doesn’t overload the high tones and has enough signal amplitude in the low frequency range. Each measurement is reliable (even without an average mean)."

I don't claim to understand it all. I would take a stab and define it as quasi-anechoic derived from a pseudo-pinknoise signal. However, I am very satisfied with the measurements I do get which, to me, display more accurately than others I've seen on the web like Clio. The measurements with ATB Pc show the same problems below 200 hz as others and I typically do a NF measurement of the woofer and splice or overlay it with the FF measurement. In the case of the KLH measurements I chose to show only the relavent portion of the freq. range above 200 Hz.

I did not make a FF mesurement beyond the initial 1M.

With regard to the term 'tight' Dickason uses the term 'taut' (which, IIRC, is a synonym of 'tight') to describe the sound quality of loudspeaker Qtc's in the range of 0.5 (1st ed. of TLSDCB pg.4).

Roy: The 2500-7000 switch was on 'HI' and the >7000 was in the middle position.

Let's see what I get for the pair of 5's I now have. These are 'as factory built' with no mods or upgrades. Fortunately for my customer, the caps in these have yet to fail so no xover rebuild has been done.

ATB_pc_PPN_bar_graphs.doc

[Carlspeak]

Some additional technical comments regarding my testing methodology:

1) I use a Behringer ECM8000 mic which recently underwent a calibration test by Cross-Spectrum Labs. It measured +- 2.2 dB from 20-20,000 hz.

2) The next series of tests will include HolmImpulse resp and impulse resp.

[tysontom]

Hi Tom:

I measured these speaker the same as I do most - in my living room. The test speaker is elevated off the floor by simply having it sit atop its mate. Measuring distance is 1 M using a special test signal developed by Kirchner elektronik, the maker of ATB PC Pro. Below is an excerpt from their manual that explains the test signal:

I don't claim to understand it all. I would take a stab and define it as quasi-anechoic derived from a pseudo-pinknoise signal. However, I am very satisfied with the measurements I do get which, to me, display more accurately than others I've seen on the web like Clio. The measurements with ATB Pc show the same problems below 200 hz as others and I typically do a NF measurement of the woofer and splice or overlay it with the FF measurement. In the case of the KLH measurements I chose to show only the relavent portion of the freq. range above 200 Hz.

I did not make a FF mesurement beyond the initial 1M.

With regard to the term 'tight' Dickason uses the term 'taut' (which, IIRC, is a synonym of 'tight') to describe the sound quality of loudspeaker Qtc's in the range of 0.5 (1st ed. of TLSDCB pg.4).

Roy: The 2500-7000 switch was on 'HI' and the >7000 was in the middle position.

Let's see what I get for the pair of 5's I now have. These are 'as factory built' with no mods or upgrades. Fortunately for my customer, the caps in these have yet to fail so no xover rebuild has been done.

Carl,

Thanks for your reply. It's really quite amazing what can be done with PC-based measurement software these day.

What surprised me was the KLH Five on-axis energy loss in the 15-20 kHz range; serious rolloff. It looks like as much as 10 dB down from the 12 kHz range, granted an area of highest energy. Do you think that microphone might have a loss in the highest octaves? Did they send you a calibration curve with it? The off-axis response loss was dramatic, as well. One might think it was 60 to 70 degrees off axis rather than 45 degree.

Nevertheless, that software does a great job of nulling the reflected energy, thus giving a good idea of what the speaker is probably doing at 1 meter out. There were also surprisingly few interference effects (possibly blanked by the gaiting) from the two midrange drivers as well as the complex 3-way crossover. All in all, the KLH Five looked pretty good.

The reason I asked about the term "tight," it tends to be a misnomer in describing a woofer. There are few transients reproduced by the woofer, in any event, but it is a common expression. Perhaps a woofer with no damping might be termed "loose," so it's easy to see how this came to be.

Thanks for sharing those curves with us.

