Noise analysis
It’s not the purpose of this short review to get in a detailed technical explanations about all the possible turntable noise sources, their causes and measurement techniques. There is an excellent app note from Bruel&Kjear for that, which is a recommended read. Rather let’s work our way through identification process and make it more apparent by actually listening to what we measure.
Establishing the Base Line
First let’s do a base line measurement for establishing the noise components, when turntable is not operational. For this purpose we have stylus resting on a metal cylinder which sits on a table chassis. Here is the resulting FFT of 30 sec. capture, with infinite averaging turned on:
Not much to comment here, most prominent is the table suspension resonance at 14Hz. Here is a more detailed view without averaging:
Red trace is when table sits on suspension, black trace – when transportation screws are tightened. The inevitable compromise of table suspension is obvious here. We get better broadband noise attenuation in audio band (20-60Hz) in exchange for infrasonic 14Hz peak. It’s an open debate, but I would argue that it’s a bad deal. Later we will see why. It would be really educational to compare 3rd trace with table sitting on heavy wooden plinth. Maybe next time. Another thing worth mentioning is how sensitive these type of subsonic measurements are to seismic activity. Street is good 50 meters away from here but despite all the efforts, it takes just one car to make it to this graph!
Not much to listen here, we all know how hum sounds, but nevertheless. Let’s import this track to Reaper, normalize the level and listen, using narrow band-pass filter too better isolate frequencies of interest.
Isolating the Moving Components
Now let’s try to identify different drive components and their noise contribution to overall spectrum. It would be ideal to be able to better couple stylus to drive shaft, using something like Thorens did with their Rumplemesskoppler (rumble noise coupler?). Something to consider for next time maybe.
For now, let’s run the same setup as previous, but with platter spinning and motor turned off. For this to happen, we start the table in 78 rpm mode, let the needle rest on metal block and disconnect the power. It’s not ideal, as the platter speed constantly drops and speed dependent frequency components will drop too, but it will give reasonable general view.
Here we have the sum noise off all moving parts – platter bearing, idler wheel and motor idling. We can brake it down even further by analyzing differences between just platter spinning from inertia and drive-train engaged.
Now it’s clear, that spectrum is totally dominated by platter bearing rumble. Actually I was quite surprised by how little idler wheel contact noise there was without torque transfer from motor. Let’s listen to our frequencies of interest at 40, 110 an 180Hz.
Understanding the Motor Influence
It’s a bit more involving process. Let’s begin by capturing and averaging the overall view of a fully functioning turntable.
At first glance there is couple obvious things here. We see increase in 50 Hz mains and it’s harmonics peaks and a lift in 180Hz region, which we previously associated with idler wheel friction noise. But what’s a deal with all that inter-modulation products ? Let’s dig a little bit deeper.
Now that’s a wood you can get lost in ! Ok, let’s begin from lowest peaks of 2.55, 5.1 and 7.7 Hz. They are clearly drive-train related, but how ? Let’s analyze motor rotation transfer to a platter.
Mystery solved! This is a rotational frequency of idler wheel and it’s harmonics. Usually higher level harmonics indicates misalignment or eccentricity of a drivetrain components. Upon further inspection idler wheel was found to have a little wobble, which can be rectified on lathe. Also there are new peaks at 12 and 25 Hz, that is to be expected, as they’re spot on torque and rotational frequencies of a 4 pole motor running in a 50Hz network. Everything else is inter-modulation products between these frequencies. Here is audio of a blue trace.
That 200Hz peak needs further attention. Let’s examine the resulting waveform, after band passing.
It’s a 200Hz wave AM modulated by 2.5Hz idler frequency. Also there is some ringing apparent. My guess is platter excitation by idler friction. Could be lowered by potting downside of a platter with some silicon rubber. This way platter balance would go out the window and needed to be redone, but it would really help damping the ringing.
Using the Test Record
Track 6 on Side Two has unmodulated grooves for residual noise measurement. This is a resulting spectrum of 30 sec. averaging.
Not much has changed, as we already identified most of the noise components, but there is a new peak at 9.8Hz. Logic dictates that it must be tonearm resonance. Let’s find out. Here is a Track 2 of a Test record – 25-5 Hz lateral sweep.
What a mess… It seems like this dude just sits there, randomly pronouncing numbers and turning frequency generator knob up and down. Even though I totally understand how problematic is cutting tracks bellow 20Hz, that’s not what you would expect for 50$. Anyhow, it’s audible from pilot tone pitch variation at 10Hz, that resonance is starting there. Decreasing FFT bin size and turning infinite averaging on produces more usable picture.
Resonance peak would be around 9.5Hz it seems, but because frequency jumps in between this range, it’s not excited enough. That agrees to our previous findings from silent track and shows, that it’s not necessary to have a specialized test record for tonearm resonance detection. It’s always there, no matter what you play. Here is simple tap on resting stylus in time domain.
Quite good overall damping from this light tonearm and loose counterweight. It means, there shouldn’t be very pronounced resonance peek anyway.
Interpreting the Results
First conclusion to make here is that we are not limited by a noise floor of a test record. There is a slight elevation in 50-150Hz region, but I guess it’s more related to the actual measurement place (platter vs. chassis) then anything else. So, what does it all mean in terms of numbers? What rumble specs does this table has ? Well, it depends on how you look at it. Or what weighting filter you apply.
IEC98 and similar standards for rumble measurement states two weighting filters. You can download A and B text files with their exact characteristics. These files were used to FIR filter 30s of silent groove track and resulting spectra was recorded.
There isn’t a lot of difference between A weighting and plain spectrum in this case. But if we had tonearm resonance lower, it wouldn’t made to the published specs. So A weighted rumble specs most always will have nothing to do with a quality of a motor or bearings. It will only show record surface irregularities, boosted by tonearm and table suspension resonances, and will solely depend on tonearm/cart combination used. B weighting, on the other hand, will mask most of the motor and bearing noise, again leaving user clueless of the issues at hand. That’s why spectral analysis is the only useful specification, when defining turntable rumble. It clearly shows where improvements can be made for manufacturer, and what can be solved by user, when selecting different tonearm/cart combo.
It’s funny to note how these figures correlate with ones made some 40 years ago. Above are reviews from popular magazines. Different setups and unimaginably different hardware, but same numbers. Always gives me warm feeling inside.