AR-303 Comments and Observations

[Pete B]

The AR-303 was mentioned in regard to the experimental vertical AR-11 that I prototyped. I read the Stereophile review of the AR-303 years ago when it came out but have not had a chance to listen to them.

One factor that has been on my mind is the very high moving mass of the AR-303 woofer as discussed here:

http://www.classicspeakerpages.net/dc/dcbo...7431&page=#5243

On reviewing the moving mass again in light of the Stereophile review something does not seem right since the 303 should have a fundamental resonance that is much lower than the AR-3a's 41 Hz, the Stereophile review shows 38.5 Hz:

http://stereophile.com/images/archivesart/303FIG1.jpg

I took another look at the data and it seems that the very low Vas of the 303 woofer is the reason for the less than expected reduction and there is fairly good agreement with the Unibox simulation showing an Fc of 38.12 Hz for "walls covered" damping, and 36.31 for "heavy fill". The amount of damping strongly influences the height of the impedance peak at resonance and the Stereophile data suggests even higher damping than the "walls covered" value in the simulation. There seems to be a small discrepancy and there could be several reasons, however, it's certainly in the ball park.

One thought I had was that the 303 woofer might have more cone area as with a true 12", however the data from Ken's tests suggests the opposite:

Sd

DUT#1 (303) 416 cm^2

Others 450 cm^2 more or less

It would help if someone measured the piston diameter of the AR-303 woofer as a sanity check.

I've mentioned the importance of moving mass and how stronly it influences efficiency/sensitivity. Sensitivity was measured at 84dB/2.83V in the Stereophile test report.

Pete B.

[Pete B]

Here is the AR-303 "anechoic response on tweeter axis at 50", averaged across 30 degrees horizontal window and corrected for microphone response, with nearfield woofer response plotted below 300Hz"

as measured in the Stereophile review:

http://stereophile.com/images/archivesart/303fig2.jpg

The large peak at 60 Hz does not pass a sanity check, unless something was done in the crossover to cause it. I have wondered about the accuracy of their measurements for some time now.

The response overall is reasonably smooth.

Pete B.

[Pete B]

The 1991 PSB Stratus Gold also tested by Stereophile as a modern comparison: "anechoic response on midrange axis at 48", averaged across 30 degrees horizontal window and corrected for microphone response, with the nearfield woofer and port responses plotted below 300Hz":

http://stereophile.com/images/archivesart/PSB91fig5.jpg

Once again I do not completely trust this data but still it is interesting to take a look.

The manufacturer claims +/- 1 dB from 36 Hz to 20 kHz as measured in the Canadian NRC lab. I'd expect this to require some smoothing and one should consider that it is their marketing literature.

Some may believe that the dip around 3 to 4 kHz is a design error however I believe that it is the well known "BBC dip" and is deliberate.

Pete B.

[soundminded]

>The large peak at 60 Hz does not pass a sanity check, unless

>something was done in the crossover to cause it. I have

>wondered about the accuracy of their measurements for some

>time now.

Pete, do you think the proximity effect of the microphone was taken into consideration? I'm with you, frankly, I never thought much of this magazine and from what little I've seen of it, I do not believe they are the technical equal of any of the three popular hobbyist magazines of the 60s and 70s High Fidelity, Stereo Review, or Audio Magazine all of which seemed to have competent labs.

[Pete B]

A pressure mic is used and therefore the proximity effect is not an issue.

I've been staring and thinking about these curves and I have a theory. The problem seems to be below 300 Hz and it looks like there is extra baffle step compensation in the AR-303 response. A nearfield measurement method is used below 300 Hz and this essentially measures volume velocity, which only has an amplitude correction to half space, but it would also have a baffle step correction, which would remove the rise seen in the curve, when translated to full space or anechoic conditions. I believe this is the error.

Pete B.

