Description image

Blackmagic Cinema Camera Audio Quality: Using External Preamps to Get Better Sound

11.13.12 @ 10:02PM Tags : , , , , ,

The folks at juicedLink have already presented us with strong evidence that using the company’s Riggy Micro/Assist low-noise pre-amplifier configuration with DSLR shooting has major advantages. First and foremost, it greatly improves the noise-floor performance of on-board recorded audio, even matching high-end dual-system recorders in its signal to noise ratio — additionally, though, using any good preamp avoids the mobility and convenience you lose with having to use a separate recorder, plus the need to sync in post. Now, juicedLink’s Robert Rozak performs another exhaustive test, this time to illustrate the benefits of using the Riggy Micro/Assist setup with Blackmagic’s Cinema Camera — plus major audio issues that must be addressed with that camera system. Check out the video below to see the results.

The video is about 20 minutes long, but I think important to watch for a better understanding of the considerations that BMCC users or on-board audio shooters in general will have to make in the field:


One has to give serious credit to Mr. Rozak and juicedLink here — they’re clearly trying to illustrate the genuine benefits of using preamps to shoot on-board audio at better-than-acceptable quality levels here, all the while explaining important considerations in the audio world, and not just trying to push their products on us. Not to mention the fact that here they’re using the results to highlight legitimate concerns with the way the BMCC handles audio recording, and even reaching out to Blackmagic to address them. I for one have great respect for the level playing field they’ve established to balance the test results, and I think the outcomes here are pretty difficult to argue with.

As far as performance goes, it’s once again indisputably clear that, if you have to or want to shoot single-system on-board, a preamp is really the way to go. Now, I’d venture to guess pretty much any dedicated and mountable preamp will give you better results than going straight from the mic into the camera, but obviously juicedLink wants you to hear the advantages of choosing their system over other competitors’  similar offerings. What’s interesting is that the BMCC design team seems to have kept audio in mind during development — listen to the non-preamped recording with the BMCC, and you hear how much better its performance is than that of a DSLR. In fact, I’d go as far as saying that non-preamped audio with the BMCC is still actually better even than the DSLR audio recorded using the Riggy Micro/Assist in juicedlink’s last shootout.

Where the Cinema Camera falls short in the audio realm is very important to note, however. The most major concern is its audio system’s DC offset — this is a malady pretty easily correctable in post (not that you should have to correct for this at all, under anywhere-near-ideal conditions), but the concern is mainly that it may cause you to mis-calculate your peak levels in a way. Basically, after correcting for the DC offset, you may find that your audio wasn’t actually as loud as it was reading — this is well demonstrated in the video. Additionally, the BMCC uses digital gain and attenuation to manipulate your levels. Robert explains that digital gain increases noise at unacceptable proportions, while digital attenuation decreases headroom, which hurts your signal-to-noise ratio by bringing your levels closer to the noise floor.

Because of these considerations, juicedLink contends that audio bracketing — the company’s innovative use of the same input signal across two separately-dialed channels (one about 16dB lower than the other) which allows you a bit of a safety net if you clip in your primary channel — is not truly possible with the BMCC. For one thing, juicedLink feels this to be a more adaptable safety measure than something like a limiter, which may have to be calibrated for each camera situation. For another, they claim their ‘properly implemented’ analog circuitry is far better suited to this practice, and given analog’s tendency to bend before breaking, this seems an entirely plausible assertion to me.

Beyond all this, Robert has asked Blackmagic to look at fixing a few issues in firmware updates down the line, including the DC offset. They’re also asking for toggle-ready audio metering akin to what Magic Lantern brings to many DSLRs, the removal of an automatic feature which switches the audio from mic to line level when mic level channels clip, and some more intuitive gain/attenuation controls. There’s a lot more to find out (and just as importantly, to hear, too) in the video, so I’ll leave the rest of the information for Robert to explain.

Were you guys surprised by the BMCC audio recording performance versus that of the DSLRs we’re used to? What about the BMCC’s shortcomings, particularly its DC offset? Is the investment in a preamp adequately justified to you for on-board audio recording?

Link: Blackmagic Cinema Camera: Audio Test — juicedLink Blog

COMMENT POLICY

We’re all here for the same reason: to better ourselves as writers, directors, cinematographers, producers, photographers... whatever our creative pursuit. Criticism is valuable as long as it is constructive, but personal attacks are grounds for deletion; you don't have to agree with us to learn something. We’re all here to help each other, so thank you for adding to the conversation!

Description image 19 COMMENTS

  • From this test, it seems that just like DSLRs, dual system sound would be essential on a BMCC shoot for any kind of critical audio capture.

