Best Of New Adventures in Loco Hi-Fi

[kiwi_bnsf]

I've recently had some time on my hands to finally work on converting my Kato locos to Loksound V5

I have around fifty Kato locos of various types to convert, so I decided to design and 3D print my own speaker chambers — which proved a good project to learn AutoDesk Fusion 360 CAD software, and get to grips with my 3D printer. I have also spent a lot of quality time milling frames to make room for the speakers and keep alive capacitors (the last time I did any metalwork was 30 years ago at school, so there was a bit of a learning curve).

I decided to do some A/B testing to compare various speaker types, speaker orientations, and enclosures using a pair of identical loco testbeds…


Bill of materials

Loco testbeds:          Kato SD40-2 (Early) Snoot Nose x2


Speakers:                 Soberton 8x12mm [SP1208]
                               https://www.digikey.com/en/products/detail/soberton-inc/SP-1208/6099104
         
                               CUI 9x16mm [CMS-16098-30-SP] 
                               https://www.digikey.com/en/products/detail/cui-devices/CMS-16098-30-SP/6243982
                               (note that there are some slightly newer versions of this speaker now available)

                               Ole Wolff 9x16mm [091630LA-8B]
                               https://www.digikey.com/en/products/detail/ole-wolff-electronics-inc/OWS-091630LA-8B/13683680


Decoder:                  ESU Loksound V5 Kato [58741]
                               https://www.esu.eu/en/products/loksound/loksound-5-micro-dcc-kato-usa/


Keep Alive:               Iowa Scaled Engineering Run-N-Smooth PowerKeeper (940 microFarad)
                               https://www.iascaled.com/store/ModelRailroad/Run-N-Smooth/CKT-DCCSA


Milling Machine:        Proxxon MF70
                               https://www.proxxon.com/en/micromot/27110.php


3D Printer:               Anycubic Photon Mono X 6K SLA printer with standard Anycubic grey resin
                               https://www.anycubic.com/products/photon-mono-x-6k



I started off by designing and printing speaker enclosures with the maximum possible internal cubic volume that I could shoehorn into the Kato SD40-2 (Early) frame.

The upper rear "shelf" of the frame has the biggest available volume, and I milled a separate space for the keep alive in the fuel tank…

[ Guests cannot view attachments ]


My initial designs achieved an internal volume of approximately 1250 cubic mm, and so I was feeling pretty confident. I left room for the stock Kato rear glass for the headlights and number boards, and created a mount for an SMD LED (as many others have done). All speakers were sealed with CA, and the outward firing versions had their wire holes up sealed up as well.

[ Guests cannot view attachments ]   [ Guests cannot view attachments ]

[ Guests cannot view attachments ]


I did some fairly exhaustive A/B tests and have come up with a few interesting results. I ended up going down a bit of a rabbit hole of design iterations and further testing…

[ Guests cannot view attachments ]


I've made a series of comparison recordings of some of the better sounding combinations of speakers and chambers using a Sennheiser MKE600 shotgun microphone. It might seem strange, but I ended up loading a Dash 9 sound file onto the decoders for testing, as this has a deeper engine note that makes it easier for me to hear how much bass extension a given setup achieves (the SD40-2s will get the correct EMD 645E sound file once I'm done with this test phase).



Observations:

1.  All three speakers sound better firing inwards (i.e. with their wires on the outside) rather than firing outwards (i.e. wires on the inside). The outward firing versions all suffer audible muffling of some frequency ranges which is particularly noticeable when playing the horn sound. This effect was reproducible for both the 8x12mm and the 16x9mm speakers across a range of enclosure types. Doing a bit of reading, these cell phone speakers seem to have been designed to fire inwards (and there are versions made with peel-off sticky surrounds that are only available in this orientation). This is a convenient fact, as it makes wiring up the speakers and ensuring a good cabinet seal a lot quicker

These recordings compare identical setups with the only change being the speaker orientation:

   Soberton 8x12mm inward vs outward firing:
   

   Ole Wolff 9x16mm inward vs outward firing:
   

2.  All three speakers (in both inward or outward firing orientations) produce an audible hiss/distortion when playing the prime mover sound above run 3 (and especially so when sounding the horn as well).  The hiss/distortion is still audible running the speakers at very low volumes — less than 30%

I tried designs that fire downwards into the chassis, and upwards into the shell. In all cases care was taken to make sure no part of the speaker diaphragms touched anything, and that the speaker diaphragms didn't get any CA on them.

