Need Suggestions For Capturing Train Audio

[Scott Hiddelston]

Hi All,
I need some real advice to a very vexing problem. I like to record trains climbing hills. I am using a Panasonic HMC40 with a Panasonic AG-MYA30G XLR adapter and an AT 875R mic. The problem is capturing good audio while the train is at distance, yet not distorting when it gets close. I have tried all manner of settings, from using the limiter built into the camera along with my best guess, to setting different levels for each channel and blending the two later. But every location is different, with differing locomotives and distances from them.
The most frustrating part to this is that I have a small Sanyo xacti I carry in my pocket, and with it set on auto it does a wonderful job of capturing both distant and close up sound. Much better than the Panasonic and attachments. From what I can gather the Sanyo has Auto Gain and the Panasonic doesn't. I'd really like to make the Panasonic work as I like the camera itself, but I'd be open to suggestions of another camera if it would cure my problem. Also I'd appreciate any other advice on solving this problem. I'm open to changing any of the equipment but have no clue where to start. Can I buy a unit that would do Automatic Gain? I feel the sound I get is lacking in higher frequencies, but that just may be me.
I've added links to audio clips recorded by the two cameras. Thanks for any help.
Scott

http://dl.dropbox.com/u/19575495/sou...0panasonic.wav
http://dl.dropbox.com/u/19575495/sou...st%20sanyo.wav

[Greg Miller]

I haven't listened to trains close up since I was about 15, so I can't assess the "realism" of your two files. But let's talk about audio in general.

First off, what's the closest distance between you and the train? The 875R has a fairly narrow pickup pattern, which would be fine if the train is a mile away. But when the train gets close, since that mic is fairly directional, you won't get a good representation of the entire "sound stage." A cardioid, or better yet an omni, will sound more realistic if you're really close. (And of course a mono track from a single mic won't sound as realistic as a stereo recording, all other things being equal. I really think stereo would be important here, as the train moves across the frame, from one side to the other.)

I hear significant AGC in the Sanyo track. For example, when the whistle blows (around 17.3 to 18.8 seconds), the sounds of the engine become significantly softer, then when the whistle stops, the engine sound gradually comes back up in level. To my ear, that makes the Sanyo track terribly unrealistic. Another example: as the engine goes past, the level in the Panasonic track goes up, then down, reflecting the changing distance between engine and mic. In the Sanyo track, the level stays fairly constant (because of the AGC); it would not sound like that in real life.

The latter part of the Panasonic track, from about 1:36 to 2:00, does seem to have a pretty constant level. I wonder whether the camera's ALC is kicking in at that point, to prevent clipping. You might want to back off on the record gain by 6dB or so, to avoid that situation.

I don't particularly hear any overload or clipping in your tracks, nor do I see any in the waveforms. Do you think you hear any significant distortion? If so, what point in the time line? The 875R can supposedly handle 127dB SPL; I doubt that you're experiencing levels higher than that (or you would be experiencing severe physical pain).

I do notice a difference in frequency spectrum between the two recordings. Levels are similar up to around 300 Hz. Above that, the Sanyo has more level than the Panasonic, by roughly 6 to 12 dB. The difference is even more pronounced above 10 kHz. Since I don't know which model xacti you were using, I can't look up specs, so I can't really account for that. Again, since I don't have the sound of that train "in my head" I can't really say which track sounds more "realistic" to me. (The Sanyo track, with more HF, might be more "interesting" but maybe that's not actually realistic.)

I'd suggest you search for specs on your particular model xacti, and see if you can find out anything about the frequency response. If there is any kind of gradual LF rolloff available on the Panasonic, you might try that, in an effort to get the frequency spectrum closer to the xacti, if you prefer that sound.

Otherwise, hopefully someone who has experience with the HMC40 will step in with some suggestions.

[Jim Michael]

I wonder would it be possible to capture more of the low frequencies associated with the train, say with some type of boundary mic on a cross-tie, etc?

[Greg Miller]

I wonder whether the files that Scott sent are flat, or whether he applied some LF rolloff at some point.

