Completely Re-Designed RIAA Pre-Amplifier for Beogram 4002/4 (Types 551x and 552x)

As usual, it took a bit longer than planned. The Beolover is definitely used to the concept of 'delayed gratification'!...;-).

But here it is: The completely re-designed Beolover internal RIAA pre-amplifier for the DC-motor Beogram 4002 and 4004 (Types 551x and 552x). An AC version will be coming up soon based on the same design.

When it comes to RIAA preamplifiers the important benchmarks are:

• How accurate is the RIAA de-emphasis? How closely tracks the pre-amp the standard RIAA curve?
• What is the noise floor of the pre-amp?
• How much distortion ("THD" - Total Harmonic Distortion) does the pre-amp inflict on the amplified signal?

The 'internal pre-amplifier' situation has an additional benchmark that is not shared with external or in-amplifier pre-amps:

• Is there any interference from the platter motor? RIAA pre-amps have gains up to 60dBV, which corresponds to an amplification by a factor 1000. This means that even small ripple in the power supply will generate audible signals in the output. Electromagnetic loads like motors are notorious for introducing ripple and disturbance on the power rails due to their back-"electromotive force" (back-EMF) that is caused by the generation of a voltage as the motor spins. This voltage is polarized opposite the voltage that drives the motor. Additionally, brushed motors like the DC motor in the Beogram Types 551x and 552x tend to emit significant voltage spikes due to the instant reversal of the coil polarities, which is the reason that these motors have spark suppressors in the first place.

It turned out that especially the last point is not trivial. And this is the reason it took me a while to come up with a solution that is on par with quality commercially available external phono pre-amplifiers.

My new design is a 'plug and play' solution that simply replaces the entire original output PCB. This picture shows a rebuilt original board (right) and my new design (left) in direct comparison:

The new RIAA board has the same form-factor like the original PCB, i.e. can be installed by simply replacing the original board. The new design also has the appropriate connectors for installation in a Beogram 4004 that has an additional connector underneath the keypad. I also included a connector for direct connection to the Beolover Commander remote control module if it is to be used in a 4004 together with the RIAA pre-amplifier.

The BeoloverRIAA board can also be configured as a standard output board bypassing the RIAA amplification. For this reason I implemented a dual 2x9 input header and two output sockets:

The lower row is used for the RIAA signal path, whole the upper row serves as connector for the conventional non-RIAA path.

This shows the board installed in a 4002:

With the 'set' switches above the red standoff the gain can be selected and the B&O standard 56p input capacitance can be connected if desired. The gain can be switched between a pre-set 40dB mode, and an adjustable mode where the gain can be tuned from ~30-50dB. This can be helpful if one wants to use a vintage cartridge where one channel is a bit weaker, or to level the Beogram output with other input sources connected to the amplifier.

The filter sections of the RIAA amplifier section, shown here,

are built with quality 2% tolerance Panasonic Polyphenylene Sulfide (PPS) Film capacitors and 1% resistors. The use of PPS capacitors enables ultra-low distortion figures. PPS capacitors have an only very weakly voltage dependent capacitance, which keeps non-linearities to a minimum. The frequency response of each channel can be tuned with two trimmers, one for bass and the other for treble. This can be helpful if there is a deviation from the ideal RIAA behavior due to component tolerances or aging.

Since the Beogram only has a uni-polar power supply, the amplifier had to be built with a virtual ground to enable AC signal amplification. This posed a challenge due to the noise introduced on the power rail by the platter motor, as well as the tendency of virtual grounds to raise the noise floor of the amplifier. Only the use of a buffered precision virtual ground made it possible to achieve a noise floor that is similar to external RIAA amplifiers, and to completely eliminate motor interference.

Characterization of the BeoloverRIAA pre-amplifier:

In the following I will show some key measurement data that demonstrates the performance of this amplifier. 

Measurement Conditions:

I used a QuantAsylum QA400 audio analyzer for my measurements. The QA400 is a predecessor of the currently sold QA403 (if it is available at all). It essentially is a very low noise high resolution Analog-to-Digital converter (ADC) integrated with a quality waveform generator, which in turn is basically a Digital-to-Analog converter. This combination allows measuring the 'transfer function' of an audio device by sending a suitable input signal into the device and measuring what comes out at the other end. The beauty of it is that the QA400 can analyze the output signal via computer software that generates a Fast Fourier Transform (FFT) of the signal. This yields a spectrum that shows the output signal divided up into its individual frequency components. This goes back to the fact that any signal can be synthesized by a combination of (or 'integral' over) sine waves of appropriate frequency and amplitude. 

