Constantin Posted October 3, 2018 Report Share Posted October 3, 2018 Following on from this thread: I would now like to post an abbreviated version of my transmitter noise comparison. As mentioned in the linked thread, on the request by my regular post team I have set up a test to compare the noise created by various transmitters. Originally, this as I didn't know what I was looking for, this test also included several microphones, but here we are only really discussing the noise of digital transmitters, so I'll focus on just one microphone, the DPA 4017B. First, the boring stuff: The setup: I recorded myself reading an excerpt from a random book I had around. This is in German I'm afraid... To record this I used an AKG 4000B, because it has very low self noise and because this mic was certain to not be part of the mic comparison. I recorded this via cable into a 788T. I played the file back from this 788T and recorded the comparisons into a second 788T. I played the file back through an Alesis (I think) amp into HHB Circle 5 speakers. Not exactly a high-end setup, but out of what I had available it seemed the quietest setup. Since these parameters are the same for all transmitters it seems fair. The speaker sat on my desk and the mics were mounted on a stand at 95cm from the treble cone. Once the mic was mounted I didn't touch it until I had recorded all variables. Before recording I played a Line-up tone through the speaker at -20 to attempt to get all tx to the same gain setting. It's not very easy to set all these tx at exactly -20... After recording, I loaded all files into ProTools, cut off the beginning and end and then normalized them. So the peak level should be the same on all files. That's it, I think. The following transmitters are featured here: Audio Ltd. A10, Lectrosonics HM, Zaxcom ZMT3-phantom. The fourth recording was made straight through a cable. I originally included the Zaxcom 742.6 as well, but decided against it here, as I think it's more balanced to have just one transmitter per manufacturer. Please note, the order in which I listed the tx above is not the order in which they are listed below. I'd like to keep it anonymous for the time being, so everyone can have an unbiased listen. Please note: This comparison is only meant to highlight noise differences. It is not really suited to reveal differences in sonic quality. Or rather, it does reveal differences, but they are not conclusive. Also, obviously, things like range, etc. do not factor in here. EDIT: for an unbiased listen, do not look at the spectral diagram or read the comments before you‘ve had a listen. Otherwise you‘d know which one is Zaxcom, by just looking at it, but that’s not the idea. So... now it's your turn. I have to say, I am very very curious to read what everyone thinks... 4017-3.wav 4017-4.wav 4017-1.wav 4017-2.wav Quote Link to comment Share on other sites More sharing options...
OnTheSoundSideOfLife Posted October 3, 2018 Report Share Posted October 3, 2018 Based on what you said in the previous thread, two files should be noisier than the other two, because they are full digital. Listening to the ambience only and watching the spectrum in Pro-Q, to me file "4017-1" is Lectro (least noise out of the three systems), file "4017-2" is cable (no noise), file "4017-3" has the most noise and a cut off at 16k, so Zaxcom is my bet. File "4017-4" has slightly less noise than Zaxcom and it has slightly more emphasis on lows und highs, while Zaxcom generaly is more mid heavy, so Audio Ltd is my bet. While I can hear the differences in noise, to me the sonic characteristics stick out more than the noise that would be buried in ambience tracks anyways. I guess it becomes more of a problem once you start to compress the dialogue tracks in post (which I never do btw, I always try multiband compress/tame the parts that are too dynamic and gain automate the parts that are too quiet, but I never throw a single band compressor on the whole dialogue track). But again I don't mix TV shows, and I guess there are certain standards, lack of time etc. that force people to do it. I think the problem gets more obvious on crappy playback systems that have an emphasis on that problematic frequency area. Quote Link to comment Share on other sites More sharing options...
glenn Posted October 3, 2018 Report Share Posted October 3, 2018 You should include the TRX742.6. It has neverclip and our best A-D in a transmitter. Glenn Quote Link to comment Share on other sites More sharing options...
Kvark Posted October 3, 2018 Report Share Posted October 3, 2018 The noise is a bit more suppressed on the Audio Ltd compared to the Zaxcom. Either way... This can be solved in post. Of course not ideal, but not a huge problem. Really wished you would have included Wisycom, but appreciate your effort with those four clips. :) Quote Link to comment Share on other sites More sharing options...