I think a mic on a rail tie would have some problems. First, I'll bet the SPL is huge at that location, so you might need a special mic, some pads, etc., to avoid distortion. Also, there's the issue of perspective. If you got a *realistic* recording from a rail tie, you would hear the train approaching, coming within a few inches of the listener, then receding. You'd be unlikely to hear that in real life, unless *you* were located within a few inches of the passing train.

A boundary mic (or pair of them) located a reasonable distance from the tracks would be an interesting approach, and might be worth a listen. However, I don't think that would necessarily capture any more LF information than an omni (or pair) at the same location. Omnis tend to have pretty good LF response.

Maybe we need to think more about the issue of perspective. I recall reading an article that said you can never realistically capture and reproduce sounds (except for single-point sources). That's because you are recording only the sound waves that reach one (or two) points -- your microphone(s), and playing back all those sounds through only one (or two) speaker(s). The goal is to create an illusion that makes the listener think the recording sounds the same as the original sound.

So Scott, where are you, in relation to the train, when you make the recording? And do you want the recording to sound like what you hear at that distance, in that location? Or do you want it to sound like a train sounds from a different distance/perspective? Do you want to concentrate on the sound of the engine as it approaches, passes, and receeds? Or do you want a "realistic" sound of the entire train as it goes past... concentrating on the sound of the wheels on the rails and other hardware sounds after the engine has gone past? Do you pan the camera to follow the engine as it passes? If so, the engine will always be centered between channels, and one ear will end up facing toward the train, while the other ear ends up facing away from the train. If you pan the mic when you pan the camera, the perspective will change. On the other hand, a good recording of a passing train will have a fixed perspective, looking perpendicular to the closest point of the tracks.

Scott, please explain in detail what you want to hear, and what you dislike about your present track.

Darn, I might need to drive down to Altoona and see what I can record at the horseshoe curve... not today, though, when the weather is a little bit warmer. Before I do that, I'd like to hear from Scott about what distance and perspective he's trying to capture.

[Scott Hiddelston]

Thanks Greg and Jim.
After reading your replies, I've realized some things: One, I know very little about this subject, two, I guess I am trying to buy my way out of this problem rather than learn, and three, I should have offered a link to a clip including video to let you see what moves and distances were involved in the recording.
So I will include a video link at the bottom of this post.
The closest distance I assume is around 150 feet. I know you will instantly think this is way too close, and I agree. I have had much better results when I could get more distance. But most times I simply cannot get further away, due to trees, private property, etc.
The sanyo model is the sanyo vpc-fh1a. I looked for audio specs but could only find 48K 16 bit.
Generally I am after the sound of the locomotives working hard. I am looking to get that sound at approaching distance, passing in front, and receding. However sometimes locomotives are placed in the center of the train, and I would like to capture them going past in front while the lead locomotives are in the frame. You can see an example of that in the link attached. I am not really looking for the "whole train" experience, but with a priority on the locomotives, and the railcars included only to show the size of the train. In fact many times I edit out most of the railcars.
I feel the panasonic/AT875r combo did not capture the "hiss" (for want of a better word) of the locomotive exhaust. I feel that is present in the sanyo clip. Again I'm far from knowledgeable, but it sounds to me like the AT875r has missed the higher frequencies. In particular, the sound coming from the locomotive at the rear of the train seems to be blown out by something.
I appreciate your answers and I hope this info is what you were looking for.

http://dl.dropbox.com/u/19575495/sou...0panasonic.mov
http://dl.dropbox.com/u/19575495/sou...ip%20sanyo.mov

These clips ar 80 megs or so. If thats too big just let me know what is appropriate.

[Greg Miller]

Hi Scott,

Nothing wrong with spending some money to solve the problem... if that will solve the problem. But first it's best to understand the problem and try to figure out a reasonable solution, to avoid throwing away your money [u/]without[/u]solving the problem. ;)

I don't think there's anything wrong with 150' distance. I have heard some train recordings that were supposedly made between 25' and 50' and they sounded quite different from yours. But those were not engines laboring up a grade, those were trains sailing by, probably not working very hard. They were moving a lot faster, there was a lot more noise of wheels on the rails, rail scrape, etc., but not the sound of engines laboring. But since those recordings sounded OK at that close distance, I think 150' is probably quite acceptable.