How were things hooked up? The input signals from the waveform generator were fed into the amplifier via BNC-to-minigrabber cables (which is the reason that there is a bit of 60Hz interference in these spectra). The output signal was fed back into the QA400 inputs by connecting the standard RCA output cable of the Beogram with RCA-to-BNC adapters to the inputs of the QA400. 

During all measurements the platter was turning and the arm was lowered next to the platter with a cartridge mounted. This was done to characterize the amplifier in-situ under realistic operational conditions.

Frequency Response Measurement - Does the Amplifier match the standard RIAA de-emphasis curve?

The frequency response of any RIAA pre-amplifier is key to a faithful reproduction of the sound stored on vinyl records. It is in my opinion the most important characteristic since it really is the only characteristic that affects the 'sound' of the pre-amp. From a scientific point of view any discussion of the 'sound of an amplifier' are in reality a discussion of the frequency response, and its deviation from the ideal behavior.

I measured the frequency response using the "FR" function of the QA400 audio analyzer. I used the 'white noise' stimulus, which basically sends pulses of white noise into the input of the amplifier and simultaneously measures the Fourier spectrum of the transmitted signal at the output. 

Since white noise is composed of all frequencies at the same amplitude, the output spectrum can directly be matched to a standard RIAA curve since it is a direct 'fingerprint' of the transfer function of the amplifier.

I obtained the standard RIAA curve in conveniently tabulated form from here and then loaded it into my data processing software. This shows the result of the measurement for both left(red) and right(blue) channels in comparison to the dotted standard RIAA curve:

Essentially both right and left curves are identical and they match the RIAA curve very well. There is a bit of a 60Hz noise due to the partially unshielded input connections. I do not know exactly what the 'wiggle' between 20 and 30Hz means, but I think it is a measurement artifact. 

Total Harmonic Distortion (THD) Measurements:

Total Harmonic Distortion (THD) of an amplifier represents the percentage of the output signal that is composed of harmonics of the input signal that were not present at the input. The origin of such distortions is usually the non-linearity of components in the amplifier circuit. Especially capacitors have non-linearities since their capacitance changes with the applied voltage. This distorts sine waves during the amplification and filtering process, and that means that the FFT of such a signal will show higher harmonics related to the input signal. In order to see and quantify this measurement, a pure single frequency is sent into the amplifier and then the output signal is evaluated for the presence of harmonics. An evaluation and summation of the individual intensities of the harmonics and subsequent comparison with the principal frequency yields the distortion as a percentage of the principal signal.

This graph shows the THD measurements I performed with the pre-amplifier:

a) Measurement of THD using the QA400 internal waveform generator (red spectrum):

This measurement basically corresponds to the above 'best case scenario', where a high quality almost ideal 1kHz signal is fed into the amplifier and the response is measured. Since the input signal is nearly perfect, all distortions that are measured can be assumed to have been introduced by the amplifier, and that makes it easy to calculate the THD as a percentage of the primary signal. 

The red spectrum was measured with the QA400 internal signal generator set to 1kHz. I adjusted the output volume of the generator to get an exact 0dBV (=1.0Vrms) output signal from the RIAA pre-amp. This seems to be the standard measurement condition employed by most amplifier manufacturers, although they are often quite circumspect with regard to how their measurements were done exactly.

It is obvious that there is only a very limited degree of higher harmonics present in the red spectrum. The THD evaluation function of the QA400 reported a THD value of only 0.012%. This is similar to the numbers stated for many commercially available external RIAA preamps.

b) Measurement of THD playing a 1kHz tone on a test record (blue spectra):

I wanted to understand what this THD result meant in practical terms, and so I wanted to see how the above best case scenario would compare with an actual 'record playing situation'. A while ago I had purchased a copy of the 'Ultimate Analogue Test LP'. It features a "1kHz reference tone 7cm/s lateral in phase (mono)" track on side one.

I think it is supposed to represent a reference signal representing the typical maximum signal level as encountered on many records. I played this track with two vintage B&O cartridges that I use for regular listening, and that in my opinion still sound pretty good.

1) MMC20CL cartridge:

The dark blue trace was measured with a MMC20CL. We see that there is a significant amount of harmonics in this spectrum. The QA400 THD evaluation yielded 0.95%, i.e. about 86 times more distortion than it determined for the red curve. 

2) MMC20EN cartridge:

The second measurement with a MMC20EN cartridge was even worse, coming in at 1.2% distortions.