I found this to be highly dependent on the amount of free air around the exposed side of the speaker. On the workbench, I found that the hiss/distortion drops off as you give the speaker enclosure a bit of clearance relative to a flat surface, and then comes back once you move towards completely free air. There seems to be some form of interference from the sound waves bouncing off a close surface and interacting with the speaker.  It appears that there is a balance between the internal volume of the enclosure (the "speaker cabinet") and the space around it (the "room"), and the reinforcement that this offers.

In the end I tried a series of speaker enclosure designs that included differing amounts of free space between the speaker and the shell in an attempt to minimise this unwanted effect. My final iterations of this process leave 3mm of free space above the enclosure (leaving an approximately 500 cubic mm space above the speaker inside the shell). This is counter intuitive as it leads to a much smaller internal volume for the speaker chamber itself, but sounds way better to me — especially in Run 8 (and with the horn) where this effect otherwise leads to the prime mover sound becoming a "roar" of the higher frequencies all blending together.

This recording compares the same Ole Wolff 9x16 speaker in my original "max cubic" design versus my later "balanced cubic" enclosure (both cases are inward facing with identical decoder settings)…

   

This is probably one of my most subjective design choices. To me, the "max cubic" clearly sounds better when soloing the horn, compressor, or bell, but once you add the prime mover, the balanced design avoids a lot of the high frequency hiss/distortion (albeit at the loss of some detail in the horn and bell). In this case the recording doesn't really do this one justice — in person I find the difference is very clear as soon as the prime mover is on. The balanced version effectively sacrifices treble and detail to achieve more bass.

This recording compares the Ole Wolff 9x16 speaker in my later "balanced" enclosure with and without the loco shell present…

   

The shell is doing a good job of reinforcing the speaker — I'm definitely not going to be drilling any holes in my loco shells or fuel tanks.



3.  The 9x16mm speakers with a smaller enclosure volume consistently outperform the 8x12mm with a larger enclosure volume when it comes to bass extension. The Soberton 8x12mm speakers have a noticeably "brighter" sounding treble that can be very harsh depending on the sound file used. The 9x16mm speakers reach quite a bit lower, and to my ear create a fuller representation of the prime mover rumble…

   Soberton 8x12mm vs Ole Wolff 9x16mm:
      

   Soberton 8x12mm vs CUI 9x16mm:
         

Again, the horn and bell sound better soloed on the Soberton, but as soon as you add the prime mover, the 9x16 generates more bass and GE chug™
At Run 8, the Soberton has a lot more high frequency going on — which is a big part of what I find fatiguing about these tiny speakers. If you are mainly switching, perhaps the Soberton would be the winner.




4.   The new Ole Wolff 9x16mm speakers are a lot more sensitive than the 9x16mm CUIs and therefore are a lot louder at the same decoder setting. For all the recordings, I ended up matching all three speakers to the same sound pressure level using a meter. I had to reduce the master volume by 35% for the Ole Wolff for all my tests. Once matched, the differences between the CUI 9x16mm and the Ole Wolff 9x16mm are quite subtle. I prefer the Ole Wolff as it has lightly more detail, and is better able to cope with multiple sounds playing at once without distortion…

   CUI 9x16mm vs Ole Wolff 9x16mm:
      

5.   I found that thinner walled speaker enclosures were more prone to ringing and buzzing. I have settled on 0.6mm as the minimum wall thickness, and added 0.5mm fillets to all the internal edges to increase rigidity.



Just as with Hi-Fi systems, locomotive speaker performance is obviously a very subjective field, and what sounds good to me might not sound good to someone else. In some ways, this tuning odyssey became mostly about tuning the speaker enclosure design to minimise the effects that I personally dislike. I find high frequency and distortion to be very tiring to listen to, even though I run my Loksound decoders with the prime mover sound set fairly quiet.

