Characterization of a MMC 6000 Cartridge Restored by Tonabnehmerservice.de in Comparison to Two Good Sounding Original Cartridges

I recently restored a Beogram 4002 (Type 5523) for a customer in Louisiana. The customer also wanted a newly rebuilt cartridge. He settled on getting a restored MMC 6000 from tonabnehmerservice.de. Franz Josef at tonabnehmerservice has provided a number of cartridges to some of my customers over the years, and they all seem pretty happy with his work. This particular cartridge is one of his top-of-the-line restorations with a beryllium cantilever and a nude Shibata diamond. 

In this case, my customer asked Franz Josef to send me the cartridge directly so I could check it out before sending it along with the restored Beogram. I received the cartridge last week. It arrived in a little box fitted with a proper plexiglass protector:

I carefully extracted it:

Looks very nice! This shows the beauty playing on the restored Beogram:

The test setup for cartridge characterization is simple. All I had to do was plug the Beogram DIN5 output plug into proper adapters to interface it with the BNC inputs of my new QuantAsylum QA403 audio analyzer. 

A few years back I wrote a couple blog posts about this type of measurement. This one discusses the frequency response measurement and this one the 'trackability' measurement.

Let's focus on the frequency response first:

There are several ways to do it. In the old days, they used test tracks that actually played a frequency ramp from 20 to 20000 Hz. The cartridge was hooked up to an AC voltmeter, and a voltage vs. time plotter recorded the signal as the test track played. Add a proper scale to the time axis that translates the time it took to play the track into the proper frequency scale, and the frequency response becomes visible. 

These days we can play a noise track and use an audio analyzer to record a Fast Fourier Transform Spectrum (FFT). A FFT spectrum shows the volume of all individual frequency components of the measured audio signal on a frequency axis. A clean 1kHz test tone would, for example, show up as an isolated peak at 1kHz. A noise signal would look more like a line parallel to the frequency axis since noise is a composition of many frequencies.

I used Tracks 2 and 3 on Side 2 of the Analogue Productions "Ultimate Analogue Test LP" as noise tracks and played them directly into the QA403 analyzer. 

Tracks 2 and 3 are "Pink Noise" tracks: Unlike white noise, pink noise distributes its energy equally per octave rather than per individual frequency. Because of this, its Fourier transform (FFT) is not ideally a horizontal line along the frequency axis, but rather a downward-sloping line that drops at exactly -3 dB per octave (which translates to a -10 dB drop per decade). In plain English: pink noise emphasizes the lower frequencies and gradually reduces the volume towards the higher frequencies contained in the noise. White noise features all frequencies at the same level, i.e., would be a parallel line to the frequency axis.

Now let’s go from the theoretical construct of pink noise to its real embodiment on a test record. Records are usually cut with RIAA pre-emphasis, which superimposes an upward tilt of roughly +6 dB per octave (or about +20 dB per decade) to the signal. If you go to the link, you will see that the RIAA curve is not a line, but has some wiggles in it. That is a consequence of the fact that in real life electronic filters are used to produce the pre-emphasis. But for simplicity's sake, let's assume a +20dB/decade line.

Therefore, if a pink noise test track is played back from a record without using RIAA de-emphasis, it has the RIAA pre-emphasis superimposed on the original pink noise signal. In practical terms, this means that the -10 dB drop of the pink noise signal is overcompensated by the ~ +20 dB RIAA signal increase per decade. This results in a measured FFT spectrum that should roughly slope upward at about +10 dB/decade.

At this point we did not consider the influence of the electromechanical characteristics of the cutting and playback processes on the signal. The good news is that in a first-order approximation we do not have to! To understand why, consider the ‘constant velocity characteristic’ of the cutting and reproduction processes by the cutting machine and the cartridge. Both have cantilevers that end in iron pieces inside coils. On the cutting side, signal voltage changes fed into the coils translate into cantilever movements cutting the groove, while the cartridge does the exact opposite: It translates the 'wiggles’ in the groove back into a voltage signal at the coil leads. So in a first-order approximation, we can say that the influences on the signal by the cutting machine and the cartridge basically cancel out! Therefore, for understanding and discussing the measured FFT curves on the test track, we only need to consider the pink noise characteristic and the RIAA pre-emphasis (since we plug the cartridge signal directly into the QA403 without RIAA de-emphasis).