Tuhiva Posted October 4, 2018 Report Share Posted October 4, 2018 Thank you for putting together a listening test comparison while controlling as much variables as you could given the fact that it is hard to have access to a facility and equipment that meet the requirement of EBU or ITU requirements. I've listened to the samples through protools playing back on old Mackie HR 624...I could not hear any difference in noise. At least nothing that is not going to be burried under ambient noise on normal shooting environment. You "can see" (not hear) nothing above 16KHZ on Sample 03 so for sure it is zaxcom and Sample 02 goes above 20khz so it might be cable. I noticed the difference in tone between the samples: the second one being the most "balanced" and sample 01 being my least favorite. But again nothing that you cant tweak by using some EQ. And again some systems might respond in a different way depending on the language being recorded and the characteristics of each personn own voice. For me at this point and price tags all of the system tested are equivalent sound wise. Special features, customer service or durability are probably what is going to differentiate those high end systems. Quote Link to comment Share on other sites More sharing options...
mikewest Posted October 4, 2018 Report Share Posted October 4, 2018 Great comment! Bien recu! mike Quote Link to comment Share on other sites More sharing options...
Tony Johnson Posted October 4, 2018 Report Share Posted October 4, 2018 10 hours ago, Constantin said: First, the boring stuff: The setup: I recorded myself reading an excerpt from a random book I had around. This is in German I'm afraid... To record this I used an AKG 4000B, because it has very low self noise and because this mic was certain to not be part of the mic comparison. I recorded this via cable into a 788T. I played the file back from this 788T and recorded the comparisons into a second 788T. I played the file back through an Alesis (I think) amp into HHB Circle 5 speakers. Not exactly a high-end setup, but out of what I had available it seemed the quietest setup. Since these parameters are the same for all transmitters it seems fair. The speaker sat on my desk and the mics were mounted on a stand at 95cm from the treble cone. Once the mic was mounted I didn't touch it until I had recorded all variables. Before recording I played a Line-up tone through the speaker at -20 to attempt to get all tx to the same gain setting. It's not very easy to set all these tx at exactly -20... After recording, I loaded all files into ProTools, cut off the beginning and end and then normalized them. So the peak level should be the same on all files. That's it, I think. All of the files sounded good to me and i guess we should expect that as it was a very evenly modulated reading with time and the opportunity to set up each system to its optimum gain etc. In a real world scenario where an actor may give several different reads with a 20db or more difference in performance on the fly then it comes down to what features you have at your disposal to deal with that. That would be the deal breaker for me. But good on you for taking the time to record these and post them. Tony Quote Link to comment Share on other sites More sharing options...
Constantin Posted October 4, 2018 Author Report Share Posted October 4, 2018 9 hours ago, OnTheSoundSideOfLife said: Listening to the ambience only If you are listening to the ambience you are listening to the wrong this. What I was talking about only manifests itself in the dialogue. 8 hours ago, Kvark said: The noise is a bit more suppressed on the Audio Ltd compared to the Zaxcom. Either way... This can be solved in post. Of course not ideal, but not a huge problem. Really wished you would have included Wisycom, but appreciate your effort with those four clips. I did too, but I don’t have access to a Wisycom tx. 8 hours ago, Tuhiva said: Thank you for putting together a listening test comparison while controlling as much variables as you could given the fact that it is hard to have access to a facility and equipment that meet the requirement of EBU or ITU requirements. Yes of course. It’s not a scientific test by any means. I don’t even have a vocal booth. Well... I do, but it’s not set up yet. 8 hours ago, Tuhiva said: I could not hear any difference in noise. At least nothing that is not going to be burried under ambient noise on normal shooting environment. Well, that is not the case. Two people sitting at a table indoors talking, there’s no ambience to speak of. It’s not hiss I am talking about, think Mp3 compression... I‘m going to have to refer to my regular post team here, and believe them when they say that they cannot easily bury it or get rid of it. This is even more so when they need to EQ a passage anyway, maybe because the boom couldn’t get right on top or whatever and they just want to add some air, that becomes far more difficult as that will accentuate the transmitter noise 8 hours ago, Tuhiva said: You "can see" (not hear) nothing above 16KHZ on Sample 03 so for sure it is zaxcom and Sample 02 goes above 20khz so it might be cable. It‘s a bit of a shame you did it that way, as you are meant to find out by listening, not looking Quote Link to comment Share on other sites More sharing options...