The final prototype designs incorporate a whole bunch of changes to geometry and refinements based on all this testing. I've also managed to simplify things further by locating the 9x16mm speakers where their spring contacts can be bent to be soldered directly to the ESU 58741 Kato board, which creates a combined drop-in decoder and speaker which makes loco maintenance a lot easier…

[ Guests cannot view attachments ]

[ Guests cannot view attachments ]


(The BLW logotype started as somewhat of a joke with some of my N scaler mates who nicknamed my obsessive working on loco internals as "Benson Locomotive Works").


In summary, I'm going to use the Ole Wolf 9x16mm for all the locos that can fit it — as it runs with the lowest current and sounds similar to the CUI.

I'm going to try a few ported designs this week to see if bass can be extended any further — but I think there is not much more to be found with such tiny enclosures.


I'd like to say a big thanks to a number of Railwire members who have strongly influenced me to embark on this project — the information from @jdcolombo @RBrodzinsky @Steveruger45 @nightmare0331 and @peteski has been incredibly helpful in getting to this point. I hope this post adds something to the growing body of N scale DCC sound info that The Railwire hosts 

Cheers

[peteski]

That is one impressive piece of research Tim! Acoustics is a pretty complex field. As you mentioned, it is also highly subjective.

But yes, even the positioning of the speaker (with its enclosure) inside the model (and whether  the speaker is directly exposed to the outside or fully hidden inside the model will affect its sound).  It is not just the enclosure.  All the enclosure does is to isolate the front and back of the speaker cone to prevent nullification of sound waves.  The loco shells are fairly thin and even they can be resonating at certain frequencies, amplifying certain frequencies whilst suppressing others.  High frequencies are can also be deflected.

If you have a HiFi set at home, play some music and move your hand closer and father from the tweeter and you will hear the high pitch sounds sounds change pretty dramatically. Or if you have a car with high mounted tweeters (on the top of the dash, or in the A-pillars) you could do this experiment in your car.  So as you describe, there are many variables that can affect sound.  I'm also not really sure I can explain why those speakers seem to sound better while the flat speaker cone (for the lack of a better word) faces into the enclosure.

The enclosure thickens does make a difference.  Your HiFi speakers at home are build from thick wood composite board which is rather dense and stiff.  Home speaker systems are also usually filled with sound deadening material (often some sort of fiber fill.  I wonder if placing similar  material in your enclosures would make any difference in sound quality?  Of course you would have to make sure it does not contact the speaker cone.

[ednadolski]

Nominating this thread for a "Best of" tag.

Ed

[peteski]

Nominating this thread for a "Best of" tag.

Ed

I second the motion.

[kiwi_bnsf]

If you have a HiFi set at home, play some music and move your hand closer and father from the tweeter and you will hear the high pitch sounds sounds change pretty dramatically. Or if you have a car with high mounted tweeters (on the top of the dash, or in the A-pillars) you could do this experiment in your car.  So as you describe, there are many variables that can affect sound. 

Absolutely. I have spent a lot of time with my HiFi speakers at home optimising how far they are away from the back wall in order to get them to soundstage well, and not have too much muddy bass.

I'm also not really sure I can explain why those speakers seem to sound better while the flat speaker cone (for the lack of a better word) faces into the enclosure.

I think cell phone speakers firing inwards is a design choice based on the requirements of the phones themselves… any grille or open area is a place for water ingress. So the mobile manufacturers have gone with drivers that fire inwards into a chamber with a long folded port or horn that exits the phone through a small aperture or vent. Quite a few IoT devices also fall into this design envelope, and I think that has what has spawned all the different sizes of speaker that are now available.

Apple iPhone X earpiece speaker:

[ Guests cannot view attachments ]


Apple iPhone X loudspeaker:

[ Guests cannot view attachments ]


Samsung S9 loudspeaker:

[ Guests cannot view attachments ]


It is also possible that I have just happened upon a sweet spot with these enclosure size ranges and speakers where inward works better.

The enclosure thickens does make a difference.  Your HiFi speakers at home are build from thick wood composite board which is rather dense and stiff.  Home speaker systems are also usually filled with sound deadening material (often some sort of fiber fill.  I wonder if placing similar  material in your enclosures would make any difference in sound quality?  Of course you would have to make sure it does not contact the speaker cone.