--Tom Tyson

[Carlspeak]

Carl,

Thanks for your reply. It's really quite amazing what can be done with PC-based measurement software these day.

What surprised me was the KLH Five on-axis energy loss in the 15-20 kHz range; serious rolloff. It looks like as much as 10 dB down from the 12 kHz range, granted an area of highest energy. Do you think that microphone might have a loss in the highest octaves? Did they send you a calibration curve with it? The off-axis response loss was dramatic, as well. One might think it was 60 to 70 degrees off axis rather than 45 degree.

Nevertheless, that software does a great job of nulling the reflected energy, thus giving a good idea of what the speaker is probably doing at 1 meter out. There were also surprisingly few interference effects (possibly blanked by the gaiting) from the two midrange drivers as well as the complex 3-way crossover. All in all, the KLH Five looked pretty good.

The reason I asked about the term "tight," it tends to be a misnomer in describing a woofer. There are few transients reproduced by the woofer, in any event, but it is a common expression. Perhaps a woofer with no damping might be termed "loose," so it's easy to see how this came to be.

Thanks for sharing those curves with us.

--Tom Tyson

Tom:

Below is the 1/3 octave band on-axis mic calibration response results. The results show my mic is quite flat from 20-20kHz.

BTW, the graphs had 1/3 octave smoothing. I like the ATB - Pro software a lot. Check how useful it is for testing the attenuation effects of grille cloths in the Advent discussion forum.

Freq. Resp. (dB)

16 -2.60

20 -1.45

25 -0.72

31.5 -0.26

40 0.04

50 0.26

63 0.42

80 0.48

100 0.52

125 0.54

160 0.53

200 0.49

250 0.43

315 0.39

400 0.35

500 0.29

630 0.20

800 0.12

1000 0.00

1250 -0.07

1600 -0.18

2000 -0.07

2500 0.02

3150 0.18

4000 0.57

5000 0.95

6300 1.47

8000 2.12

10000 2.22

12500 1.79

16000 0.82

20000 -0.85

25000 -4.12

[Carlspeak]

Carl,

Thanks for your reply. It's really quite amazing what can be done with PC-based measurement software these day.

What surprised me was the KLH Five on-axis energy loss in the 15-20 kHz range; serious rolloff. It looks like as much as 10 dB down from the 12 kHz range, granted an area of highest energy. ........ One might think it was 60 to 70 degrees off axis rather than 45 degree.

event, but it is a common expression. Perhaps a woofer with no damping might be termed "loose," so it's easy to see how this came to be.

Thanks for sharing those curves with us.

--Tom Tyson

Tom:

Additional thoughts on your discussion above regarding the HF response of the KLH-5 tweeter.

The KLH-5 tweeter is of a cone type design. It's been my experience that all of those types of that era had a tendency to roll off rather sharply above 12-15 kHz. (for me, that's okay anyway with my ability to hear HF these days).

Take a look at Speaker Dave's AR4x measurements here (#1 & #4 posts).

http://www.classicspeakerpages.net/IP.Boar...?showtopic=5864

The tendency is the same and even the AR4x 30 deg. off axis resp. is heading towards what I got for 45 deg. with the KLH-5.

Nature of the beast perhaps?

[Carlspeak]

Here are some more measurements on another KLH-5. These were done with 3 different software packages and each has somewhat different results. My workhorse pkg. is ATB PC pro which I trust the most. REW and HolmImpulse are also good. These were run using my new calibration file except for ATB which did not.

The ATB test is the most complex because I took the most readings using it. Blue line=on axis, red=7.5 deg off, green=15 deg. off, brown=22.5 deg off, dark blue=30 deg off, light blue=45 deg off and black 60 deg. off.

The Room Equalizer Wizard test has the "685" line drawn in it. Only 4 tests were done. ON axis, 15 deg. 30 deg & 45 deg. The vertical scale is more compact and thus the curves look smoother. However, the same general deterioration in response going off axis is evident.