    • Dual system sound is essential on any shoot for any kind of critical audio capture. This has always been the case, long before digital.

      • Thank you for watching the video! … I know it was quite long …

        Dual system audio is not necessarily essential for any kind of critical audio capture. But, there are instances when it is a benefit.

        For many instances, dual system provides no additional benefit. Assuming you are recording direct to the camera (using a juicedLink low noise preamplifier and achieving as good (or, perhaps even better, depending on the recorder) signal-to-noise performance as you would in an external recorder), then the external recorder doesn’t really have value. I’m talking about the case where your camera has headphones and meters, so you can monitor exactly what is being recorded (I understand the desire for a dual-system for the old-style DSLRs which didn’t have those features, because you couldn’t monitor in the camera itself). Hey, if your camera isn’t properly recording the audio, then it has some fundamental issue and likely isn’t properly recording the video either.

        Where I could see a benefit would be where you wanted to do some audio bracketing (multiple recordings at different levels) for backup in case you overload a channel. But, the new juicedLink Riggy preamplifiers have the option for implementing audio bracketing right in the camera. So again, no real benefit to dual system there, for simple setups (shooting scenes with dialog, interviews, etc).

        Now, where I do see a benefit of dual system is in something like the production of reality TV. Here, they want to have an audio engineer doing a live mix recording (although, it would typically be recorded in a separate device, with the camera simply being a scratch track (since the camera needs to be mobile)). With 5 microphones running at once, that’s a challenging live-mix. So, it’s beneficial to have each mic also recorded separately (bracketed at a lower level) so you can pull from them in post in case a section of the live mix gets blown out.

        Specifically regarding the BMC, they really should fix those issues. But, now that you understand what the issues are, they can be managed. That’s why I made the video …

        JMHO. Thank you for your comments! You can contact me on the website if you have further questions …

        - Robert from juicedLink

        • Thank you for the comparison video. This video and the other video your company released do a good job showing how higher quality microphone preamplifiers can improve the audio signal chain.

          Unfortunately, even with a nice preamp, we’re all still at the mercy of the camera’s AD conversion and associated analog circuitry. It would be great if a camera could have a digital audio input, and then you could provide a nice AD as part of your product as well.

          One question- At 13:28, was the JL/DSLR recording made at the same time as the JL/BMCC recording? The level of the voice appears to be a lot lower, and that is going to skew the results of a listening test. In any case, to my ears in this video, the BMCC sounds vastly inferior to the DSLR in sound quality. This was a surprise to me.

          It’s interesting to hear all the devices back to back, have you ever done a similar test using a high quality condenser microphone, such as a schoeps or sennheiser? That would be more similar to a real world shoot and would be interesting to listen to.

          • Bove,

            The test recording of the BMCC is new, and the other recordings were taken from a previous video/test. So, they were not made at the same time. The clips were normalized to -3dB. But, the entire recordings are longer than what is shown in the video, so perhaps the peak that the normalization operated on was outside the clip played back on the video. That could explain the difference.

            Anyhow, to compare signal-to-noise, what we are really listening to is the difference in the noise floor.

            These tests came from a challenge that was posed from a blogger, to use a dynamic mic to really stress the preamps to see which performed the best in terms of signal-to-noise with a low input signal level.

            Regarding the analog circuitry and A/D in the camera, that doesn’t bother me at all. The analog circuitry in the camera is completely inconsequential with the use of a preamp (see “cascaded noise figure of amplifiers”). You’re going to have an A/D somewhere in the chain … 16b versus 26b A/D … I don’t know about you, but I can’t hear 144dB of dynamic range (the AES had a nice paper basically showing this as well).

            Now, where I would agree with you where there can be more of a difference between recording direct to camera and the really high-end of external recorders, is frequency response. The juicedLink has a super-flat 20Hz-20KHz frequency response, just like the high-end of external recorders. To get down to 20Hz, you need big honking coupling capacitors, which they can’t fit into the cameras. So, a SD702 recorder will have a better low-end frequency response than recording into the camera. Now for video production (scenes, dialogue, etc) no big deal. Even indoors, many people engage the high-pass filter on their mics (to avoid boom handling noise, and you even get low rumble wind noise INDOORS when moving a mic while booming). So for those productions, no big deal. Now, would I record the Chicago Symphony to a camera or a SD702 … for that, the right tool is the SD702.

            Great discussion … thanks!

            -R

          • Dave Kendricken on 11.14.12 @ 4:40PM

            Great discussion is right, thanks for the comments and the great work on the presentation Robert! Appreciate the nuts-and-bolts insight as well!

