Sound advice(?)

[GaryHinshaw]

[Apologies in advance for the text-heavy ramblings.]

I've always been a bit skeptical about N scale sound, but I've recently become smitten with the ESU decoders - mainly for their fine motor control.  Since they now offer drop-in-place sound decoders for Kato locos, I figured I take another look at sound while I'm thinking about converting decoders.

Requirements:
I model Tehachapi in a 10' x 20' finished garage and I host operating sessions with 4-6 road crews in a typical session.  The crew are all pretty busy with a throttle for most of the session, and much of the time is spent struggling up grade in run 8 or holding back a descending train in dynamics.  I would love for those particular sound effects to be compelling but not grating, i.e. an operator should feel the presence of their train and any train they meet, but not be too cognizant of the other trains in the room.

This places a premium on clarity over volume.  A good bass response is desirable, but your brain can fill the bass in if the harmonics are clear.  (Musical analogy: low C on a cello has a frequency of 65 Hz, but a sound wave of this frequency has a wavelength of 5.28 metres - much larger than the cello itself!  In fact most of the sound power in the cello note resides in the higher harmonics, but your brain perceives the fundamental frequency nonetheless.)

Example 1:
My starting point was two locos with factory sound: a Scale Trains Dash-9 and a Kato/Kobo SD70ACe, both with LokSound 5 decoders and 9 x 16 mm sugar cube speakers (not 100% sure about the Dash-9 speaker, actually).  These are delivered with the volume set rather high and most of the sounds are quite good - if a bit over-powering.  But the run 8 sounds get fairly muddled and the dynamics sound a bit hissy.  With the volume turned down, the lower throttle notches sound rather nice, but the higher notches are still muddled and frankly a bit annoying.

Example 2:
Inspired by @kiwi_bnsf 's explorations, I next turned to some Kato installs with the ESU 58741 board and a variety of speaker and soundbox options.  The speakers I tried were from CUI: 9x16 mm, 9x25 mm, and 14x20 mm.  The first two fit in a Kato GEVO without any frame modification.  For the 9x16, I used a short ESU soundbox, while for the 9x25, I printed a custom box, along the lines of the type printed by Tim, though not quite as fancy.  The sound file was a late style GEVO with a lot of 'GE thump'.





Both of these installs sound as good or better than the stock models in Example 1, but the run 8 and dynamic sounds are still disappointing.  I think there is just too much harmonic distortion at the low end of the spectrum to produce the desired clarity.  But the installation is dead simple, so that's a significant plus.

Example 3:
I tried the 14x20 mm speaker in a simple rectangular printed sound box with the whole assembly just taped to the top of the frame, with no loco shell present.  The sound was not noticeably better than the Example 2 samples.  I may proceed to try milling a frame and fashioning a more sophisticated soundbox/fuel tank, but the early tests don't give me much confidence that the results will be worth the extra work.

Example 4:
I came across a new speaker offering from TDK: piezo-electric wafers which vibrate when a voltage is applied to them:

https://product.tdk.com/en/products/sw_piezo/speaker/piezolisten/index.html

They are designed to be integrated into consumer electronics, for example, to drive sound directly from your tablet screen and such.  I was very curious to see if I could couple one of these directly to a loco shell and have the shell be the speaker itself.  The upside is that the these wafers are very thin and can easily fit in a number of tight spaces.  The down side is that they're expensive and they require a bit of extra resistance to play well with the ESU audio driver.  Here is a Franken-Loco(!) I rigged up for a test:



The wafer sizes in the photo are 30x15 mm and 20x10 mm.  The wafer is billed as a 2 Ohm speaker and the ESU board would not drive it very efficiently without some additional resistance (as far as I could tell), so I cobbled together a 2 Ohm resistor and wired it in series with the larger speaker.  I then coupled it to the inside of a shell with some double-sided tape (which is what TDK recommends).  The quick & dirty results were pretty neat!  But it will take a lot more experimentation (and a price decrease) to make this an attractive bulk option.  I do think they have a future though!

Summary:
For now, I'm going to convert a handful of Kato locos with the 58741 board and the 9x25 mm speaker, since it is so easy.  The goal is to convert enough locos so we can run a session with a sound equipped loco on every train, at least in the first half of a session.  Then I'll re-evaluate.  The remainder of my Kato fleet will remain silent for now, and will be converted to the LokPilot Micro for its fine motor control.

In parallel, I would love to get some larger speakers to try out on the bench with the 58741 board.  I'd like to determine whether the lack of run 8 clarity is due to the speakers, the decoder, or the sound file itself.  I'm guessing it's the speakers, and that significant improvements will be challenging.  That said, I do think the experience of switching (alone) with sound can be quite satisfying.

