Hi Eric,

Just to make sure we're talking about the same thing: When you say "chassis", do you mean the 'main chassis', a term used by B&O for all the electronics incl. EEPROM?

Was the work carried out by an official B&O service engineer or a hobby service engineer? And they did reprogram the chassis with the respective driver data when they swapped them, right?

Eriperi wrote:

is it for instance possible that something is wrong with the microfone, for instance a constant hum which affect the measurement and causes the dsp to play everyting at a lower volume?

Since this is such a simple question, I can already answer that irrespective of my questions above:
Yes, of course something can be wrong with the mike! A hum is rather unlikely, but if it existed it would come from the electronics that's inducing this hum into the mike signal. It's more likely that one of the electret condenser mikes deteriorated. Electret mikes are fairly stable over time, but they can drift. Real high-end measuring mikes normally need to be calibrated every one or two years. An even bigger problem than aging is humidity and moisture - if there ever was some water damage on one of the microphones it would surely explain a substantially changed sensitivity, thus a smaller signal from a given source, which the DSP might interpret as reduced sound pressure and it would boost the output to compensate this.

I'm not sure whether B&O uses a simple self-test procedure for the microphone to determine whether the calibration circuit is ok before it starts the real calibration process. If not, the software assumes that whatever signal the mike delivers is trustworthy and it will try to compensate that by applying a corrective filter curve to the audio output on that speaker.

You can easily verify whether the mike is the problem: just swap the mike between both speakers and recalibrate. If the error follows the mike, the mike is the problem!

All work is done by an Official B&O engineer, in fact B&O themselves directed me to the Dealer and their engineer. When I say "we fixed" i Actually mean "the service engineer fixed it" while I was looking and trying to help him diagnose the issue. 

Yes I mean the "main chassis" which includes all electronics (Power supply, DSP, crossover, protection and 4 ICE power amps) but without the EEPROM, the EEPROM stays with the speaker because it contains the measurements of each driver in the BL5.

This is actually one thing I forgot to mention: in the speaker that plays at normal level (we also compared the SPL with a 3rd loaner BL5) one leg of the EEPROM broke and B&O programmed and a new EEPROM with the correct settings for this serial number. I asked (through the dealer) if maybe these EEPROM settings for SPL had changed, but that was not the case, B&O responded.

Currently the speaker with the issue has an older version of the main chassis (iSW could not be updated to 3.30) than the other speaker but B&O claims this makes no sonic difference and we also tried it with a newer version of the chassis.

During the repair process they had to fix also 3 connectors with broken/loose wires: the subwoofer connector and one position sensor contact for the microphone in one speaker and the white wire of the microphone in the bad speaker.

The current chassis now also has a failing ICE amp on the midbas and this is the reason why a failure in the midbas speaker itself or its wiring is one of the causes I suspect. .Because of the other wire issues we had it is not unlikely there are more potential wiring issues. The Failing Ice amp is the reason why I have not swapped the microphones, there is quite a loud 3000hz noise coming out of the speaker.

The dealer offered me to buy some BL50's as replacement, but they are not within the budget I plan to spend, the 28's are too small for my living room (110sqm) next to that sound and size wise the BL5 is close to exactly what I want

Next step is that the dealer will pick up both speakers and try to repair them in their facilities, hoping they have a better environment and can communicate directly with B&O when they are working on it.  They also send a letter to B&O because basically 2 out of 5 refurbished main chassis failed very shortly (minutes) after installing. My comment to this was: maybe they fail because something else is wrong?

Here is a screenshot of the pink noise response of both speakers I measured on a 1m distance with a calibrated mic. (the average SPL difference is about 6db, the cursor was a a unlucky position so it seems more in the screenshot)

any suggestions are welcome

Eriperi wrote:

Next step is that the dealer will pick up both speakers and try to repair them in their facilities, hoping they have a better environment and can communicate directly with B&O when they are working on it.  They also send a letter to B&O because basically 2 out of 5 refurbished main chassis failed very shortly (minutes) after installing. My comment to this was: maybe they fail because something else is wrong?

🤒 I would hazard the very same guess! 😕
But don't be too hard on those guys. They're probably just B&O-authorized service guys, which means they got training on those products but they also have to deal with 200 other, different products probably from different manufacturers throughout the year... Such specific issues they can only fix by getting back to the right support team in the B&O HQ.