This! Adding some damping foam in the void between the rear of the speaker and the shell is on my list to test as soon as my next batch of 3D prints are run (hopefully in the next week or so).


I'm also going to try some folded port and transmission line designs to see if these design principles even apply at this size…

Dual folded ports:

[ Guests cannot view attachments ]

[ Guests cannot view attachments ]


Long folded port:

[ Guests cannot view attachments ]

[ Guests cannot view attachments ]


Transmission line (reverse tapered):

[ Guests cannot view attachments ]

[ Guests cannot view attachments ]



(All of these designs are right on the edge of what is printable).

Cheers

[Chris333]

What frequency is it tuned to? 

[Steveruger45]

Impressive research, for sure Tim.  Brilliant stuff. 

[peteski]

What frequency is it tuned to? 

Exactly. That is where acoustics comes into play.  Tuned ports (their length) are designed to delay the air waves  coming from the back of the speaker to actually augment the waves coming from the front.  Just randomly adding ports of various length will make some difference, but might not be a desired difference.

As far as inwards firing speakers in smart phones go, they all are. But if you think about it, those are all tuned and open enclosures so the air that comes out of the enclosure is still "pumped" but the fully exposed "front" cone inside of the enclosure.  Again, these are likely designed using acoustic calculations, not just randomly put together with the main parameter being the available free space in the locomotive.  I'm not trying to put Tim's work down, but he is not an acoustical engineer who uses some specific software for designing speaker systems.  If a speaker in Tim's enclosure sounds dramatically better then Tim got lucky using his semi-educated design). That would also be how I would describe my acoustics knowledge. Yes, any speaker in an enclosure (which isolates the front and back of the speaker cone) will make it sound better than just a bare speaker.

Still I applaud Tim's experimental efforts, and I can say that he is taking this much, much further than some "real" model train manufacturers who don't seem to put much thought into speaker enclosure design, or even not use enclosures at all.

I'm also amused seeing BLW on the enclosures. That instantly makes me think of the beloved Brooklyn Locomotive Works, where I shopped for decades,and super-nice Peter Postel, the retired proprietor.

[Dwight in Toronto]

Damn, this is interesting stuff!  I’m impressed that a 3D printer might be able to form those sinusoidal ports - that is really something indeed.  Seeing the extensive variety of modeling that is being done with these things, I suspect that an acquisition is destined for my near-term future. 

One of my earliest sound installs was also an ESU 58741 into a Kato SD40-2 Snoot, using a salvaged iPhone speaker and a simple styrene enclosure. The result remains hugely satisfying.  But I really admire the professional ‘factory’-like fit & finish that you achieve with those printed enclosures. 

Your cell phone speaker examples take me back to a discussion we had here some months ago, wherein I relayed how an old iPhone 4 speaker yielded the best loco sound I had experienced to date.  That particular speaker had an unusually complex geometry (no doubt a key contributor to its superior sound quality), but the dimensions would not fit within an N scale loco.  Out of curiosity, I took it apart to examine the internal baffles, chambers and porting … there was even a very tiny cube of soft foam rubber in there! 

Looking forward to following your ongoing experiments. 

[kiwi_bnsf]

This has definitely been an empirically driven process based on actual observations

I started with the Kato SD40-2 (Early) because it has extremely limited milling options, and I figured that whatever speaker performance I could squeeze out of the available space would likely translate well to my fleet of SD40/45/45-2 and SD40-2 mids, and bigger SD70Ms.

I foolishly thought that with such a limited design envelope, the best speaker design would only take a few rounds to confirm… but that was not what the testing led to.


Full disclosure — I've never studied acoustic design, nor audio engineering. However, I did study Chemical & Process Engineering, so I can at least read a paper or two, and plug things into a formula.

With port design, the frequency reinforced is primarily proportional to the vent length, vent diameter, and the cubic capacity of the speaker box. A longer narrower port reinforces the lowest frequencies.

My single ported CAD design theoretically reinforces 276Hz:

[ Guests cannot view attachments ]


My dual ported CAD design theoretically reinforces 475Hz:

[ Guests cannot view attachments ]



You can play with the numbers here:       http://www.mh-audio.nl/Calculators/WVC.html


These figures will not really hold for my labyrinthine port folding, but they do at least hint that low frequency reinforcement is possible with longer vents.