[peteski]

Yes Gary, the inportant thing to keep in mind that the lack of clarity or overall quality of sounds can be due to the sound recording itself (not just the speaker system).  I guess a good way to determine that would be to temporarily hook up a much larger higher quality speaker system (maybe from a small compact stereo, system) to the decoder's output and listen to it.

As for connecting lower impedance speakers to the decoder, using resistor to match the impedance will basically waste the excess audio power as heat, so the sound volume coming from the speaker will be lower.  If you have a 2 ohm transducer and a 2 ohm resistor, 50% of the audio power will be wasted.  But if you connect two 2 ohm transducers in series, that will result in total impedance of 4 ohms, in-spec for getting maximum power from the audio amp, and without burning up the amplifier.  You would likely have to experiment with phase of those transducers for maximum volume (that will likely be frequency dependent if they are not bot placed in a same air tight enclosure).

BTW, thanks for the info about 9x25 speaker. Did not see that one last time I bought a variety of speaker sized from Digikey.  Must be fairly new, and perfect for the Kato narrow-body locos.

[Cajonpassfan]

Hi Gary, nice to see you experimenting with another dimension of model railroading. When it comes to sound, I've learned there's a fine line between enjoyable and annoying, but once you find where that line is, I  bet you'll be hooked and find silent running no longer adequate to the task, especially as the technology continues to improve.
Your reference to music made me think of my own notion of what I find enjoyable, and that's driven by size of the layout, its setting, number of operators, type of operation, i.e. the "venue". You and I have similar "venues" in terms of size and a long mountainous run. With five guys in my trainroom, keeping the volume way down is essential, with only horns/whistles and maybe dynamics getting a few extra decibels. (And those super annoying bells get the least). I also find that with lesser overall volumes, I get less picky about the sound quality, not that it's not important.
I'll be following with interest.
Fun stuff, enjoy!
Otto

[jdcolombo]

Regarding sound clarity:
When designing a speaker enclosure for sound installations, pay attention to the interior volume of the enclosure (in cubic mm).   In general, more is better until you hit a "sweet spot" and then more is not better.  I've found through trial and error that larger speakers (e.g., a 13x18 or that 9x25 you found) need an enclosure volume close to 1000 cubic mm to sound good.  If you mount the speaker firing down into the enclosure, that means that a simple box enclosure needs to be about 5mm thick (not including the bottom, which will add a half mm or so, depending on the thickness of the material you use for enclosure) - 9x25x5 = 1125 cu.mm

The other key to sound clarity is to make sure the speaker is completely sealed to the enclosure; no air leaks permitted.  I use thick CA to accomplish this, but other solutions (silicone sealant; light-cured adhesives) also work.  I'm assuming you did this with your speaker/enclosure interface, but sometimes people try to use a sticky foam surround to attach the speaker to the enclosure, and that will not work.

You can find what works best for a particular speaker with trial and error: start with a minimum of 1000 cubic mm, and then go up and down by 100 cubic mm (or thereabouts) until you find what works best for you.  You might be surprised at how a "too small" enclosure really hampers both sound clarity and low-end frequency response.

John C.

[ednadolski]

I would love to get some larger speakers to try out on the bench with the 58741 board.

I've done that with some Bose speakers and found the sound quality of the Loksound recordings to be quite good. You may have to fiddle with adjusting the different sound slot volume levels individually to get the effect that you want.

To connect any stereo system input to a decoder speaker output, you will have to use an audio output transformer and lower the output volume on the decoder to a pretty low level, otherwise the output power amp on the decoder will overload the line-in level.  It is after all build to drive a speaker directly. That said, it may not have enough power by itself to drive much more than a small speaker.  IDK if there is a way to tap into the pre-amp audio signal on these decoders -- likely not, tho perhaps @peteski has looked closer into this.

Ed

[Jim Starbuck]

N scale sound is extremely subjective to the end user so what I do may not apply to your specific tastes in sound. Something that I find helpful in my own engines is to turn off or way down the extra things that are running along with the prime mover such as traction motors, air compressor, sanding valves etc. To my ear all these other sounds simply muddle up the overall effect. More is simply not better. One of the cool parts of the ESU decoders is that each part can be individually adjusted.
Some of my switchers have only the prime mover and a slight amount of air compressor running. I’ve never had anyone comment like hey, nice sounding traction motors.

Jim

[peteski]

I've done that with some Bose speakers and found the sound quality of the Loksound recordings to be quite good. You may have to fiddle with adjusting the different sound slot volume levels individually to get the effect that you want.

To connect any stereo system input to a decoder speaker output, you will have to use an audio output transformer and lower the output volume on the decoder to a pretty low level, otherwise the output power amp on the decoder will overload the line-in level.  It is after all build to drive a speaker directly. That said, it may not have enough power by itself to drive much more than a small speaker.  IDK if there is a way to tap into the pre-amp audio signal on these decoders -- likely not, tho perhaps @peteski has looked closer into this.

