Refurbishing a pair of JBL L150A loudspeakers

I acquired  a secondhand  pair of these in 1988.  They had the classic JBL 'live music sound' so lacking from normal  domestic loudspeakers.   But in about  2002  I realised that they were not behaving as they should - the vibrant  'slam' and 'crash' of percussion was softening.  Inspection revealed that the foam materials of both the 12" drivers and passive radiators were disintegrating.  I removed the four units and delivered them to  Wembley Loudspeakers,  a business in London specialising in repairs to driver units.   They obtained replacement cone surround kits and restored much of the original sound quality.

But also I'd noticed that the mid-range and tweeter 'level controls' were rather scratchy, so I thought I'd  replace the potentiometers.  But to do that, I'd really need the circuit diagram.  Then I thought - well, if I've opened the box,  why not convert the divider network to 'tri-wired'?.  I hadn't read of any explanation of why tri-wiring was a good thing, what the details were, or even how it worked.  I guessed that it was the effects of voltages generated in common lead impedances formed by cables from the amp to the speakers.

And I'd noticed quite recently that there seemed to be some distortion coming from the mid-range units.  I downloaded a utility from http://www.electronics-lab.com/downloads/pc/index.html  which allowed me apply  sine frequency sweeps and white noise.  Listening to noise is very revealing, and is able to discriminate between nominally identical loudspeakers. 


                             

Sure enough, it sounded as though both the LE5 voice coils were rubbing.  So, off to Wembley Acoustics again for a replacement cone job.

As a result of all this, I came across the JBL Legacy web-site forum www.audioheritage.org    I submitted two enquiries concerning 'crossover networks' and 'tri-wiring'.  Two very heIpful replies came from 'Regis' and 'Guido' as follows:

'Regis'  offered his views on 'do_s and dont's'

Hi Don,
The L150A is a great speaker, much better than the L-150 as it utilizes the same crossover as the L-112. This crossover is a variant of the 3113b. When you're talking about tri-wiring the crossover, do you really mean that you want to tri-amp the speaker? In other words, are you going to use three separate amplifiers to power the individual components (tweeter, midrange and woofer)?

Because if you do this, you're not going to need a stock internal crossover. What you're going to have to have is an external electronic crossover that separates the outgoing signal from the preamp or reciever into the separate amplifiers and then to the individual components.

I really don't think you have to go this route. Dirty and corroded connectors are easily sanded and cleaned (pull one at a time, if you have any doubts about where they should go). The biggest improvement will be in cleaning your adjustment controls with a high quality electronic cleaner spray like Deoxit. Dirty or old controls will have a huge effect on the sound. You have two options, cleaning them and replacing them. If cleaning doesn't work, you'll have to replace them.

Try to squirt the deoxit into the vent holes (if the pots have them, some models don't). If that doesn't work, you'll have to pull them out of the cabinet to get at them. Be careful, this stuff has a lubricant as well and you don't want to get any, anywhere else as it will stain the flat black with a grease mark.

To remove the adjustment pots, you have to carefully remove the L-150 foilcal/foilcals off the speaker with a thin, flexible putty blade and blow dryer (see link below). Heat up one corner and gently work the putty blade under it as the glue softens. Keep gently working the whole decal off. Do not hurry this process, because those nice looking silver and black foilcals are no longer made or available.

Three screws hold the crossover controls to the cabinet. If the knobs block removal, you may have to pull the knobs off. Pull the adjustment knobs off with a fine pair of needle nose pliers, cushioning the knob sides with a fold of cloth (so you don't damage them) while you pull straight out Once you pull the knobs, the board should come out.

Unplug all the wires, they are color coded anyway, but it's always a good idea to write down what goes to what. You can remove the caps off the back of the controls and attempt to clean them. Use a pair of channel locks or adjustable jaw pliers to wiggle the caps off. You can then get at the wirewound resistors inside.

You can test them out with an ohm meter and see if they drop out intermittently (you should have a smooth ohm reading from minimum to maximum turn of the control, if it jumps around, then the potentiometer is bad). You can get replacements at Parts Express for a reasonable cost. I replaced mine with newer JBL pots and it now has a much-improved sound.