[Guest richd]

Some questions have been asked about the difference

between the crossovers of the AR 11 and AR 303a

Attached are Multisim (Spice) runs of the crossovers

transfer function at the speaker terminals

In Figure 1 an impedance curve of the 12inch

driver using the model parameters in the Classic Speaker

page library is shown. This data was supplied by Ken Kantor

In figure 2 the original measured plot of the

impedance is shown, Note how close this is to figure 1

which shows I have the crossover in the simulation

wired correctly

Figure 3 shows the complete Multisim schematic with

the AR 303a crossover in place. Note how I reflected

the mechanical parameter of the woofer into the

electrical model of the woofer.

Figure 4 shows the resulting transfer function from crossover

in to voltage on the speaker terminal out. Both the AR 11

and AR 303a crossovers are simulated.

We can see a significant resonance in the AR 11 transfer

function that is absent in the AR 303.

In the AR 11 a simple LC crossover is used with the -3dB

point set to 600Hz. The value of L is 2.3mH and C is 30uFd.

These values come off the schematic in the Classic Speaker

pages web site.

The schematic of the AR 303a can also be found on

the classic speaker page web site. The AR 303a

crossover is more complex (see figure 3). For frequencies

far beyond the -3dB point only the 50uFd contributes

to the frequency response since its reactance is well below

5 ohms. This calculates as a 400Hz. Thus the out off band

response acts like a 400Hz 2nd order low pass filter. Below

about 600Hz, which the value that a 50uFd cap will be 5 ohms

or less the 80uFd cap in series with the 5 ohm resistor comes

into play.

Please note that these simulations do not include the

acoustic model of the speaker in the box. For that

simulation I would need the electrical equivalents of

the box parameter. Also note that taking the 2nd derivative

of these transfer functions will not yield the frequency

response of the speaker. To get that I would need to

simulate the complete 2 port model of a closed

box speaker with the parameter for the box as well

as the parameters of the driver supplied by Ken Kantor.

While not including the box acoustic parameters

into the impedance model used in this

simulation has certainly affected the accuracy

of the results it appears the more complex

network in the AR 303 is to reduce the effect

of the complex impedance characteristic of the

woofer on the crossover transfer function.

David Rich

802.txt

803.txt

804.txt

[tysontom]

I think it is great to have David Rich contribute to this forum. In my view, this adds another important dimension to the AR site, and we are fortunate to have his insight into speaker design and performance. It will give a slightly different perspective to sound quality of the older AR speakers in relation to newer technology, but this is a good thing.

This reinforces the feeling that this is an excellent forum and contains great insight into sound quality in general and AR speaker technology in particular!

--Tom Tyson

[Guest richd]

Attached are a couple other figures. In these I reduced the inductor

in the AR 303a woofer crossover from 3mH to 1.4mH. This alights

the stopband with the AR 11. From the stopband both now have

an effective -3dB point of 600Hz although the in band response

is significantly different go to the speaker loading characteristics.

I do not know if the 3mH value of the crossover schematic for the AR 303a

shown on the schematic posted on the classic speaker pages is

a mistake. I have a physical AR 302 crossover board and can verify

the capacitor and resistor values. I do not have a way to measure inductance.

David Rich

[kkantor]

Hi, David, very nice to see you here!

It will be a few weeks until I can open up my own reference pair to double-check the inductor value, but I will. Perhaps someone here can do it sooner.

-k

[Guest Bret]

drich's attachments 2-4 with extensions corrected for easier viewing.

Bret

http://www.classicspeakerpages.net/dc/user_files/819.jpg

http://www.classicspeakerpages.net/dc/user_files/820.jpg

http://www.classicspeakerpages.net/dc/user_files/821.jpg

[johnieo]

Hello David:

>In the AR 11 a simple LC crossover is used with the -3dB

>point set to 600Hz. The value of L is 2.3mH and C is 30uFd.

>These values come off the schematic in the Classic Speaker

>pages web site.

Regarding the woofer L, C values you used in your Spice model: The correct value of the inductor is 2.85 mH, as those of us who have re-habed AR-11s have discovered. I believe Rich Laski pointed out that the scanned drawing likely had been folded in the middle of the numeral eight causing it to be mis-read as numeral three.

Second, there is an error in library crossover drawing. The AR-11 Model A woofer capacitance was derived from two devices: a 72-uF in parallel with a 50-uF totalling 122 uF. In the AR-11 Model B, the two capacitors were replaced with a single, 120-uF unit. The value of 30-uF woofer you used assumed that the two were series-connected, as they are incorrectly shown in the library drawing.