Ed

I was not advocating connecting the ESU's audio amp output to a home HiFi amplifier using line-in jacks. I was just thinking of connecting somewhat bigger and higher quality speaker in decent enclosure (like one from a small home stereo system) directly to the ESU amp's output to better evaluate the quality of the sounds in the sound project in the decoder.

I have also not really done any reverse-engineering of the audio circuitry on the decoder.  The audio amp is a Class-D amp for efficiency (analog amps would get too hot).  Basically it is a PWM device like what is used on decoders to provide variable voltage to the motor and function outputs.

There likely aren't any LC or RC filters on the output of the amp to smooth out the high frequency pulses (again due to space constraints), but the speaker itself being an inductor can handle those PWM pulses without any problems.

A resistor network (basically a voltage divider) could be used to bring the signal which is designed to drive the speaker to line-level voltages, but I'm not sure how the HiFi amp would handle the high frequency PWM signal.   A coupling transformer could be utilized to match voltage levels but as I mentioned, my goal was not to blast the ESU decoder's output through 100W amp and Bose speaker system -- just to listen to the output through a somewhat better speaker system than our minuscule sugar cube speaker setups, to see if the tiny speaker causes the "hissy" sound, or the audio in the recorded samples is not the best quality.

I also think that in N scale sound we need to set our sound quality expectation lower than with larger scale models due to the fact that we have to resort to using those tiny speakers.

[ednadolski]

... my goal was not to blast the ESU decoder's output through 100W amp and Bose speaker system -- just to listen to the output through a somewhat better speaker system than our minuscule sugar cube speaker setups

What about checking out the demo sound samples on the ESU download page?

https://projects.esu.eu/projectoverviews/search?cat=18&type=all&q=C44

Ed

[ednadolski]

I also think that in N scale sound we need to set our sound quality expectation lower than with larger scale models due to the fact that we have to resort to using those tiny speakers.

I actually find the default sounds in many HO locos to be more grating than N.  Doubtless due in no small part to the volume levels set by the mfrs, as a larger model is expected to "sound big".

Ed

[peteski]

I actually find the default sounds in many HO locos to be more grating than N.  Doubtless due in no small part to the volume levels set by the mfrs, as a larger model is expected to "sound big".

Ed

You are probably correct, and thankfully volume control is easily done  on any DCC system by at least changing the master volume  (a single CV which can be done using POM or service mode). Or further volume tweaks can be made to individual sound effects. I'm a fan of "scale sound" levels unless the model is operated during a public show with lots of ambient noise.

[GaryHinshaw]

Thanks for the feedback.  I definitely agree that low volume is important, especially in a session setting.  I have been using a master volume setting in the range of 30-50 (out of 192).  I have varied a bunch of the individual channels, but have not done so exhaustively.  I'm especially curious to try Jim's suggestion of singling out the prime mover to see if that improves clarity. 

In terms of box volume, I'm getting about 700 mm^3 for the 9x25 mm speaker, so shy of John's rule of thumb.  I can try larger volumes as a test, but they wont fit in the shell unless I mill the frame.  Maybe it's a better trade to fire a smaller speaker into a bigger box?  (I believe my speakers are well sealed to their boxes, BTW.)

I'm perfectly willing to believe that the sound files and audio amplifiers are not the weak link in the system.   I have a few larger speakers to try for fun, but I'm not planning to invest too much time exploring that side of the coin for now.  I'm mainly looking to maximize the quality per unit of effort that goes into it.

[kiwi_bnsf]

Hi Gary,

Great to see you evaluating speaker enclosures — there is definitely a lot of room for improvement over factory sound (especially in comparison to the stock Scale Trains 11x15mm speaker in the Dash 9 / ET44s).


Looking at your test install photos, one thing that really jumps out at me is the kapton tape that you have over the speakers. In my experience this is very bad — you want totally clear air around the moving component of the speaker (and the rear of a cell phone speaker has considerable vibration and corresponding air movement when they are firing inwards).

During my testing I found that fully sealed with inward firing speakers sounded best, and you ideally need at least 1-2mm of space above the speaker to minimise distortion. Without this I found there to be a lot of audible hiss. I 3D print keys into my enclosures so that they are firmly clamped by the shell halves. Alternatively you could use double sided tape on the underside of your speaker chambers.

Also, I would not put too much stock in what a speaker sounds like without the shell present. I found that you really have to test with the shell on from the outset, and consider and optimise the negative space within the shell above the speaker chamber. My final designs sound much better with the shell present than with it missing.