Feel free to ask me any other questions you may have.

Removing FoilCals

http://audioheritage.org/vbulletin/s...light=pots+ohm

Sincerely,

Regis

I noticed that in my models,  Ser.Nos 16379 and 16359, it is possible to remove the level controls by pulling off the knobs and removing the nuts retaining the potentiometers.  So no need for that potentially ruinous removal of the decals.
 
During my time of using the speakers, I'd settled on the fully clockwise +3db position, so I thought I'd eliminate the controls and just use a fixed resistor.  I'd also taken to heart the suggestion in the UK hi-fi culture that push-on connectors were bad news - especially so many of them with a fully removeable network printed circuit. That meant soldering all the connections....hmmm....risky stuff.
   
The N150A crossover network

'Guido' directed me to  http://manuals.harman.com/JBL/HOM/Technical%20Sheet/L150A%20ts.pdf which shows a pretty comprehensive data sheet for the N150A.   Being a lapsed electronics engineer from the analogue days  ( I'd  spent my graduate years 1955/56  in the audio department of EMI Music developing stereo disc cutters and loudspeakers),  I thought  I'd amuse myself and shake the dust from memories of  circuit analysis  programs.   In those days it was 'ECAP' running on a Hewlett Packard 9824, but now of course that's completely obsolete and replaced by numerous versions of SPICE running on a PC.  I used the demo version from  http://www.penzar.com  to analyse the electrical  circuitry, in a number of stages.  

This shows the network diagram from the JBL handbook, thanks to Guido.
 
            

It's interesting that the level controls appear to be simple potentiometers, so that the impedance presented to the divider network will change with the setting.  Missing from the diagram are the equivalent circuits of the drive units, so it's pointless trying to analyse the responses.    But....I have a sentimental collection of JBL brochures, and in one these, SSL250/B460 dated 11/82, for two of their flagship domestic models, the L250 and B460, are electrical and acoustic data for the drivers.  In this brochure, the writer points out the disadvantages of  slider pots, and indicates the superiority of fixed bus-bar controls.  He also refers to networks  compensating for the frequency behaviour of the impedances of the drivers.

It's a good approximation to represent the electrical circuit of a driver by a lossy parallel resonant circuit in series with the dc coil resistance and an inductance.    Using simple circuit theory and and a bit of guesswork, it's possible to select the values which fitted the measurements of dc resistance, the frequency and Q of the resonance and the impedance at high frequencies.

Although the driver type numbers  of the L150A and L250  differ slightly, they're probably close enough for the present purposes.
The high frequency driver

The unit in the L150A is the 044.  In  the L250 it's 044-1.  I derived the following equivalent circuit for the 044-1.  Note that the resistor R4 is a dummy resistor to calculate the input impedance by calculating the voltage at the junction with the dc resistance R1.

                                      

The following image shows the  voltage at the node joining R4 to R1 and is a measure of the input impedance -10mV corresponds to an impedance of 10R.

                        


In a similar manner, it's possible to represent the acoustic output by a high-pass filter with a 12db/octave cut-on response.  By setting the design input impedance at 1Kohms it's possible to add this to the electrical equivalent to calculate the acoustic output without the need for a buffer amplifier.


                                 

            
                           



The mid-range driver

The unit in the L250 is the LE5-11. That in the L150A is the LE5-12.  As before, we can calculate values for the equivalent electrical circuit and the acoustic output.                                       
       
                                                                                                                                                                                                                                                       

 






This following image shows the electrical equivalent circuit to create the acoustic response:
                      

This following image shows the acoustic output:




The Total Response
The output from the mid-range filter is inverted and added to the sum of the outputs from the lowpass and highpass filters to give the total response.  In practice, I was unable to perform a full simulation because the demo version of TopSpice is limited to a modest number of components.   I didn't feel strongly enough about it to buy the full program.   For those who have the program, and are probably more fluent with it than I, here's my script.