If you wonder how the QA400 determines these percentages, this can be rationalized by looking at the ratio between 1kHz principal peak and the first harmonic at 2kHz. For the test record measurements a difference of about 40dBV is read from the graph. 40dBV corresponds to a 100x signal amplitude difference, which means that the 1st harmonic signal has about 1% Vrms amplitude compared to the 100% principal signal. 

Looking at the red curve, a difference of about 90dB is read off the graph, which corresponds to a factor 31622 between the signals. Expressed as a percentage it would mean 0.03%, which is pretty close to the 0.012% number the QA400 spat out. This may give an idea about the potential error bars on these measurements, since the 1st harmonic peak is pretty close to the noise level, i.e. there is probably some degree of uncertainty in these measurements, especially when the percentages get small.

Luckily, it does not really matter whether the amplifier distortion is 0.012 or 0.03% since the distortion introduced by the vinyl and the cartridge is about two magnitudes larger. The above measurements just mean that the amplifier is certainly good enough in the THD department for a faithful reproduction of vinyl records.

Determination of the 'noise floor'

On to the last set of measurements: The determination of the noise floor.

The graph below shows a series of measurements under different conditions that allow exploring what is going on:

Let's start with the green spectrum at the bottom. This spectrum shows the situation with the standard output board without RIAA amplifier. It basically represents the noise that generated by the cartridge and the wiring, which is thermal noise from electrons bouncing around in the cartridge coils and the connecting wires, which represents a small fluctuating voltage in the uV range.

This 'natural' noise floor is amplified in the RIAA pre-amp when the signal from the cartridge is routed though it. 

I also measured the noise that exclusively comes from the amplifier itself. This is usually done by connecting the input to GND. The resulting spectrum is the black one. We see a drop of ~40-45dBV between 20 and 20000Hz. This about corresponds to the RIAA gain drop between these corner frequencies, and means that the thermal white noise at the input of the amplifier is simply amplified by the frequency dependent gain of the amplifier.

 

The next measurement was the blue spectrum. It was measured after removing the GND connection and lowering the arm next to the platter with the platter running. This curve is similar to the black spectrum at lower frequencies, but up to 10dBV higher towards 20kHz. This is probably a result of the thermal noise from the cartridge coils being added to the signal.

A big question of any 'internal' pre-amp design is: Does the platter motor add unwanted noise into the output signal of the pre-amp? The good news is that my design approaches are able to completely quench any noise coming in from the Beogram power rails. The ripple rejection of my approach is very strong. Earlier versions of my design had sizable motor peaks in their noise spectra between 30 and 40Hz and between 100 and 120Hz. The motor noise was clearly audible at high volumes. Not so anymore with the current design. The output is very quiet when listening to it with the arm down next to the platter and the volume dial on the receiver all the way up.

And then there is the red curve, which has a noise floor that is considerably higher than the blue and black spectra! The red curve was measured on the 'silent track' of the Ultimate Test LP, which is supposed to represent the noise floor of a vinyl record. I played it with the MMC20EN cartridge for this measurement. It is interesting to note that the red spectrum represents an average of the groove noise over maybe 20-30 seconds since I averaged the signal 30x with the QA400 to get a nice spectrum. This means that there are several klicks and pops from the record incorporated in this spectrum, which likely accounts for the 'features' on the low frequency end of the spectrum.

In conclusion it can be said that, similar to the above discussion of the THD spectra, the requirements for a RIAA pre-amp in the noise department are actually pretty relaxed. Since the noise floor of records is so high, it does not really matter what the noise floor of the RIAA pre-amp is, as long as it is a little bit below the 'natural' surface noise from records.

After these measurements it was finally time to give this new RIAA pre-amp a listening!

I selected Art Farmer's CTI issue "Something You Got" (CTI 7080) from 1977. I really like 'Flute Song' on Side 1. Of course this record was ultrasonically cleaned with a CleanerVinyl ProXL setup mounted to a UC-3360 multi-frequency cleaner. Together with the new pre-amp this record sounded quite fresh! I think I really like the sound of my new pre-amp. Very smooth and precise. Perfect for a high-quality CTI record! Here is an impression:

Excuse the rugged look of the Beogram 4002. It actually is the one that I just functionally restored, and which is currently waiting for a restored plinth and a rebuilt keypad. Perfect for testing the new BeoloverRIAA pre-amp!

A few more small changes to the PCB design and then the BeoloverRIAA will be available to other enthusiasts, like most of the components and parts we show on the blog!