In terms of the ESU sound files, I've found there to be quite a lot of variation in quality at the higher notch settings (especially with the GE prime movers). Lowering the volume right down helps, but from about Run 5 upwards things get quite messy. I'm experimenting with customising the ESU notches to top out at run 5-6 so that you get more of the distinctive thwump of the prime mover at a distance.

In contrast, I find the ESU dynamic brake samples to be excellent, and with the right speaker enclosure these sound very clean and have a good low frequency extension.

Hope that this helps.

Cheers

[ednadolski]

When designing a speaker enclosure for sound installations, pay attention to the interior volume of the enclosure (in cubic mm).   In general, more is better until you hit a "sweet spot" and then more is not better.  I've found through trial and error that larger speakers (e.g., a 13x18 or that 9x25 you found) need an enclosure volume close to 1000 cubic mm to sound good.  If you mount the speaker firing down into the enclosure, that means that a simple box enclosure needs to be about 5mm thick (not including the bottom, which will add a half mm or so, depending on the thickness of the material you use for enclosure) - 9x25x5 = 1125 cu.mm

@jdcolombo is there an optimal thickness for the enclosure wall?

TIA,
Ed

[GaryHinshaw]

Thanks Tim, sound advice indeed.  I am intending to add mounting tabs to my box base like you did, but these are still v0.1 test boxes, hence the Kapton, which was never intended to be permanent.  I'll try removing it.  (Other than the 14x20 mm speaker, I have been testing with the shell on, and it definitely makes a difference.)

Has anyone tried making the shell be part of the box?  I was picturing a design where the entire box/speaker assembly is turned upside down and the ceiling of the shell formed the base of the box, and the speaker is firing upward into the box/shell.  It would require the base of the box frame be profiled to the shape of the ceiling to maintain a seal (which could be challenging), but the acoustic coupling to the shell would be much more direct and controlled.  This is basically how the piezo wafer speakers are intended to work.

In contrast, I find the ESU dynamic brake samples to be excellent, and with the right speaker enclosure these sound very clean and have a good low frequency extension.

RE the dynamics - I'd love to see how you have set your sound slot levels on a GE project.  Any chance you could post that info, maybe as a screen shot?  I'm sure my dynamics could be made to sound better.

[kiwi_bnsf]

I am intending to add mounting tabs to my box base like you did, but these are still v0.1 test boxes, hence the Kapton, which was never intended to be permanent.  I'll try removing it.  (Other than the 14x20 mm speaker, I have been testing with the shell on, and it definitely makes a difference.)

All good — I wasn't meaning to critique your prototype designs. The 9x25mm enclosure looks smart.

It's just anything touching or attenuating the speaker front or back is bad. Also I found that any flat surface that is too close to the front/back of the speaker leads to hiss and distortion. You should be able to easily reproduce this effect on the bench with one of your enclosures moving closer and further away from a flat surface like a piece of wood.

Has anyone tried making the shell be part of the box?  I was picturing a design where the entire box/speaker assembly is turned upside down and the ceiling of the shell formed the base of the box, and the speaker is firing upward into the box/shell.  It would require the base of the box frame be profiled to the shape of the ceiling to maintain a seal (which could be challenging), but the acoustic coupling to the shell would be much more direct and controlled.  This is basically how the piezo wafer speakers are intended to work.

That's an interesting idea. My concern would be that it would require some careful sealing, and it would also require some very precise depth measurements in order to leave sufficient space between the speaker enclosure and the frame. It's definitely possible though with 3D printing. I did a quick Google a while back to see if there were any cheap laser scanners that would allow accurate mapping of the inside of shells, but they were all very expensive if you wanted accuracy.

I've essentially iterated the reverse scenario by having speaker enclosures that almost exactly fill the shell voids with 2-3mm clearance creating a negative space above the speaker. This design was accidental and only cropped up for me as a result of a ridiculous number of combinations that I ended up empirically testing. However it reproducibly sounds better to suit my taste, and has proved adaptable to a number of shell types. I have not gone so far as to create an ideal sealed negative space for each speaker as there are more fundamental limits imposed by how much I can mill each chassis and the available space in the shells I've been designing for.

I have found that 9x16mm is almost always better than 8x12mm even with limited cubic volume. However I think there are diminishing returns as the 11x26mm with a big enclosure in my BLI AC6000CW is not fantastically better than the 9x16mm with a smaller enclosure.

It will be very interesting to test the 9x16mm vs the 9x25mm in the exact same available cubic volume.

RE the dynamics - I'd love to see how you have set your sound slot levels on a GE project.  Any chance you could post that info, maybe as a screen shot?  I'm sure my dynamics could be made to sound better.

I will fire up the Lokprogrammer and post some Dash 9 and SD40-2 values tomorrow. I have not done anything fancy with dynamics apart from tune the relative volume to the prime mover.

(I am working on a lot of customisations with ESU sound files right now, but those are for another thread).

Cheers