                                                   JBL N150A.CIR
*
* CONSTANT VOLTAGE SOURCE FROM AMPLIFIER
VS    24 1  AC    1.0
*
* COMMON LEAD IMPEDANCE
R24 24 0  0.01
*
* LOW PASS FOR WOOFER
L01 1 2 2.5E-3
R01 2 0 51
R02 2 0 6.5
***********************************************************************
* BANDPASS FOR MID-RANGE
C01 1 3 13.5E-6
L02 3 4 0.75E-3
C02 4 0 6E-6
R03 4 5 2.4
R04 5 0 20
R05 5 0 6.2
*
* EQUIVALENT ELECTRICAL CIRCUIT OF LE5-12 MID-RANGE DERIVED FROM BROCHURE
*R08 5 7 5.5
*L04 7 8 .25E-3
*R09 8 9 3.3
*L05 9 0 5.5E-3
*R10 8 10 3.3
*C04 10 0 55E-6
*
*EQUIVALENT ACOUSTICAL CIRCUIT OF LE5-12 MID-RANGE DERIVED FROM BROCHURE
C07  5 17 480E-9
L09 17 23 1200E-3
R16 23 0 10
R17 17  0 1000
**********************************************************************
*
* HIGH PASS FOR TWEETER
C03  1  6  4.0E-6
L03  6  0  0.3E-3
R06  6  0  6.2
*
* EQUIVALENT ELECTRICAL CIRCUIT OF 044-1 TWEETER DERIVED FROM BROCHURE
*R11  6 11 6.2
*L06 11 12 0.1E-3
*R12 12 13 15
*L07 13  0 1.0E-3
*R13 12 14 1.5
*C05 14  0 1.5E-6
*
*EQUIVALENT ACOUSTICAL CIRCUIT OF 044-1 TWEETER DERIVED FROM BROCHURE
C06  6 15 120E-9
L08 15 16 300E-3
R14 16  0 200
R15 15  0 1000
**********************************************************************
* INPUT Rs FOR ADDER 1
R20      2     18     10K
R18  15  18     10K
R23  21  18  10K
* FEEDBACK R
RF1     19     18     10K
* OPAMP
XOP1    0  18 19 OPAMP1
*
*  INPUT R FOR INVERTER OPAMP
R21  5 20  8K
* FEEDBACK R
RF2     21     20     10K
* OPAMP
XOP2  0  20 21 OPAMP1
*
* OPAMP MACRO MODEL, SINGLE-POLE
* connections:      non-inverting input
*                   |   inverting input
*                   |   |   output
*                   |   |   |
.SUBCKT OPAMP1        1   2   6
* INPUT IMPEDANCE
RIN    1    2    10MEG
* GAIN BW PRODUCT = 10MHZ
* DC GAIN (100K) AND POLE 1 (100HZ)
EGAIN    3   0    1 2    100K
RP1        3    4    1K
CP1        4    0    1.5915UF
* OUTPUT BUFFER AND RESISTANCE
EBUFFER    5 0    4 0    1
ROUT    5    6    10
.ENDS
*
* ANALYSIS
.AC DEC 500 1e2 1e5
.PRINT    AC  VDB(2) VDB(5) VDB(6) VDB(19)
.END
                              
 Of course, all this is highly idealised - no account is taken of the time delays created by the physical spacings between  the contributions of the three drivers and the listener.






Effects of common lead impedance

I tried a common lead impedance of 1R in the returns, but could find no significant effect on the voltages at the drivers.  I'd half expected that  there would be some breakthrough, but that seems not to be the case.  Hmm?

Practical work

I disconnected the level controls and soldered all the remaining connections at the board, including two replacement  8.2ohms 10W wirewound resistors.  To create the triwiring configuration, I  fitted two further sockets.  I moved the commons connections from the network board to the new sockets.  Of course, there still are the push connectors on the LE5 and 044 drivers, and the sprung knife connectors on the 128H driver.  I ran separate commons from my SUMO Polaris amp using  a large current capacity 4-core cable. I was able to do an  A-B test between single wiring and tri-wiring, but  I couldn't hear any difference.  Well, there you go..it was fun anyway.
  
