Beomaster 8000: New Signal Path Opamps and Amplifier Performance Characterization with a QA400 Audio Analyzer

Inspired by Sonavor's recent effort to characterize the amplifier performance of a Beomaster 8000 after replacing all signal path opamps, my Australian customer asked that the same would be done to his 8000 before sending it back (after upgrading its circuit with a muting function for the FM section while the Phono input is selected to eliminate crosstalk).

And I am glad he asked. It turns out that it is a great idea to replace the opamps when performing a full restoration of the Beomaster 8000. My measurements yielded a 14 dBV improvement of the THD (Total Harmonic Distortion) performance on the left channel after implementing new opamps.

This shows the preamplifier and input selector board with the original opamps in place:

and after replacing them with new LF353 units (except the phono input, which was replaced with a low noise LM833 type):

I used IC sockets to spare the opamps the stress of soldering. Probably not necessary, but then why not. I did the same for the control panel PCB. This shows it with new LF353s in place:

After putting the control panel back together I performed a series of measurements to see if the amplifier performance would measure up to the values given in the service manual. It turned out that such a comparison is difficult to make due to the inherently different measurement methodology used in the 80s and with the instrumentation that is available today. But I think the measurements below show that this 8000 is now in good shape.

Let the fun begin:

I use a Quant Asylum QA400 audio analyzer for such measurements. Essentially this device yields a Fast Fourier Transform (FFT) of the audio signal that is put into it, and it can perform some measurements (i.e. calculations on the FFT spectrum) based on that data. The measurements include power, total harmonic distortion, signal-to-noise, and frequency response. 

Let's have a look at the output spectrum measured with a 0 dBV 1 kHz signal at the Tape 1 input of the Beomaster:

The dBV values essentially give the amplitude ('level') of the signal relative to a standard 1Vrms signal V0. Since level=20*log(V/V0), a signal change of 20dBV corresponds to a 10x change of the amplitude. As an example, if the measured amplitude of the 1kHz signal at the output of the amplifier is 10 Vrms with an input signal of 1 Vrms from the waveform generator, then level=20*log(10/1)=20*1=20 dBV. If the output signal were 100 Vrms, the level gain would be 40 dBV and so on.

Looking at the spectrum above, you probably wondered why the 1 kHz peak is only at about -8 dBV, while the amplifier operated at a volume setting of 5.0, which is close to the maximum output amplitude it can muster. The reason is that the measurement was performed via a voltage divider that was connected as load at the output of the amplifier. The divider was built from two 4 Ohm 50W power resistors and a 0.1 Ohm 3W resistor in series. This shows the setup:

Since the Beomaster is able to produce 100W output power into an 8 Ohm load, the resistors need to be mounted on a heat sink. I used a RF amplifier can that I had laying around from another project. The Y-shaped red wires that are soldered to the small 0.1 Ohm resistor connect to the BNC jack on the right side of the RF can, which then is connected to the QA400 input. The resistor chain is connected on the left to the speaker jack of the Beomaster. This setup guarantees that the QA400 input never gets more than about 1 Vpp. But it also means that the output amplitude that the QA400 'sees' is only 1/81th of the actual amplitude that is applied across the 8.1 Ohms. 

Using the above level=20*log(V/V0) formula, we can calculate that the level difference due to the voltage divider is about -38dBV (20*log(1/81)=-38.17 dBV). This means that the 1 kHz peak in the above graph would be at ~-8+38dBV=~+30dBV. I could have shifted the peaks in the graph, but since all measurements that are performed are inherently differences between two levels, this shift really does not matter, i.e. in the following all levels are just as they came out of the QA400, i.e. -38dBV lower than the real signals.

So what do we see in the above spectra? Mainly the THD spectrum to the left of the main 1 kHz peak and some noise. We can see that the 1st harmonic at 2 kHz is about -84 dBV weaker than the main peak. According to the above calculation, this means that the amplitude of that distortion is about 10^4 (= 10,000) times weaker than the main signal. Well below what a human ear could notice. The -84 dBV value is close to the THD measurement result of the QA400, which came in at about -82 dBV (=0.008%). The measurement is a bit worse than the 84 dBV value determined from the graph since there are higher order THD peaks that add to the total distortion. This was the only measurement result that changed before and after replacing the opamps in the signal path. My initial measurements of the left channel yielded a measly -68 dBV THD value, and after the opamp exchange this was improved to -82 dBV. So I think it is a great idea to replace all opamps when the boards are upgraded with new capacitors, just to be on the safe side.