HolmImpulse test has 3 positions. On axis, 30 deg. and 45 deg. I can't explain the huge dip at 3500 hz in the on-axis test that didn't show in the other two tests. The impulse response data also looks reasonably good with no large humps out beyond 0.8 msec.

These tests were done with both switches max'd out.

The flattest response seems to be slightly off axis as noted in the ATB test. Also an average resp. of the REW data show is nice gradual tapering off in response that many seem to like and may explain why the 5 has so many fans.

[Zilch]

The HF is peaked on-axis to compensate for the directivity loss; unless toed-in, we don't normally listen there.

The technique is commonly employed with beamy tweeters....

[Carlspeak]

The HF is peaked on-axis to compensate for the directivity loss; unless toed-in, we don't normally listen there.

The technique is commonly employed with beamy tweeters....

Yeah, I think the 5's will sound rather nice toed in 15-20 deg when sitting in an equalateral listening arrangment.

[speaker dave]

HolmImpulse test has 3 positions. On axis, 30 deg. and 45 deg. I can't explain the huge dip at 3500 hz in the on-axis test that didn't show in the other two tests. The impulse response data also looks reasonably good with no large humps out beyond 0.8 msec.

Hi Carl,

It looks like some tweeter on-axis issues from baffle symmetry causing some cancelation. This always tends to disappear when you move a little off axis. The path lengths to the edges diverge and cancelation notches diminish. The 3.5k dip is worse in Holm (there are hints of it in the other curves) because you are using the progressive time windowing and seeing the early response only. The other curves look more steady state so later energy has filled in the dip. (Where is Howard to argue about which view is important?) There are also differences in smoothing, although the steady state curves are inherently rougher from including much later reflections.

Still, the overall trends are very similar, as you would expect from different measuring packages. In the end, the microphone will have more effect in the measurement than properly written, and operated, software.

Good data and thanks for sharing.

David

[Carlspeak]

Hi Carl,

It looks like some tweeter on-axis issues from baffle symmetry causing some cancelation. This always tends to disappear when you move a little off axis. The path lengths to the edges diverge and cancelation notches diminish. The 3.5k dip is worse in Holm (there are hints of it in the other curves) because you are using the progressive time windowing and seeing the early response only. The other curves look more steady state so later energy has filled in the dip. (Where is Howard to argue about which view is important?) There are also differences in smoothing, although the steady state curves are inherently rougher from including much later reflections.

Still, the overall trends are very similar, as you would expect from different measuring packages. In the end, the microphone will have more effect in the measurement than properly written, and operated, software.

Good data and thanks for sharing.

David

Hi Dave and 'top '0 the [very early] mornin' to you as well.

I just completed switching over my test software from an ancient laptop (XP) to a modern Windows 7 64 bit system - not without some angst and sweat and gracious personal help from from the president of the German company that makes the ATB Pro sytem.

[soundminded]

Below are two response tests done on a recently restored KLH-5 speaker. The cabs were stuffed pretty well with blocks of pink FG. Measured Fc (box resonance w/woofer) was in the 40-45 hz range between the pair I restored. This is quite similar to the AR3a target Fc. Box Qtc was in the 0.55 to 0.7 range - pretty well damped for an AS speaker of that era. The bass response on this pair will be quite 'tight'.

However, the 45 deg. off axis test shows a pretty severe drop beyond 5 kHz.

I recall reading somewhere that the model 5 was created to compete with the AR3a. It surely doesn't with regard to off axis response.

The on-axis test shows a 2-3 dB drop in resp. from 4-12 kHz when the grille is on compared to no grille. Strangely though, at 45 deg. off axis the grill didn't affect response nearly as much.

My recollection is that to my ears this speaker sounded pretty much the way you measured it. The predominance of the the lower midrange as evidenced by the 5 to 10 db overemphesis of the output between 150 hz and around 600 hz gave the speaker a bottom heavy tubby sound. This was not true for models 6, 17, or 12. Model 6 does make male voices sometimes sound tubby but not to nearly the same degree. When I first heard model 5, it was a real disappointment to me, I was expecting a significant improvement over Model 6.