Right, there are no specs for your small recorder and its internal mics. It's possible that recorder actually rolls off the low frequencies, to minimize room rumble and wind noise. It's also possible that recorder boosts the highs (above 10kHz) to increase articulation of voice recordings. There's no way we can find this out accurately, unless we rent an anechoic chamber and test the machine.

Still, the specs say the 875R is pretty flat, so I don't think its highs are rolled off, unless it's somehow defective. (Is it new? Has it been dropped or abused?) Did you have the mic muffled in any way? Perhaps something to cut down on wind noise? How was it aimed? And is there any kind of "high cut" or "low pass" filter in the camera, that might have been turned on?

(Hoping someone familiar with that camera might chime in here.)

You will see the 875R's acceptance angle is fairly narrow. Note that a mic is not like a lens, in that there is not a specific area that is or is not "in the frame." A mic's pickup falls off gradually as you go off axis. But some mics (like yours) fall off faster than a standard cardioid pattern; and an omnidirectional mic does not fall off at all. But note this: a mic's high frequencies fall off more abruptly (as you go off axis) than the low frequencies. Therefore, the mic will sound flat only when pointed directly at the desired sound. So with your mic, unless you pan it to follow the engine, it will likely sound muddy except when the engine is directly on the mic's axis.

I would therefore suggest a cardioid pattern, or even an omni, as an experiment. Don't you know any audio folks in your town who would be interested in participating in an experiment? Borrow a nice mid-range cardioid, or start with an omni, and see how that works. That will give us some more data and a good frame of reference.

Meanwhile, I'll wait for the weather to warm up and maybe take a weekend jaunt to Altoona.

[Scott Hiddelston]

Thanks for your info Greg. Unfortunately I am kinda rural, so no good stores around to buy from. And I know no-one else interested in this filed so no chance of borrowing any gear. I'll just have to take advantage of Amazon's generous return policy (and ultimately buy from them). With that in mind have you any suggestions? The Rode Stereo Video Mic comes to mind, but it would make my XLR adapter redundant.
Meanwhile I will experiment more. I DID have a windmuff on the mic, specifically a Wind-Tech mic-muff. I'll try without it next time. I have also discovered that if I switch the limiter in the camera on, then set the two recording levels in the XLR adapter differently, the limiter will cut off both channels even though only one is overloading, ruining the benefit of setting them differently to begin with. See here in the FCE audio graph:

http://dl.dropbox.com/u/19575495/Scr...04.22%20PM.png

It's just so difficult to gauge where I should set the recording levels beforehand, what with the different locomotive types and differences. Thanks again.
Scott

[Greg Miller]

Scott, a picture is forming as we continue the dialog.

Too bad you are rural as you describe. Actually, I am too, although I'm very close to Penn State, there are no reasonable audio dealers nearby, just a few music stores. Do you have any music stores near you that could lend or "sell on approval" a reasonable omnidirectional mic? How about a Radio Shack? I'd really like you to try a traditional cardioid mic, and/or an omni. It doesn't need to be the world's greatest mic, we just want to try a different directional pattern for comparison.

Here's some more theory for you. High frequencies fall off more rapidly with distance than low frequencies. They are even somewhat "blown away" because of high winds, or because of atmospheric refraction. Go and listen to a carillon on a windy day, and listen to how the sound of the bells changes constantly. We might theorize that your recording sounds a bit muddy because you're too far away. But then, on the other hand, the recording from your xacti sounded much brighter... that would tend to contradict that theory.

I apologize again that I'm not familiar with that camera. With some equipment you can "un-link" the AGC or limiter so the channels are independent... check your menus. Otherwise, I guess it's just a matter of trial, taking notes, and garnering experience. Eventually you will know where to set the levels.

Keep smiling. What you are doing is rather uncommon. If you had come in here and asked what mic to use for dialog on an indoor set (or some similar question) you would have gotten many answers from many people who have already done that. You are exploring rather new territory. It will eventually fall into place. Meanwhile, Google and see what you can find about train recording; someone, somewhere, has done it before.