Let's have a look at what I measured:

This graph shows the frequency response of the restored MMC6000 cartridge measured on Track 2 (lateral pink noise):

I tried to 'graphically evaluate' the slope of the spectra and found maybe +9dB between 100 Hz and 10000 Hz, less than the above estimate of +10 dB per decade. This result probably has a bit of an error bar on it due to the superimposed 'RIAA pre-emphasis wiggle'. I think this probably means that the pink noise tracks do not precisely follow the theoretical -10 dB/decade energy distribution across the bandwidth. This is supported by the fact that I see the same slopes for the other cartridges I measured.

Nice is that both channels match very well. The two peaks at the bottom end of the spectrum at 60 Hz and 120 Hz are interference from the power grid. Not surprising considering the very small signals coming from the cartridge. The sharp drop-off after 20,000 Hz is probably caused by the pink noise signal itself. It probably only contains frequencies up to 20,000 Hz. Sadly, there is no detailed information about the Analogue test record available, and my emails have not been answered so far. I guess analog audio is only analog if there is a bit of woodoo in the signal path!...;-)

The next measurement I did with this cartridge was on Track 3 (horizontal pink noise):

This spectrum looks pretty similar, but there is an approximately 5 dB difference between the channels between 15,000 and 20,000 Hz. This difference means that the left channel will reproduce very high frequencies at a slightly lower volume. In the bigger scheme of things, 5 dB is not a very big difference. I certainly did not notice it when I listened to records with this cartridge. 

This is not surprising for several reasons. The most important one is probably that the 62-year-old Beolover likely has a significant roll-off at around 15,000 Hz and does not hear much of the frequencies under consideration here in the first place.

Furthermore, music has only very little information in this frequency range; only the sizzle of cymbals and similar noises are there. We also need to realize that the vertical signal contains the stereo localization (L-R), while the lateral (horizontal) signal contains the mono information (L+R). So a bit of difference in the vertical signal only affects the localization of instruments on the sound stage. Since the human brain does the locating mainly with mid-range frequencies, a small difference at high frequencies is not very significant for this process.

On to the next cartridge: My customer had sent the Beogram together with its original MMC4000 cartridge, which still sounds very good in my opinion.

Here is Track 2 played by this cartridge:

and Track 3:

Very similar spectra compared to the MMC 6000. Left and right are in good agreement in both directions.

I also measured my favorite MMC20 EN that I use for daily listening in my workshop. These are the measurements:

First Track 2:

and Track 3:

These curves also look very similar to the previous curves, but here both directions have a small difference between the channels. Here, the left channel is a tad stronger than the left in both directions.

In my opinion, this cartridge also sounds very good.

On to the trackability measurements:

I use a Clearaudio "Trackability Test Record". This record has tracks of a 333 Hz test tone at different volumes. They give the 'volume' directly as lateral deflection, i.e. the physical dimension of the 'wiggles' superimposed on the groove. The tracks start at 50 um lateral (i.e. mono) deflection, and then they go to 100 um in 10 um steps. To put this into perspective: Loud dance music tracks have maximum amplitudes around 70 um. So if a cartridge can track 70-80 um without too much distortion, it is pretty good and can play most music ever pressed on vinyl.

So I played all the tracks with each of the three cartridges and noted the total harmonic distortion (THD). THD is essentially the ratio between the fundamental peak and its harmonics. As an example, these L/R spectra were measured on the MMC6000 playing the 50 um track:   

You can see that the THD estimated from the ratio of the second harmonic to the fundamental peak is less than 3%. After measuring six sets of spectra like the above for the three cartridges for each of the test tracks, I ended up with this plot of THD vs. amplitude:

All three cartridges track fairly well up to 70 um. At larger amplitudes, the original cartridges begin to distort fairly strongly. The restored MMC 6000, on the other hand, is considerably better. It has the lowest distortion across the entire range and could even reproduce 100 um signals with tolerable distortion.

So the conclusion of these measurements is maybe that 'good-sounding' original cartridges are still good enough for everyday listening, but that for audiophile listening experiences a restored cartridge should be procured. My money is on tonabnehmer.de. I think Franz Josef does a very solid job, and you can send him your original cartridges for restoration without hesitation!