Constantin Posted October 4, 2018 Author Report Share Posted October 4, 2018 1 hour ago, Tony Johnson said: All of the files sounded good to me and i guess we should expect that as it was a very evenly modulated reading with time and the opportunity to set up each system to its optimum gain etc. In a real world scenario where an actor may give several different reads with a 20db or more difference in performance on the fly then it comes down to what features you have at your disposal to deal with that. That would be the deal breaker for me. But good on you for taking the time to record these and post them. Tony You are absolutely right. Whatever can be heard in differences here, would be far more pronounced in real life, so to speak. The thing is, when it comes to data compression rates, we have no tools at our disposal, the manufacturer gets to decide that. Of course, the differences are not „in your face“ obvious, that’d be crazy. No they are subtle here, but on location it becomes far more obvious (when listening back later). As I said in the other thread, I just watched online a recent tv movie I did. Now, once all that gets on the air it will have gone through various compression stages and data conversion/compression and whatever else. By the time they post it online this gets even worse, by adding even more of their own data compression. All of this does not help with the „problem“ I‘m trying to point out here. 9 hours ago, glenn said: You should include the TRX742.6. It has neverclip and our best A-D in a transmitter. Glenn Ok, I wasn‘t aware. NeverClip shouldn’t really come into play in these particular samples, but a better A-D could help. I can’t really add it now, though, unless I shuffle all file names, bracause it would be obvious which one would be it... Quote Link to comment Share on other sites More sharing options...
claus Posted October 4, 2018 Report Share Posted October 4, 2018 Thank you Constantin, very insightful clips. Quote Link to comment Share on other sites More sharing options...
alenK Posted October 4, 2018 Report Share Posted October 4, 2018 Hi Constantin, As a suggestion, rather than normalize the peaks to 0dBFS, it would have been possible to normalize the guide tone to a consistent level, and then use the exact same amount of amplitude required to line each of the guides on each respective sample of content. For example; if wave-1 was adjusted by +0.03dB to align the guide tone peak to -20dBFS then adjust the relative content +0.03dB. Likewise, if wave-2 was adjusted by -0.07dB to align the guide tone peak to -20dBFS then adjust the relative content -0.07dB. This procedure would account, somewhat, for the practical difficulty in aligning the guide tones with the actual wireless hardware, and it might reveal interesting inter reactive characteristics between the mic and the systems. I think it would seem interesting to see the character of the peaking before the peaks of the content were normalized to 0dBFS. For my own listening session, I went ahead and cut the start times to more exactly match each other by finding a cut point where I could identify the placement within one or two samples. I also trimmed the lengths so that they were exactly the same. This made lining up the examples for switching during playback easy. Having said that, after listening intently I can not appreciate a practical difference between any of these examples, and even though a post team may suggest that one example is more useful or malleable than some other, I suspect that the impetus for making such a comparison was inspired by the routine challenges introduced by other more influential factors than a choice of professional grade wireless system. In my opinion, this analysis suggests that issues such as talent timbre, mic selection, mic placement, ambient factors, etc. are far more influential on the final results than the choice of signal transmission system. Having said that, after listening intently I did dare to "look" at the wave files with various analysis tools. You can easily see some differences and compare what is easy to see to what seems difficult to hear. The low frequencies exhibit some variation that might excite the imagination, and the relative density of "noise" along the spectrum seems varied. Of particular note, the 4017-1.wav displays some spurts of energy up at 22kHz which is, perhaps an artifact of aliasing fold over. Is this the kind of artifact that anti aliasing filters are supposed to moderate? Regardless, of the fact that I can not hear 22kHz content, seeing the evidence of the artifact makes me curious to learn how this signal was transmitted. Thank you for sharing the wave files. Quote Link to comment Share on other sites More sharing options...
John Blankenship Posted October 4, 2018 Report Share Posted October 4, 2018 I, too, do not agree with "normalizing" them. Different systems handle peak excursions differently (key test, anyone?), so adjusting the average, or nominal, level to be the same is the way to compare. Nonetheless, thanks Constantin for doing the test. Quote Link to comment Share on other sites More sharing options...
jon_tatooles Posted October 4, 2018 Report Share Posted October 4, 2018 Constantin, I appreciate that you explained the test conditions in this speech test. It is important to note that, IMO, the source material doesn't come close to stressing these system's dynamic range, preamplifier noise performance, limiters (or lack of limiters?), and overall frequency response linearity. I believe I can identify each of the systems based more on a visual representation of the files, except for one that is sonically obvious. Choosing gear is a matter of trade-offs. Using wireless versus a cable is that first trade-off. When you can, use a cable, of course. Why use a complex, expensive wireless link when a perfectly good, high quality microphone cable at $1/foot would do? Well, we have to! One could choose a system like Sennheiser's 9000 series wireless in HD uncompressed mode and not hear any appreciable sonic difference between a cable and the wireless. However, its numerous trade-offs in HD mode (64 QAM modulation, IIR) may not be worth it. These tests are certainly very useful. At the end of the day, each of us should evaluate our tools first-hand. I encourage everyone to listen to each of these systems with their own specific setups. Quote Link to comment Share on other sites More sharing options...