If you are recording outside, you almost definitely need some sort of wind muff. If it's a legitimate known brand, it should have negligible effect on frequency response. And if you try again with your present mic, keep it pointed directly at the engine, because that mic is very directional.

As far as a good omni mic, perhaps someone will chime in with a suggestion or preference. I have read good things about (but have not personally tried) a Behringer ECM8000. It is a very small diaphragm omni condenser mic. It's designed as a measurement mic and is supposedly very flat. I've seen them online for as low as $50 so you could easily try one and if it's a winner you could get a pair and record in stereo. Not based on my personal experience, though...

Also, your brain does some amazing processing with the signals from your two ears. If you stand in a given location and listen with both ears, and then record with one mic, the resulting recording will not sound like what you heard in person. If you are going to record with one mic, then you need to listen with only one ear when you pick your recording location... plug your other ear with one finger (or wear a good isolation muff on one side). Location, location, location...

Too bad you aren't near central PA. I have Altoona's horseshoe curve here, I think there are trains climbing frequently, it should be a good place for some test recordings. As I said, when the weather gets warmer...

Keep us posted on your progress, and good luck!

[Steve J. Nordahl]

Hi Scott,

Being an avid Railfan I have come up with my own solution as to getting quality field audio at varying distances.

I've attached a link to the audio portion of Norfolk Southern's (NS) "Office Car Special"(OCS), and the actual Youtube video. The sound is approach from the right, pass, and departure to the left. I shoot with a JVC GY-HM700U with audio captured with a Rode NT4 stereo mic mounted on a shock mount inside a Rode Blimp on a K-Tek boom pole mounted on a Avenger C-Stand. The mic ends up around 15 feet high and perpendicular to the approaching trains. Mixer was a Sound Devices 302. Train is F9A/F7B/F7B/F9A + 11 Norfolk Southern Business/Office Cars. At the time, I set up about 150-175 feet from the track with the equipment.

Location is MP 22.1 on the NS Reading Line, in Longswamp Twsp., Pa. on September 26, 2010.

Audio: NS-951 OCS Train - Audio of NS-951 the Norfolk Southern (NS)"Office Car Special"(OCS) Sound is approach, pass, and departure sound. Captured with a Rode NT4 stereo mic mounted on a shock mount inside a Rode Blimp. Mixer was a Sound Devices 302. Train (Press on the player at the bottom to hear the train.)

Video: YouTube - Four Train Set..with NS "OCS" !!! - Sept 8th, 18th, & 26th 2010

Note: The video has good examples of trains in various distances using the same audio equipment. Skip to 7:08 for the OCS.

Here is a picture of the equipment setup along trackside. The other shows where I was in the large field giving a sense of the distance from the OCS train.

Steve

[Jim Andrada]

I was thinking that an Omni/Figure 8 M/S setup might work out well. Tucson used to be a railroad town - maybe I'll find some good place to try it out.

[Greg Miller]

Steve, that is excellent detail. Thanks for adding some specific information.

I'm surprised you set up only 15' from the track, that's much closer than I would have tried it (but I am not disagreeing, you're the man who's done it!).

That will give a much different sound from Scott's recordings at 150' distance. That's why I asked Scott about what perspective he was trying to achieve.

I am curious about only one part of your description:

Originally Posted by Steve J. Nordahl

The mic ends up around 15 feet high and perpendicular to the approaching trains. {snip} At the time, I set up about 150-175 feet from the track with the equipment.

Do you mean you set up the mic 15 feet from the track, then ran mic lines back to a location 150 feet away? If so, why? Why not just set up the recorder near the base of the mic stand? Maybe I'm missing something obvious...

By the way, where is Longswamp Twp. in PA? I'm in State College, near Altoona.

Thanks again!

[Jim Michael]

Where I was going with the boundary mic idea was, when you are close to a train moving by the low frequencies make you feel the presence of the train. So I was thinking an audio track that captured the low frequencies would help reproduce that presence.

[Greg Miller]

I hope this reply is posted OK. The board seems a bit confused. When I click "Reply" at the bottom of the thread, it opens a reply box that includes quotes from a post half way up the page. Hmmm.