So how does this value compare to the THD value from the Beomaster 8000 service manual? The value stated there is "< 0.05%". So we could be happy and say: "Wow this 8000 is almost 10x better than the value in the manual!". Not so fast, I would think, since the manual states that the measurements were performed according to the "IHF A-202" standard. At this point I do not know what this means, i.e. we need to postpone this comparison with the stated values. But I think we can confidently say that this Beomaster is performing reasonably well and is probably within the original specifications.

Another interesting measurement to perform on an amplifier is its signal to noise (SNR) ratio. This essentially gives us a number that qualifies how much stronger the signal is relative to the noise ("hiss") of the amplifier. This measurement is a bit more difficult to do and understand since we are comparing a defined signal peak amplitude (or power) with a diffuse noise background that is composed of a continuum of frequencies spanning the entire audible range and beyond.

I played a bit with the QA400 settings and it turned out that the SNR measurement is strongly dependent on the number of samples used for the FFT transformation of the input signal. I was able to "change" the SNR value from 65 dBV to 88 dBV simply by changing the FFT resolution from 8196 samples to 65535 samples. This means that depending on the setting the SNR value changed by a factor of 10, one full magnitude. So what is happening? A bit of reading up on the internet and a semi-cryptic response from the Quant Asylum tech support suggested that at lower FFT resolutions the main 1 kHz peak spreads out over several 'frequency bins', thereby lifting the spectral power outside the 1 kHz line, which is misunderstood by the FFT algorithm as part of the noise. This means that at higher FFT resolutions this 'frequency spill out' becomes less pronounced. So we can assume that the higher value is closer to the true value, even though the true value may be even better. Unfortunately the QA400 can only go to 65535 samples, i.e. we would need to find better equipment for an answer. 

Wondering about this topic, I performed an experiment that would allow me circumventing the FFT based calculation of the SNR. I measured the signal power of the spectrum at various FFT resolutions and it turned out that the PWR value is not significantly dependent on the resolution. This makes sense since PWR measures the power of the entire spectrum, i.e. integrates over the entire frequency range. So resolution is of limited importance, as long as the signal peak is still 'caught' in one of the sampled frequency bins.

This realization enables a basic SNR measurement: Measure the PWR value with the signal present at the amplifier input and then ground the input and measure again. Subtract the two values and the SNR is obtained.

My measurements yielded on both channels -8 dBV with 1 kHz signal (at Volume setting 5.0) and -98 dBV with the input grounded (also at Volume 5.0). The difference is 90 dBV, a bit better than the best 88 dBV value measured via FFT analysis. This compares to a ">77 dB" stated in the service manual. Again, we do not really know at this point how the 77 dB value was measured, but I think we are on the safe side and can conclude that this Beomaster 8000 is operating on spec.

Another interesting measurement is the frequency response ("FR"). The QA400 does this measurement by sending a square pulse into the input of the amplifier and then measuring the response at the output. Since a square pulse contains all wavelengths, the FFT of the response yields a true spectrum of the amplifier FR. I verified this by manually measuring the transmission for a few frequencies and the FR curves were exactly matched, i.e. I think we can believe this measurement as it comes out of the QA400. This shows the FR spectrum measured on both channels:

I cut the spectrum off at 1 kHz since there was some 60 Hz noise, and the FR drop at 20 kHz seems typically referenced to 1 kHz. So what we see from the graph is that there is a 1 dBV drop from 1 kHz to 20 kHz. The manual prescribes 0.5 dBV, but of course, again, we do not know how this was measured etc...A difference of 0.5 dBV corresponds to a signal ratio of 10^(0.5/20)=1.06, i.e. the signal at 20 kHz is 6% smaller than it should be. Not very dramatic, and most likely this discrepancy is a result of the different measurement methodologies that were applied in the 80s.

Since everything was hooked up, I decided to measure the FRs for the various filters the 8000 has. This shows the spectra:

The spectra show the FR for filter 1, 2 and both active, for bass and treble sliders set to minimum and maximum, and the flat (filter button "off") response in direct comparison. We can see nicely that the treble and bass sliders allow a ~±10 dBV change of the higher and lower frequency ranges, and that the filters cut off around 7 kHz and 10 kHz at  as prescribed in the manual.

So in summary, I think we can say that this Beomaster is in excellent condition and that everything works as it should.

Beomaster 8000 Circuit Upgrade: FM Audio Muting During Use of Phono Input

After reviving the Beomaster 8000 that came for a visit from Australia with new processor crystals, it was time to do some additional work on it. My customer was annoyed by the crosstalk between FM audio and the phono input while listening to vinyls records, and a solution to alleviate this issue was requested.

Phono inputs are especially susceptible to crosstalk since they are about 100 times more sensitive than regular high level line inputs. The result is that when the phono input is selected (with a player attached or grounded inputs) the FM audio signal can be heard weakly in the background (if a station is dialed in). And that is of course not very Beolovely. The reason for this behavior is a design flaw in the Beomaster circuit: When another input than FM is selected and previously a station was tuned in the FM tuner stays on and continues blasting its signal out to the amplifier board, even though one wants to listen to a record via the phono input.