I cannot help with specifics either, I'm afraid, because B&O hasn't released the circuit diagrams. The service manual only shows an overview and a wiring diagram. I agree that having 2 out of 5 electronics assemblies fail is worrisome - whatever the root cause may be. 

Your course of action, bringing it to the service center and have them do whatever necessary in cooperation with B&O Denmark, is probably the best option.
But maybe you're lucky and some other members in this forum have seen a comparable symptom on one of the Beolabs and might be able to point you in a specific direction.

The possible lucky shot was the reason I posted it.

I would expect that an error in measurement to lead to incorrect bass response (Adaptive Bass Control) but not a difference across the whole frequency band. I also thought about a bad connection in the EEPROM but this is a serial I2C STM 24C04WP which either works or doesn't work and it is unlikely that just part of the data (volume settings) is missing.

I hope they be able to fix it because I still find it great speakers and they fit my needs perfectly.

Well, from a chip level point of view, data corruption can occur, but it usually happens when you're flashing the EEPROM...and I'd expect that the BL5 firmware has a mechanism to capture such failures. At the very least it should write an error message to its error log.
The B&O guy should be able to read out that log with the Beo Service Tool and I think he can also read all stored EEPROM values. A spurious flip of bits or bytes in the EEPROM however is quite unlikely.

It would be interesting to know what driver data is actually being stored. Does B&O measure and save the full frequency response of each driver or do they just measure and save an SPL correction value for a specific driver frequency? Probably it's just the latter because a full frequency response characterization and correction was probably beyond the limit of affordable consumer electronics in the early 2000's.

Just out of curiosity, what's your measurement setup for the REW measurement you posted above? The full monty? (ie pro-grade audio setup on the PC, fully calibrated and proven or a "just-for-the-fun-of-it" setup?) Do you have measurements of the speakers before the chassis was changed?

I used a MiniDSP UMIK-1 (calibrated) to measure the response, directly USB into a Macbook Pro

The signal was a pink noise signal from a (ripped) testcd, the flac file was fed through Roon and a SP/dif signal 24bit 96 khz signal directly into the Beolab 5. I kept the BL's in their position and put the microphone on a mic stand 1m in front of the speaker at the height of the lover disc. 

I always disconnected the other speaker when measuring and put both on the same channel to cancel any possible channel difference in the source.

I found the response plot surprisingly linear, much better than expected. (other speakers don't plot that good normally)

TACT RCS2.0, was released in 2000 and used a 1999 FGPA as DSP and had full range frequency and impuls room correction capabilities. BL5 apparently has similar methodology but uses a more modern SHARC DSP so I think they should be able to do a full range frequency response correction. That said there used to be strong debates 20 years ago where most experts claimed that room correction above 300hz or so makes no sense. That point of view has changed with the introduction  of the BL90 I guess.  So maybe an individual driver correction is possible but practically not executable for any company who wants to make som profit, the fact that they do a db correction per driver is a tour de force on it's own.  But maybe there is a general correction for the driver response in the BL5's.

Eriperi wrote:

I found the response plot surprisingly linear, much better than expected. (other speakers don't plot that good normally)

Thanks for this info, Eric, and yes, the plot looks pretty good.

I would think though that the measuring distance is a bit too close. At 1m you're still in the near-field of the speaker and the frequencies of the different drivers haven't yet "mixed properly" to put it simply. 1m is a common standard for point sources but at BL5's height of approx 1m, it's more of a line source.
I think the rule of thumb for large speakers is 3 x speaker height, ie. if you measure a BL5 at 3m you'll get a flatter and a more representative measurement of the sound field. If necessary you can then normalize the measurement to a 1m reference mathematically by using the inverse-square law. 

Eriperi wrote:

So maybe an individual driver correction is possible but practically not executable for any company who wants to make som profit, the fact that they do a db correction per driver is a tour de force on it's own.  But maybe there is a general correction for the driver response in the BL5's.

Yes, and I would think that full driver correction may not even be as desirable as one might think because the driver characteristic will actually change with loudness, aging, etc etc etc. It's probably better to build the drivers with narrow manufacturing tolerances in order to get the same characteristic in different units and then only adjust SPL variances with software.