I've always wondered if it used the same woofer as model 6. It appears to have a more extended low end response than model 6. I thought it was an improved 12" woofer also used in Model 12. It uses the same tweeter as models 6, 12, and 17. It adds two 5" full range drivers used as midranges, shades of AR2a but the KLH midranges are far superior to the Jensens used in AR2a. I think they are even slightly angled towards each other the way AR2a's were. This improvement in models 5 and 12 is one more acknowledgement of irreconcilable tradeoffs of a 2 way design without extraordinary measures. When you add a subwoofer to a two way design the way many modern systems do, you are turning it into a three way design because you have exactly the same tradeoff.

The tweeter response is as expected. The measured and claimed limit for it is 16 khz. This was the result one of the reviewers either in High Fidelity Magazine or Julian Hirsch reported. The falloff as a function of angle is no surprise either. The dustcap/cone arrangement make me think of it as a dome inside of a horn. Could it be the world's first Constant Directivity speaker a la Geddes? :-) Clearly the dispersion of the AR 3/4 inch dome is far superior but its output is much lower. With both AR2ax and KLH model 6 set for flat output using their level controls and preamplifier tone controls out of the circuit, KLH model 6 sounds much brighter to my ears than AR2ax. Even at maximum output, AR2ax does not have HF output balance comparable to KLH model 6 at its "normal" setting. Equalization changes the whole picture. With sufficient and judicious boost, AR2ax can be made to have a perceived flat output. KLH model 6 also requires some adjustment to its high end but not nearly as much.

[oldarnut]

In the late 1960's, I owned acoustic suspension speakers as did most all of my friends. I had AR2a's and my dad owned KLH6's. My best friend had AR3's until he got sick of replacing midrange units and replaced the AR3's with KLH5's. Another friend had AR3a's, still another large Advents.

As a result of a lot of speaker talk---mostly bragging, we built a switching board to AB compare five different systems. Instead of drinking beer and playing poker that night, we all brought one of our prize speakers and hooked them up to the switches and a Dynaco Stereo 70 bridged to mono.

We listened to our favorite "test" vinyl records and fairly quickly agreed we all preferred my dad's KLH6 to all the others. Seems funny that the KLH5 didn't impress us with its better lows. The AR3a went lower than any others, but not much. The AR2a had better dispersion of highs. The Advents didn't sound as good as any of the AR's or KLH's.

Just thought I'd share a look back at how we spent our time in the 1960's.

[JKent]

Hey oldarnut. You make me want to do a real comparison of my own. Right now my rec room "vintage" system drives a pair of AR2ax's. I have some other speakers of similar size that "could" take their place, awaiting restoration: AR2a's, Original Large Advents, KLH Fives and JansZen J-Z800s. The AR and KLH are 3-way, Advent and JZ 2-way.

I also have a pair of KLH Twelves that I love. But here's the weird thing: The Twelves and the Fives use exactly the same drivers. The only difference is the cabinet volume and the Twelve's external crossover box. But even though the Twelves are in a hallway where they can't be properly appreciated, they seem to sound much better than the Fives.

Can't put my finger on it but the Fives just don't satisfy. So maybe after all the speakers have been refurbished, I'll hook them up to my old speaker switcher and do a comparison head-to-head as you describe. My prejudice is that the AR2ax's will win (or maybe the 2a's--never compared the a & ax either). Bass is no problem, since I also have a VMPS original sub that I built from a kit a couple of decades ago: 15" driver and 15" passive radiator. I used to use a bridged Hafler 220 amp but last year installed a plate amp. Plenty of bottom (love those organ pedals!)

So it'll be back to the 60s. Too bad the ol' bell-bottoms from Sunny's Surplus no longer fit

Kent