The frequency response curve for all these small cell phone speakers is very similar… they reach full loudness at 1kHz, and drop off quickly as you move down the frequency range…

[ Guests cannot view attachments ]

Any reinforcement in the 100-500Hz range would therefore theoretically make a big improvement to prime mover sound. Or it could sound a mess because the Total Harmonic Distortion at lower frequencies is so much worse with these speakers…

[ Guests cannot view attachments ]

(I think my problem of increased distortion with the GE prime mover sound playing is partially explained by this THD curve).


I'm really up against it with how much vent pipe I can fit into the overall space, and how small a diameter will print. There is also an issue that very small diameter vents can whistle due to the volume of air flowing through them. It is surprising how much air these speakers shift — I can really feel it when testing.

It may well be that a sealed box is the best I'm going to manage here, but it definitely seems worth trying some ported designs seeing as that is how the cell phone and IoT device manufacturers are extracting so much performance from these tiny speakers.


One the most fun aspects of this project for me has been the ability to think of something, then 3D print it, and then test it within the space of a week. 3D printers and other additive manufacturing really are game changing for this type of design.

I'll be sure to share my further test results (both good and bad).

[Cajonpassfan]

Loving it! Really impressive work, Tim!
Otto K.

[peteski]

Good stuff Tim.

[jdcolombo]

All I can say is . . . Wow.  Very impressive (the actual printed enclosures are very impressive, too!).  I'm mostly converting steam these days, but this does give me some ideas about the size of enclosures inside tender shells.  And the inward firing vs. outward firing comparison has me convinced: inward it is.

The increased performance of a 9x16 over an 8x12 is very noticeable even in simple box enclosures.  But that's to be expected.  The size of the vibrating diaphragm in an 8x12 is 96 sq mm vs. 144 for a 9x16 and 165 for an 11x15 (not useful for an N scale diesel, but very useful for a N-scale tender).  13x18mm gets you up to 234 sq mm of diaphragm area, and in my experience, it sounds an order of magnitude better than the 9x16 or 11x15 and can usually fit in larger N-scale tenders.

I have my doubts that you'll get much better performance out of a ported enclosure given the overall frequency response of these speakers, but in the interest of science, i'll be watching

John C.

[kiwi_bnsf]

Righto, so I managed to get five variations of vented speakers 3D printed on Sunday.

The good news is that they all printed successfully, and the vents were correctly rendered even down to 0.6mm…

[ Guests cannot view attachments ]

[ Guests cannot view attachments ]


The bad news is that the vented versions all sound worse than the "balanced" version with the completely sealed enclosure.

I've not made any recordings, because it's unnecessary — the sealed enclosure is such a clear winner. The vented versions have no improvement in bass extension, but lose volume, lack detail, and generally sound muddier. So that's a bust!

I think this brings my testing phase to an end, and I will move onto making "production" versions.


@peteski — I have tried a foam insert in the free air above the "balanced" version, and it does seem to calm down some of the brighter harmonics further. I need to do some final random A/B/X tests once I have several locos converted so that I can be categoric about this, as the difference is subtle and I think my mind is now biased towards preferring the one with the foam

Thanks for all the positive comments — I'm glad this is of interest!

[kiwi_bnsf]

The increased performance of a 9x16 over an 8x12 is very noticeable even in simple box enclosures.  But that's to be expected.  The size of the vibrating diaphragm in an 8x12 is 96 sq mm vs. 144 for a 9x16 and 165 for an 11x15 (not useful for an N scale diesel, but very useful for a N-scale tender).  13x18mm gets you up to 234 sq mm of diaphragm area, and in my experience, it sounds an order of magnitude better than the 9x16 or 11x15 and can usually fit in larger N-scale tenders.

Thanks for the additional info — I'm already thinking about whether I can fit an Ole Wolff 11x15mm, 11x23mm, or 11x33mm into my BLI AC6044CW   

I will also try the Ole Wolff 8x15mm in my narrower hood units.

(It's annoying that they don't make a 9x23mm, as that would fit in quite a few of my larger locos)