Doc Justice Posted October 4, 2018 Report Share Posted October 4, 2018 4 hours ago, jon_tatooles said: Choosing gear is a matter of trade-offs. At the end of the day, each of us should evaluate our tools first-hand. I encourage everyone to listen to each of these systems with their own specific setups. Perfectly said, Jon. I couldn't agree more! Quote Link to comment Share on other sites More sharing options...
glenn Posted October 4, 2018 Report Share Posted October 4, 2018 "You are absolutely right. Whatever can be heard in differences here, would be far more pronounced in real life, so to speak. The thing is, when it comes to data compression rates, we have no tools at our disposal, the manufacturer gets to decide that." We have 5 modulations with 3 different data rates. So in our case the end user does have some control. I would be curious what modulation was used in our test. "Mono" is our highest data rate with the least data compression. Whatever was used in the test I thought the results were very good. Thanks for running the tests. Glenn Quote Link to comment Share on other sites More sharing options...
Ilari Sivil Posted October 5, 2018 Report Share Posted October 5, 2018 All of these sound good to me. I could tell which one was Zax before I looked at the spectrogram (which I did after the first listen). On this test, the difference is more subtle than I've heard on some other tests that used Sanken Cos11s. Personally, I would not mind any of these files in post, they'd all be perfectly acceptable to me. My gut feeling would be that #1 is Audio LTD, #2 is cabled, #3 is Zaxcom and #4 is Lectro. There might be more differences with a more dynamic reading, but that wasn't the point of the test, was it? I think I might now know what Constantin means by the "noise" in question. I think I heard it on the second part on the phrase "Der ausfühlichen angabe seine handlung", around "angabe". Sorry if I'm misspelling stuff, my German isn't great. Am I hearing the same thing in the upper mids? Thanks for the comparison! Quote Link to comment Share on other sites More sharing options...
Vincent R. Posted October 5, 2018 Report Share Posted October 5, 2018 thanks for the upload @Constantin. I don't have a good listening setup at my current location, but will check them when I have. Quote Link to comment Share on other sites More sharing options...
Tuhiva Posted October 5, 2018 Report Share Posted October 5, 2018 On 10/3/2018 at 10:09 PM, Constantin said: Two people sitting at a table indoors talking, there’s no ambience to speak of. It’s not hiss I am talking about, think Mp3 compression... I was listening through my monitors (old mackie HR624) and could barely hear any differences in noise between the samples.....Then today I listenened to those samples on my headphones and ok now I can clearly hear the difference. Samples 03 and 04 are noticeably "noisier" with sample 03 being the worst. Geez I thought I was going death with everyone on this discussion hearing things that I could not 😅 Quote Link to comment Share on other sites More sharing options...
OnTheSoundSideOfLife Posted October 6, 2018 Report Share Posted October 6, 2018 12 hours ago, Tuhiva said: I was listening through my monitors (old mackie HR624) and could barely hear any differences in noise between the samples.....Then today I listenened to those samples on my headphones and ok now I can clearly hear the difference. Samples 03 and 04 are noticeably "noisier" with sample 03 being the worst. Geez I thought I was going death with everyone on this discussion hearing things that I could not 😅 I think Constantin is talking about digital noise/artifacts (similiar sounding to those of lossy mp3 files), not the general noise floor. But it's so subtle, that you can barely hear it in the raw files. If you heavily compress them, it starts to become more audible. At least I think that's what he is talking about. Quote Link to comment Share on other sites More sharing options...
jli Posted May 24, 2022 Report Share Posted May 24, 2022 @ConstantinNow that it's been a few years do you want to post the true TX for each clip? It's obvious which one is Zaxcom but between #1 and #4 things are unclear. I am quite curious to know as I also noticed this: On 10/4/2018 at 9:09 AM, alenK said: Of particular note, the 4017-1.wav displays some spurts of energy up at 22kHz which is, perhaps an artifact of aliasing fold over. Is this the kind of artifact that anti aliasing filters are supposed to moderate? Regardless, of the fact that I can not hear 22kHz content, seeing the evidence of the artifact makes me curious to learn how this signal was transmitted. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.