Setting the Scene
In one of the earliest blogs, I wrote about the importance of silent operation, and in this blog and the next I am going to explain how I approached this. To summarize earlier blogs, there is no point in cruising all day on battery power if you make a noise recharging it overnight, so if you can recharge silently, why not recharge while cruising? This dictates the battery capacity and type (small capacity, Li-ion to give the required current and to avoid the reduced current charging of lead batteries). It also set the boat geometry with the generator in the bow, after all why not make use of the basic shape of a narrowboat?
With this groundwork done and the generator selected, the goal is to make the generator as quite as possible. Complete silence is not achieveable, but a level of sound below the noise of the wash or the breeze in the trees is what I am aiming for. There are two types of silencing to take care of, namely the exhaust noise, and the noise emanating from the rotating machinery.
Before going into details, I would like to issue a small word of caution to budding builders. The different generator manufacturers quote a noise figure for their generators but there is no consistency about how these levels are measured. For example, many are related to a distance of 7m, while Vetus report their value at 1m. As sound diminished as the square of the distance, this gives a 49:1 difference (-17dB). To cap it all, none explains how they treat the sound from the exhaust, so I would recommend that you treat the “Noise level dBA” with a pinch of salt.
Now let’s get into the detail.
As loud as it gets
Here is the spectrum for the unsilenced, uncovered generator. (See "footnotes for nerds" at the end of this blog for details of how the measurements were taken).
The frequency range from 10 Hz to 1000 Hz encompasses all the sounds of interest. On this chart the harmonics of the firing order are shown, with the first harmonic at 38Hz (a 3-cylinder diesel running at just over 1500 rpm and firing on alternate revolutions). As you would expect from any engine, the noise is strongly harmonic, with the first ten tones easily identified. Interestingly the second harmonic is far stronger, and almost saturates the microphone (0dB is full scale input).
The two minor peaks between each of the dominant harmonics is caused by differences between the cylinders, for example the different inlet and exhaust paths or perhaps injector tolerances.
Simply put, the more you can contain and absorb the sound the better. In this regard the BetaGen series of generators are not ideal because they have air cooled alternators and noise inevitably gets out through the air ducts in the alternator. Similarly, they are designed to exhaust hot air from around the generator, which offers another path for noise to escape (it also provide a source of warm air for drying clothes!) The balancing factor is that their soundproofing cabinet contains, from visual inspection, far thicker sound deadening material than the competitors’ generators. I suspect it’s a bit of swings and roundabouts.
So what effect does it have? Here is the difference in spectra between the generator with the covers off and on.
As you might expect, the attenuation is fair at higher frequencies, but at very low frequencies I think that the large metal panels tend to drum, which results in an increase in sound levels from 40Hz downwards.
On Perseverance we have placed the generator on a 20mm thick slab of steel, which will help to reduce sound transmission into the hull, and I have asked Ortomarine to line the engine bay with a second layer of soundproofing to keep airborne noise from reaching the bulkhead and so passing into the cabin. Only time will tell how effective these measures are.
Here is the aspect where we really can make a difference. I collected silencers of three different types and a collection of 1.5in BSP pipework and set about measuring their performance.
As well as a normal absorption type silencer, I chose to test a hospital silencer and a cowl silencer. The absorption silencer was 8” diameter x 22" long, the hospital silencer was 12" diameter x 21" long, and the cowl was 10” diameter x 5.3” length. The comparative size of the hospital silencer (silver) and the cowl (black) are shown here.
To start, I measured the attenuation of the three different silencers.
I was surprised by some of these characteristics.
Firstly, the cowl silencer was almost as good as the hospital silencer which is three times bigger. Normally silencer performance goes with size, so the performance of the cowl silencer was remarkably good. A cowl silencer one size larger might have won the day!
Secondly, both the cowl and hospital silencers give some amplification at low frequencies, and in fact I could feel the flat endplate on the hospital silencer drumming.
Thirdly, the absorption silencer has a quite different characteristic to the other silencers, justifying the recommendation from the supplier to fit two in series.
Finally, I had expected the large hospital silencer to do a better job of attenuation at low frequencies and for the absorption silencer to be better at high frequencies, but in fact the reverse was true. Just goes to show you can’t judge a book by its cover.
Interestingly, none of these results compared well to the published attenuation charts from the suppliers. Here, for example, is the curve supplied by the cowl manufacturer.
I like to plot attenuation downwards, not upwards, which is why the plot is upside down, but the shape and values are all over the place. It doesn’t remove the fact that the cowl silencer was very good. Why didn’t I fit it? At one stage it was in the design, but when we found there was space enough to fit the slightly better hospital silencer, I settled on that. If anyone out there wants a very slightly used, very good small silencer, it's free to a good home!
Performance of Hospital + Absorption Silencers
With the two complementary silencers in series, here is the overall attenuation using conventional silencers.
The detailed spectrum still shows the expected tones, with the slightly unexpected result that it is the second harmonic at 76Hz which dominates.
If you are looking to compare this with the first graph, be aware that the scale is changed by -20dB, or 1/100th the sound level. The key thing is that we still have the expected tones, but they are all well suppressed. The sound in this configuration is of a well silenced diesel engine with a gentle low hum.
But for Perseverance the goal is silent operation, so we need to do better. We need to be unconventional, and I will talk about that in the next post.
Notes for Nerds
The generator test setup has been described in an earlier vlog. The microphone was a SubZero SZC-500-USB device, selected for the good low frequency response of condenser microphones, quoted as 20Hz -18kHz for this unit.
The software was the excellent Arta program by Ivo Mateljan (firstname.lastname@example.org). I used the spectrum analysis mode, with a sample rate of 8kHz and 16384 samples in the result. The sample period was therefore about 2 seconds, frequency response from 0 to the Nyquist frequency of 4kHz. Averaging over a (possibly excessive) 100 samples meant that each measurement took 3 minutes 20 seconds.
No attempt was made to scale the results, which were all dB with respect to the microphone full scale input. The results presented are either (a) relative harmonic amplitudes, where the overall signal level is not important, or (b) attenuations where only difference between the before and after conditions was the addition of silencing panels or addition of a silencer. In all cases the generator under test was a BetaGen 7, running at 4.5kW load.
The available results cover the range to 4kHz, but for clarity only the range from 10Hz to 1kHz are shown. Analysis was carried out using a program written in Python using the Numpy and Matplotlib libraries.