Having gone this far, it seemed like an opportunity for some more fun,  making some spectral response measurements.  It seemed worthwhile measuring the responses of the crossover and the outout from the complete loudspeaker system.  This latter is famously difficult to do in a domestic environment, because of the multiple reflections that occur.  Measurements with 'pink noise' i.e. constant power per octave are rather easier and can give useful results.   I tried several freeware/shareware/licensed applications which offered sine-wave and white-noise generators, but in the end I bought from trueaudio.com the 1/24 octave version of the signal generator and spectrum analyser.  I was unsuccessful in using the signal generators to create useful audio signals from the loudspeakers, because of feedback.  I'm probably doing something wrong?

I used instead a CD of a very wide range of sine and noise signals.  I was thus able to do all sorts of acoustic measurements without using the internal generators.  But for electrical measurements I found the signal generator from http://www.electronics-lab.com/downloads/pc/index.html to be extremely useful.

Electrical Results

This following diagram shows the theoretical results plotted on a log scale:                        

                          
                            

To  measure  the frequency responses of the three filters I played four times the 27seconds pink noise band on the test CD and averaged the FFT outputs  over 100 samples. And these are the results:

                   

                    

                               


Generally, the comparison is good.  But there are interesting spurious peaks both in-band and out-of- band.  I sort of suspect some coupling between the inductors in the mid-range and tweeter circuits - they seem rather closely spaced on the network card.

With the benefit of hindsight, I should have made the measurements first, and then fitted the electrical models -  I recall that when fiddling with the values of the equivalent electrical circuits for the drivers, I observed that there were values which created the pronounced peak in the mid-range and the slight overshoot in the tweeter.  I don't really want to try this again - the effects are quite modest.
 
                         
Acoustic Results 
I placed the speaker asymettrically in my sitting room, which in plan is two areas, 12ft x 10ft, and 11ft x 9 ft with 8ft 6" ceilings  ( this is an old English Victorian house!).  I set up a simple PC microphone at 1m distance from the axis of the mid-range and measured the output in 1,1/3,1/6, and 1/12 octaves for the bass and mid-range units with pink noise.  The following 8 images are with 'peak value' setting.




 







Even with pink noise, interference effects are very obvious.  There does seem to be also a pronounced bass peak.

For the mid-range:








  

Interference effects still seem present.

These following images show the results with all the drivers present - it's probable that the  HF response cannot be trusted, because it was only an uncalibrated PC microphone.




I then moved the microphone to 3 metres and temporally integrated the signals.  There are some really useful facilities in the TrueRTA software!




The curves are now much smoother and the bass peak seems to have disappeared.  Mind you,  I've always been aware that there is a slightly heavy bass emphasis.

Conclusions

It looks and sounds as if these beautiful loudspeakers are restored to their original quality.  Soldering the connections definitely had a positive effect.  But rather disappointingly, all the effort that went into tri-wiring them seemed to do nothing that I could hear.   But it was all good fun!


Update 5 June 2006

Guido has pointed out that my assumption that the level controls are simple potentiometers is incorrect, and that they are actually constant input impedance L-pad networks.  This would certainly make more sense for a JBL loudspeaker - and is more consistent with the network diagram.

With the full range of frequency responses now available, I was bound to rerun the model and adjust some of the empirical values to better fit the measured results.  Note that I had replaced ( er...mistakenly) the level controls with 8.2R each.

This following image shows the new calculated results for the networks.


                             

There is now a very gratifying fit between the calculated and measured values.

This following image shows the results of putting the level controls at their maximum setting i.e zero shunt impedance.



                   


The trends are rather more pronounced.  It seems therefore that the electrical circuits do have quite a marked effect on the responses of the filters - on paper anyway.

Updated 11June 2006

Thanks to John Murphy of TrueAudio, who created the RTA spectrum analyser and pointed out that I had not correctly set the configuration in my sound card, I am now able to make
wide dynamic range comprehensive measurements of the network.





                   

So, each section has interesting but almost certainly trivial anomalies
1.  The too gentle falloff in the bass.
2.  The low frequency peak in the tweeter
3.   The high frequency peak in the midrange.

All very strange.  Comments welcome...things that come to mind are resistances in the inductors, and maybe some common resistance in the printed circuit of the network.