Thank you for your simulation, David.

Cheers,

[Pete B]

(wish I could delete this)

[Pete B]

Hi David,

I simulated and measured the AR-11 woofer section with the correct values of 2.8 mH (I measure it as 2.8 not far off from 2.85 measured by others) and 120 uF. I did these some time apart but from memory agreement was good. There is some peaking, and it is the series resonance of the cap and inductor, with the primary losses being the inductor DCR and the cap ESR. The peak is somewhat sensitive to ESR because the ESR value is significant, compared to the resistive loss in the inductor.

My point here is that due to this relatively high Q in the AR-11 crossover, it is more sensitive to inductor DCR and capacitor ESR than the more common lower Q designs would be. The capacitor ESR should be included in any simulation.

And as I mentioned in another thread substituting large (low DCR) inductors and low ESR caps, without adjustments to alter Q, will certainly make it sound different, it will have an even larger peak in the mid bass. This is not the case in the 303 where the larger inductor, and smaller cap lower the electrical Q.

I believe, based on my experience and the Stereophile measurements that the 3 mH in the 303 is correct. The inductor values in both the AR-11 and 303 seem to be a bit high from my experience, and based on listening tests (AR-11 only) where there is a bit too much baffle step compensation. Again I don't have the notes in front of me but it seems to be about 2 dB IIRC more than is needed not as much as is shown in the Stereophile review. I believe that this was done to give them a bit more low bass (50-60 Hz) thump for mass market appeal. It does not sound natural in my experience.

Pete B.

>Attached are a couple other figures. In these I reduced the

>inductor

>in the AR 303a woofer crossover from 3mH to 1.4mH. This

>alights

>the stopband with the AR 11. From the stopband both now have

>an effective -3dB point of 600Hz although the in band

>response

>is significantly different go to the speaker loading

>characteristics.

>

>I do not know if the 3mH value of the crossover schematic for

>the AR 303a

>shown on the schematic posted on the classic speaker pages is

>a mistake. I have a physical AR 302 crossover board and can

>verify

>the capacitor and resistor values. I do not have a way to

>measure inductance.

>

>David Rich

[Guest richd]

>Second, there is an error in library crossover drawing. The AR-11 >Model A woofer capacitance was derived from two devices: a 72-uF in >parallel with a 50-uF totalling 122 uF.

Please refer to:

http://www.classicspeakerpages.net/ar/ADD/...(11A)%20(3).jpg

The capacitors are in series not parallel. The total capacitance is 30uFd. A quick check of the resonant frequency of the LC network with a 2.35mH inductor yields 600Hz.

David Rich

[johnieo]

>Please refer to:

>

>http://www.classicspeakerpages.net/ar/ADD/...(11A)%20(3).jpg

>

>The capacitors are in series not parallel. The total

>capacitance is 30uFd. A quick check of the resonant frequency

>of the LC network with a 2.35mH inductor yields 600Hz.

Hello David:

I respectfully disagree with the AR-11A x-o diagram posted in the Archives. Pleas examine a detail of that drawing reproduced here:

http://www.classicspeakerpages.net/dc/user_files/829.jpg

The partially obscured middle numeral in the woofer inductor is an eight, not a three. One can see this by comparing the font of the numeral three in the resistor value lettered at the top of this detail, and the font of the numeral eight lettered in the mid-range inductor--clearly it is not a three. One can also see the vertical fold line disrupting the coil symbol and everything in its vertical path because the two pieces of the drawing are not correctly alligned horizontally.

Measurement of the AR-11 woofer inductor verifies that its value is 2.85 mH.

Those who have opened an AR-11 case to replace the capacitors have found the two caps wired in parallel and not in series, as is incorrectly shown on this diagram. It is for this reason that I prefer to compare the real device to the drawings, which unfortunately may contain errors.

Cheers

[Guest Bret]

John, I hope neither you, nor Mr. Rich, nor Mr. Laski will object to my interjecting these links to photos of both types of crossovers (paralleled and single cap) from real AR-11s.