Jim,

I see your point about the boundary mics. It comes back to the question of perspective. That would definitely capture the "up close" perspective, especially if played back over a system with a beefy sub.

I see now there is a wide range of expectations. Steve records from 15' away. His sound reminds me of my teenage years when we'd hike through the woods to the nearby Pennsy ROW and wait for the trains to come up (or down) the river. It was fun standing that close watching (and hearing) them fly past.

Another perspective: I grew up in a hilltop community about a mile from the above mentioned ROW. Going to sleep every night, we'd hear the trains come whistling down the track, and the rather mournful sound of the whistle echoing off the hills on the opposite side of the river. (Couldn't hear the trains themselves from that distance.)

At another point I was in a town which was bisected by a B&O line, standing at the crossing gate less than ten feet from the rails, as the freights crawled through town on their way to the steel mills. (The hydraulic gate was pumped down by hand, by a watchman in the crossing tower 20' or so above the street.) That's when I heard the repeated ka-thump, ka-thump as the wheels slowly rolled past and deflected the tracks up and down. My memory isn't vivid enough to recall whether I actually felt it through my feet, but I probably did at that distance.

So yes, different perspectives and different trains flying, crawling, laboring, will give distinctly different sounds. I've ridden passenger trains up around the horseshoe curve, and I've watched them from the highway a mile away, but I've never stood near that part of the line to hear the engines making the climb. I've already checked Bing to see how to get there from here, and I will definitely make the trip once the weather warms up. Not to mention the historic steam excursion line about an hour from here...

Thanks for starting an interesting thread, Scott!

[Jim Andrada]

Go for steam - so many more great sounds. When I was a kid steam was IT - they had the great looking sort of art deco GG1 electrics once you got to the electrified line into NYC but north of there it was steam all the way. Steam sounds so much more - well ALIVE, I guess. So many more emotions than diesel. Pants like a big dog.

Pennsylvania State Locomotives - Steam Locomotive - Electric Locomotive

The Japanese bullet trains don't even slow down as they whoosh through the stations where they don't stop - there's an announcement that they're coming and a request to step back from the edge of the platform - but boy do they ever make a great sucking sound as they whip through - never thought of recording them but now I have the bug in my head I might just do it next trip

Great thread indeed. Thanks

[Steve J. Nordahl]

Originally Posted by Greg Miller

Steve, that is excellent detail. Thanks for adding some specific information.

I'm surprised you set up only 15' from the track, that's much closer than I would have tried it (but I am not disagreeing, you're the man who's done it!).

That will give a much different sound from Scott's recordings at 150' distance. That's why I asked Scott about what perspective he was trying to achieve.

I am curious about only one part of your description:



Do you mean you set up the mic 15 feet from the track, then ran mic lines back to a location 150 feet away? If so, why? Why not just set up the recorder near the base of the mic stand? Maybe I'm missing something obvious...

By the way, where is Longswamp Twp. in PA? I'm in State College, near Altoona.

Thanks again!

Greg,

My distance from the track varies dependent on the location and available shooting space. The shot showing the equipment was in Duncannon, Pa. on the NS Pittsburgh Line along the Susquehanna River. I try to get 25 feet but some locations are tighter. 15 feet overall, is the closest I ever get to the side of the tracks.

In my comment, the 15 foot number is about how high the mic ends up above me. as you can see in the picture, it's slightly closer to the action then me. When this close I have to be watching the levels on the 302 right as the engines arrive. But when set correctly the first time, every train thereafter is usually fine. As long as I'm lighting the led's near the top threshold on the 302, and not hitting the limiters I'm fine. The entire audio package is always kept right next to me. I use 25 foot cables, and have longer, but I've never gotten in a situation where I needed them.

On the shot with the OCS train, the cam, mic, and audio equipment were set up at about 150-175 feet from the train. The only thing I did was keep the height of the mic lower and sloped out in front of me about 5-7 feet but high enough as to not be in my shot.

Longswamp Twsp, Pa. is on the Berks & Lehigh County line. not far from Kutztown, Pa.

Steve