Part 4 – Research Methodology

Before embarking on any research project, it is important to define methodology.

Methodology (research methods) – the design of a particular research study: a set of procedures according to which it is undertaken, including techniques of data gathering and data analysis (Dictionary of Media and Communication).

I will be measuring tonearm resonance multiple times, therefore the methodology must remain unequivocal and fair for each configuration.

1. Test Record

Tonearm vibration occurs at a specific sound frequency. For my testing, I will need a record with a 20Hz to 20kHz frequency sweep track. Playing this special test signal will allow me to measure the relationship between sound frequency and tonearm vibration.

It should be stressed that I will only be measuring vibration level and not sound quality. This is because vinyl is not a perfect recording medium and any sound distortion may not necessarily be caused by vibration.

I will be using Ortofon Test Record where the first four tracks are frequency sweeps from 800Hz to 50kHz on both channels. The Ortofon record is missing the 20-800Hz frequency which I’m going to test with HiFi News Analogue Test LP Cartridge record. Apart from using test signals, I will also check resonance on a standard record – I have picked Pink Floyd ‘The Wall – C1 Hey You’ (for no apparent reason).

This is when things start to get a little trickier. How could I measure resonance? I will somehow have to detect, record, map to frequency, and present vibrations on a graph. All this could take at least as much work as completing an engineering thesis!

2. Vibration Sensor

There are various vibration sensors available. The mass-market 3-axis sensors are commonly used in toys and mobile phones. Unfortunately, these are cheap, generic, and…of little use to me.

What I need is a professional sensor with a large sensing bandwidth, fairly small measuring range, and wide vibration frequency range.

I have found this one:

  • Weight: 1 gram – not much, which is good as it adds weight to the tonearm edge, increasing its effective mass.
  • Number of axis: 1 (Z).
  • Measuring range – the manufacturer provides two values which is a bit confusing:

     a) 20m/s^2 found in the product specification (so the sensor should detect +/- 2,038g, assuming that g = 9,81m/s^2).

     b) Measuring range value +/- 50m/s^2 (5,09 g).

I will go with the specified +/- 2,038g but the range shouldn’t make any difference as I’m only detecting resonance.

  • Sensitivity: 20mV/m/S^2, which gives 196,2mV/G at 160Hz (the higher the value, the better the sensitivity).
  • Frequency range: 10Hz to 15kHz
  • Power supply voltage: 0-33V (offset voltage: 1,65V).

The sensor will be glued directly to the headshell surface.

3. Microphone

Why microphone? I would like to be able to simultaneously monitor vibration level (acceleration) and output signal frequency.

I won’t be measuring sound quality so the mic doesn’t need to be fancy. Anything with the 20Hz to 20kHz range will do.

Both the microphone and the sensor have analog outputs which will require employing an appropriate data conversion method.

4. Microcontroller

This should be a little easier as I had previously used microcontrollers in LED motors, switches etc.

Microcontroller will collect and plot data on a graph from both the vibration sensor and the microphone (in fact, the mic is a sensor too).

It sounds easy in theory but is more challenging in practice. I encountered and resolved the following issues:

  • To accurately plot sensor data onto a line graph, I had to change analog signal into digital signal using 12-bit resolution ADC converter. I went with an internal converter as external ADCs (even 16-bit multichannel) are not that efficient.
  • Microcontroller must be fast – 16MHz is definitely too slow. This is because converting single microphone signal into frequency signal is not that easy. Microphone generates sinusoidal signal of certain intensity and waveform, therefore an algorithm is required to transform that into wide range (20Hz to 20kHz) frequency information. As this task exceeded the capabilities of standard accelerometer sensors, I had to apply the Fast Fourier Transform (FTT) algorithm (look it up if you’re feeling brave). Apart from an effective algorithm, fast conversion of signal from its original domain to a representation in the frequency domain requires high computing power. In the end, I settled on a 84MHz microcontroller (as strong as Pentium 90 processor back in the day, if you’re old enough to remember).

5. Microcontroller Driver Software

I programmed the controller to:

  • Sample microphone signal (128 times) after 12-bit conversion.
  • Transform accelerometer signal into 12-bit data.
  • Plot the highest values from 10 of the above waveform readings on a graph.

I normalised data values and presented time on the X-axis. Sound frequency and accelerometer movement (in g, where 1g = 100kHz) are shown on the Y-axis.

Graph values will not exceed 0,4g.

One more important thing – I want to avoid unnecessary vibration generated by microphone through speakers, especially at low frequencies. For that reason, I will disconnect the speakers and place headphones near the mic to monitor its signal. The test is quiet and may seem a bit weird to onlookers  – ‘I can’t hear anything, why are you looking at the screen like a madman?’

Exactly. Madness!

What was the test result? I sampled 800Hz to 20kHz signal and analysed both vertical and lateral vibration, using sensor attached to the centre and then to the side of the headshell.

The test was carried out on the most recent version of BennyAudio Immersion turntable supplied with Denon DL103R cartridge.

The above cartridge model was used for the High Fidelity magazine test of BennyAudio and performed really well in low and medium but not necessarily (IMO) in high frequencies.

Blue line – sound frequency. You can see it climb up to approximately 15kHz. I will increase it to 20kHz by compromising on quality. This is about being able to capture the common vibration point.

Yellow line – lateral vibration. The graph shows approximately 0,12g at 10kHz (10,000Hz = 0,1g, so I multiplied g by 100,000 for graph scale).

Green line – vertical vibration, inverted for clearer comparison.

You can see that lateral and vertical vibrations occur at the same moment in time, therefore from now on, I will only analyse the vertical vibration values. I may repeat the lateral vibration analysis later.

I didn’t sample frequencies above 800Hz but will get onto that at some point. Looking at the graph, it’s clear that first significant vibration occurs at 8-9kHz, subsides for a moment, and then rises again to higher amplitude (up to 0,37g at 12-13kHz).

Make sure to check out the next blog entry for more performance testing, including tests on my old Project Debut Carbon turntable with different cartridges (2m RED, DL103R etc).