This should help clear-up any confusion there has been.

Thank you again, Mr. Laski, for taking these and posting them.

http://www.classicspeakerpages.net/dc/user_files/335.jpg

http://www.classicspeakerpages.net/dc/user_files/336.jpg

Bret

[Guest richd]

First I apologize to John, I did not read through his post completely enough to understand that the schematic diagram on the web site was wrong. So 2.85mH and 120uFD is what should be simulated. I have done so in the attached plots. As before the simulation uses the Ken Kantor 12 inch AR woofer model supplied in the library. Remember that this model is in free air. The parameters for the resonance frequency and the maximum resistance would change if the speaker was in a box but those values are not available to me. Yes I have modeled the DC resistance of the inductor. No I have not modeled the ESR of cap as I do not have that data for the old caps used in the AR 3 and 11.

Now some quick background. To design a 2nd order LPF we use L and C to set the cutoff frequency. With both L and C available it looks like we can pick any L and find a matching C. In reality we do not have freedom because R of the woofer sets the Q of the filter (not be confused with the Q of woofer rolloff). So given say 4.27ohms for R (the resistive component of the Ken Kantor model) and a desired 550Hz -3dB cutoff point we need an L of 1.7mH and a C of 50uFd. This also plotted as the idea curve as a reference. Note that woofer is represented only as a 4.27ohm resistor in this plot

I have also plotted the AR 303 curve as I did in the first post.I kept the L at 3mH.

As can be seen from the attached plots the AR 11 corner frequency as inferred from the stopband characteristics of the filter is 275Hz. That is exactly what I calculate from the L and C values provided. I once again point out that the voltage on the speaker terminals is not directly related to the pressure that comes of the speaker at a given frequency so I am not drawing any conclusions from this data. I do find these results do look surprisingly similar to the measured data I posited yesterday of the AR 302 and AR 5 woofer frequency response. This was totally unexpected given the simulation results I posted here before.

Note that these simulations (with all the caveats) and my measured results of the 5/302 pair posted yesterday show the Ken Kantor designs have not been peaked up in the bass relative to the classic designs. Indeed some energy loss may be occurring in the 200 - 400Hz range.

David Rich

[tysontom]

>John, I hope neither you, nor Mr. Rich, nor Mr. Laski will

>object to my interjecting these links to photos of both types

>of crossovers (paralleled and single cap) from real AR-11s.

>

>This should help clear-up any confusion there has been.

>

>Thank you again, Mr. Laski, for taking these and posting

>them.

>

>http://www.classicspeakerpages.net/dc/user_files/335.jpg

>

The attached drawing is from an AR-11 board-layout drawing, and it confirms the layout (above) of Mr. Laski's single-120mfd-cap version AR-11. I have added the correct part numbers (for ID purposes) for each component except for resistors.

http://www.classicspeakerpages.net/dc/user_files/839.jpg

(Fig.1) This is part of a huge drawing #800028 D, entitled "AR-11 Crossover Assy One-Piece Board," dated February 26, 1976. Two mistakes: (1) crossover coil #9, 2.85 mH, has part number 805009, not shown, and (2) 40 mfd crossover capacitor is number 803009.

http://www.classicspeakerpages.net/dc/user_files/840.jpg

(Fig.2) This is an early pencil-drawn AR-11 crossover schematic, also from about 1976 I believe, showing the addition of the single 120 mfd capacitor across the woofer.

[Pete B]

Your idealized blue curve, looks like a Butterworth response, is a text book design, and it is well known that the important response in loudspeaker design is the cascade of the electrical response with the acoustical response of the driver, and therefore most often the electrical response does not follow a text book formula. It is very interesting that your 1.7 mH calculated value is very close to the old 1.9 mH value for the AR-3a which I have stated does not have full baffle step compensation, which your curve supports. Indeed your curves show the reponse of crossovers with and without baffle step compensation.