This diagram shows the solution I developed after a bit of head scratching and downloading the data sheet for the CA3189E IF stage integrated circuit.

The CA3189E chip is responsible for shifting the audio modulation of the FM signal that is received in the front end of the tuner to a lower frequency and amplifying it before the audio signal is then extracted in the detector. The CA3189E was designed with radio applications in mind that do not have a microcontroller at their disposal. The Beomaster mutes the FM audio while tuning between stations to prevent the speakers from emitting loud static noise. It does that by extracting signal strength information from the CA3189E via pins 13 (TP22) and 12, and feeding it into the microcontroller after processing in some additional circuitry. The microcontroller then decides when to ramp the output volume down using the main volume attenuators. 

Alternatively, this chip can do this on its own via its audio mute input (pin 5). The CA3189E data sheet contains a circuit that shows how to do this if you are interested. In the Beomaster 8000 design this pin is simply grounded via R37 to permanently turn the signal output on. 

The circuit on the data sheet in combination with some measurements of the signal strength related output of pin 12 (it seems to be close to 0V between stations and at about 5V with a strong FM station dialed in) suggested that pin 5 could be controlled directly via the microcontroller signal that is fed into the two AD10/278 input selector chips on the preamplifier board. The AD10/278s have three inputs for selecting FM, Tape 2 and Phono. When one selects one of these inputs on the control panel of the Beomaster 8000, the corresponding input goes high (~4.8V), activating that particular signal path on the chips. So my idea was to simply connect the Phono selector line to the audio mute input of the CA3189E via R37 after severing the ground connection of R37 by unsoldering the resistor on the ground facing end. That way, whenever Phono is selected the FM audio muting circuit is automatically activated muting any FM audio in the system that could crosstalk with the phono input.

This shows the practical implementation of this scheme. I elected to make the AD10/278 connector directly at the ribbon cable coming from the microcontroller board. That way the additional connection can be removed from the amplifier board when it needs to be taken out.

I soldered the jumper wire to the plug contacts

and then modified the black cap with a bit of Dremeling to fit over the jumper wire:

Then I wrapped the plug with some electrical tape to keep the jumper wire from moving so it will not break off. This shows the plug plugged in next to one of the AD10/278:

I connected the other end of the jumper to the disconnected end of the R37 resistor after securing it with some shrink tubing to the capacitor located next to it. The solder spot was secured by some additional white shrink tubing:

And then it was time to turn the Beomaster back on. And it seems to work smoothly. FM still works and when Phono is selected there is no more FM audio superimposed. Beolovely!

Beomaster 8000: Replacing the Processor Crystals

A Beomaster 8000 that I restored a few years ago for an Australian customer unfortunately had to come back to my bench due to a malfunction of the processor board. It exhibited rapid-fire power on/off events, which are a sign for a dead crystal on IC4 (slave processor). See here for more detail on how I first ran into this issue on the 8000 that is in my living room. Anyway, here are a few pictures of how I fixed this one. This shows the original setup:

As you can see someone in the past soldered jumper wires directly to the pins of the microcontrollers (savages!) to bridge the often failing vias that connect some of the processor pins to other components on the board. I usually re-solder the vias instead of putting in jumper wires. The reason that these vias fail is poor initial soldering. In the 80s electronics manufacturing just started becoming fully automated and some outcomes were not as high quality as one would desire...

Anyway, here I had to deal with these wires since it is good practice to take the microcontrollers out before replacing the crystals. They can be charged with high voltages when you take them out of the packaging, and this can fry the ICs...and that essentially would mean that a donor Beomaster 8000 would need to be procured. I was able to remove IC3 (right) without trouble but IC4 on the left was stuck. When the wires were soldered on a lot of solder was used and some of it penetrated the lower regions of the socket. I unsoldered the affected legs of the socket and then pulled the IC out. It came out including the three soldered-on pins from the socket. Once it was out it was easy to clean the pins up since I was able to get to the solder without having to heat everything for a dangerously long time...much better for the survival of the precious IC.

I implanted a new IC socket

And then exchanged the crystals and oscillator capacitors (the original ones are 12 pF while the modern crystals need 18 pF caps):

I elected to not solder the jumper wires back on to IC4. Instead I fixed the vias. I left the wires on IC3 in place since every time one heats up a pin on an IC there is a slight chance that damage can occur...

Once it was all back together, I fired the Beomaster up for the first time, and it came on normally, no more relay rattling etc...so I hope this did the trick.

Beolovely!