Baffle step compensation is most often implemented in modern designs by making the woofer, and perhaps midrange, inductor much bigger which causes the filter to start attenuating at a much lower frequency with a much lower slope than pure second order, then as frequency increases the slope becomes second order due to the capacitors influence. This is what's going on in the AR-11 and 303 with the larger than text book inductors. The peaking or shelving in the low pass response for the AR-11 is highly sensitive to capacitor ESR, and your curve shows the worst case of a perfect capacitor with an ESR = 0 ohms. There are ways to estimate the capacitor ESR based on historical measurements.

You mention that the Ken Kantor designs have not been peaked up in the bass I take it you mean in contradiction to the Stereophile response graph. I agree having commented above that it is probably a measurement error. You say that the 200 to 400 Hz range may be attenuated, well another view is that the below 200 Hz range is elevated. Just in case there's any confusion my comments about peaking in the crossovers, and in the bass are for the AR-11, not the 303. The low bass problem in the AR-11 is the low moving mass and indeed the 303 has significantly *increased* mass, ~105g for AR-3a, ~68g for AR-11, and ~139g for AR-303.

Pete B.

>First I apologize to John, I did not read through his post

>completely enough to understand that the schematic diagram on

>the web site was wrong. So 2.85mH and 120uFD is what should be

>simulated. I have done so in the attached plots. As before the

>simulation uses the Ken Kantor 12 inch AR woofer model

>supplied in the library. Remember that this model is in free

>air. The parameters for the resonance frequency and the

>maximum resistance would change if the speaker was in a box

>but those values are not available to me. Yes I have modeled

>the DC resistance of the inductor. No I have not modeled the

>ESR of cap as I do not have that data for the old caps used in

>the AR 3 and 11.

>

>Now some quick background. To design a 2nd order LPF we use L

>and C to set the cutoff frequency. With both L and C available

>it looks like we can pick any L and find a matching C. In

>reality we do not have freedom because R of the woofer sets

>the Q of the filter (not be confused with the Q of woofer

>rolloff). So given say 4.27ohms for R (the resistive component

>of the Ken Kantor model) and a desired 550Hz -3dB cutoff point

>we need an L of 1.7mH and a C of 50uFd. This also plotted as

>the idea curve as a reference. Note that woofer is represented

>only as a 4.27ohm resistor in this plot

>

>I have also plotted the AR 303 curve as I did in the first

>post.I kept the L at 3mH.

>

>As can be seen from the attached plots the AR 11 corner

>frequency as inferred from the stopband characteristics of the

>filter is 275Hz. That is exactly what I calculate from the L

>and C values provided. I once again point out that the voltage

>on the speaker terminals is not directly related to the

>pressure that comes of the speaker at a given frequency so I

>am not drawing any conclusions from this data. I do find these

>results do look surprisingly similar to the measured data I

>posited yesterday of the AR 302 and AR 5 woofer frequency

>response. This was totally unexpected given the simulation

>results I posted here before.

>

>Note that these simulations (with all the caveats) and my

>measured results of the 5/302 pair posted yesterday show the

>Ken Kantor designs have not been peaked up in the bass

>relative to the classic designs. Indeed some energy loss may

>be occurring in the 200 - 400Hz range.

>

>

>David Rich

>

[Pete B]

Hi Tom,

Thanks for the documentation. You probably noticed this but the assembly drawing (Fig. 1), and the photos show single pole switches, whereas the schematic (Fig. 2) shows double pole. The XO designs are slightly different there are 5 and 10 ohm resistors in the assembly drawing and none of that value in the schematic.

The AR-11s that I have here have the single pole switches and are identical to the photo and assembly drawing.

Pete B.

[kkantor]

To the extent that I am able to understand (calculate, model, measure) their behavior, the 303's have a Qtc of 0.83. I am including the source impedance of the xover in this figure. So we can expect a peaking of, what, about 0.3 or 0.4 dB dB? Certainly more than I would use for an NHT product, but likely to be swamped by other factors at 40 Hz.

I writing this from CES. If I run into JA, I will ask him if he remembers his measurement conditions.

[r_laski]

Tom,

Thanks for sharing the drawings!

Fig 1 and my picture of the single board AR-11 X-O do match up almost precisely. A minor detail, but the HR 5 Ohm resistor is wired directly to the negative terminal, not to one of the 120 mfd cap lugs, like the MR 10 Ohm resistor is.

>>>>(Fig.2) This is an early pencil-drawn AR-11 crossover schematic, also from about 1976 I believe, showing the addition of the single 120 mfd capacitor across the woofer.<<<<

http://www.classicspeakerpages.net/dc/user_files/336.jpg

http://www.classicspeakerpages.net/dc/user_files/840.jpg

I agree with Pete B’s post about Fig 2. It matches up better with the two-board AR-11 X-O. The double pole switches and the resistor values match what is found on the back board of the two-board X-O. May be hard to see the double pole switches in the photo because of the yellow wire, but the resistor values are clear and match the schematic. Note that all four (4) center pins of both double pole switches are wired together and directly to the positive terminal by the thin wire with clear insulation. The yellow wire hides this detail.

I noticed in addition to the 120 mfd cap, the fuse is deleted. It is possible the 120 mfd cap represents the paralleled 50 mfd and 72 mfd, although we know these two caps paralleled equals 122 mfd.*

Another very real possibility is that there are two-board AR-11 X-Os out there with a single 120 mfd cap.

*Note that for at least some early AR-11’s (and AR LST’s) AR paralleled two 20 mfd Sprague caps for the 40 mfd midrange cap value. You can see the pink pair of 20 mfd Spragues in the two-board X-O photo. We know AR also paralleled the woofer caps in early AR-11's. Whether its a single cap or paralled pair is not as important as knowing what the capacitance value should be, which is what your revised drawing provides.

Again, thanks for sharing the drawings. They provide invaluable information.

Rich

[Pete B]

I would just like to clarify that the intent of this thread was not to discuss the 60 Hz peak in the Stereophile response curve. As I said from the start I believe that it is a measurement error. And I want to point out that I understand that Ken's objective was to design a system with a "family" sound similar to the original 3a which I believe he achieved and more.

I mention the AR-303 and the response curve because it was brought up in response to the AR-11 vertical experiment.

My point was that I have no intention of holding to the AR family sound, rather I'm interested in getting the best performance from the drivers with minimal crossover modifications.

I was interested in discussing the differences in the vintage AR sound as compared to modern designs.

Pete B.

>Here is the AR-303 "anechoic response on tweeter axis at

>50", averaged across 30 degrees horizontal window and

>corrected for microphone response, with nearfield woofer

>response plotted below 300Hz"

>as measured in the Stereophile review:

>

>http://stereophile.com/images/archivesart/303fig2.jpg

>

>The large peak at 60 Hz does not pass a sanity check, unless

>something was done in the crossover to cause it. I have

>wondered about the accuracy of their measurements for some

>time now.

>The response overall is reasonably smooth.

>

>Pete B.

[Pete B]

Here John Atkinson comments and supports my theory about the peaking in the 303 measurement, however it seems that in this review they chose a compromise level for the woofer output, less peaking around 60 Hz, and droop around 200 to 300 Hz. This is from a review of the KEF Reference 207 loudspeaker, Stereophile February, 2003:

"Fig.3 shows the 207's overall response averaged across a 30 degrees horizontal window on the Uni-Q axis, spliced to the complex sum of the low-frequency nearfield responses. (Both acoustic phase and the different distances of each driver to a nominal farfield point are taken into account.) The 3dB rise in the midbass level will be due to the nearfield measurement's assumption of a 2pi (hemispherical) environment."

http://stereophile.com/images/archivesart/kef207fig3.jpg

"Fig.3 KEF Reference 207, anechoic response on Uni-Q axis at 50", averaged across 30 degrees horizontal window and corrected for microphone response, with the complex sum of the nearfield responses plotted below 300Hz."

Pete B.

[Pete B]

Note the similar rise, however as I recall Don Keele's review in Audio of the Paradigm Studio/100 did not show this low end lift:

http://stereophile.com/images/archivesart/Parfig4.jpg

Fig.4 Paradigm Studio/100, anechoic response on-axis at 50", averaged across 30 degrees horizontal window and corrected for microphone response, with the complex sum of the nearfield midrange, woofer, and port responses plotted below 300Hz.

Pete B.