  • I find these videos so frustrating. I’m coming to video from audio, and it’s clear that there is a lot of misinformation that gets posted in charts and passed off as scientifically measurable facts.

    I’m all for preamps, and while I don’t have any experience with these I’m sure they’re useful in many situations. The size and form factor are also big plusses for filmmakers. But the suggestion that recording with a middling condensor directly into a camera establishes your pre’s superiority to a dedicated recorder (let alone one doing 24/96) is snake oil.

    You call this a “low noise” preamp, but I think it’s more accurate to call it a high gain preamp. If I’m reading the specs right, the pots give you a dynamic range of about 35dB of gain. To get the amount of gain necessary to boost a dynamic mic’s signal to within the range one would need in this test, it’s got to adding gain before the signal hits the pot (or, more likely (and I can’t say for sure without taking it apart and looking, but I know enough about this type of circuit to be reasonably sure), it’s adding some large amount of gain (we’ll say 100dB for argument) and then attenuating with the gain control). So the dynamic range of this pre isn’t 0-100dB, it’s 65-100dB. For a good percentage of what most people shoot, this is extremely useful, and that’s great. But if I shot a live rock band in a small club, even with the gain at zero I’m going to be blowing out these preamps unless I attenuate the signal before it gets to the input. Not so with a Zoom H4n (for instance).

    But of course it would be crazy for me to claim that the above proves your pre is a crappy product. It doesn’t. It proves that in a specific high-stress situation, yours will fail. Similarly I’d suggest NOT claiming that your test establishes your pre as superior, because it doesn’t. It establishes that your high gain circuit has a better noise floor when stressed, but let’s be honest: those audio samples all sound like crap. This test is like comparing which lens shoots better through nylons.

    The way to establish that one tool does a better job than another is to test it in an application it’s going to be used for. Which is why I, too, would rather hear practical real-world tests of mics filmmakers might use recording the sorts of things filmmakers might record. Why not record a range of sounds (more than just “testing 1-2-3-4″)? Record some ambience, record some boomed dialog, record some lavs, record some foley – record the sorts of things filmmakers do. If you want to push it, use it like an audio engineer would – record some voice of god VOs, use it with a LD condenser and a range of different SDs. Even a ribbon (you want to show off high gain with beautiful results? There you go).

    Whatever advantages this pre gets from it’s high gain will then be moot, and we can listen to what matters: how it sounds. My guess is that this preamp will do okay in that sort of head-to-head, landing somewhere in he middle of the pack. Which is great! The price point and form factor will make it an easy sell to filmmakers.

    My point is you have a product with plenty of advantages for filmmakers, and I’m sure you can market it without resorting to hype and non-science to sell it.

    And say what you like, but a quality A/D converter (especially in a dedicated device) will sound better better than any DSLR’s A/D converters. Cameras don’t do A/D particularly well – it’s a processor-intensive activity, and whatever processing power cameras have is dedicated to making pictures. I’m not sure which AES paper you’re referring to implying that it doesn’t make a difference, but I’m going to guess that that’s an oversimplification of what the paper says.

    • ** correction: I’m coming from audio to video.

    • Colin,

      Thank you for your thoughtful comments!

      You might also be more referring to a blog post that I did a couple of months ago, comparing recording directly to a camera through a JuicedLink low-noise preamplifier, vs. numerous recorders (Sound Devices 702, Tascam DR680, Edirol R44, Tascam DR100MKII, Zoom H4n):
      http://juicedlink.com/blog/2012/09/audio-recorder-shootout-sound-devices-702-dr-100-mkii-h4n-dr-680-and-more/

      As an audio guy, when you use an audio analyzer, such as the HP8903B (which injects a tone to measure the signal output from the DUT (device under test), and then turns the tone off and measures the noise output from the DUT, to caclulate SNR, EIN (@ a specified gain), etc), you need to make a determination as to what signal level you’re going to inject into your device under test. That choice has a significant effect on which your measured signal-to-noise performance is going to be. You could use an unrealistically high signal (which would give you very favorable signal-to-noise measurements). But, you don’t want to use an unrealistically low signal level that would never be seen in the real-world either If I post tests using the HP8903B, how do I communicate to people that I’m using a realistic signal level?

      Then, the bloggers from the ProVideoCoalition challenged me to use a dynamic microphone which has low sensitivity, to really stress the system to evaluate signal-to-noise. I thought that was a beautiful suggestion, because right there in the recording, I can demonstrate that this is a real-world signal level,
      and it is going to be on the low-side of the expectation to really stress the front ends (but still be real-world applicable). Then, people can listen to the silent period and compare the noise floor for themselves (acting like a human HP8903B). So, I ran with that and conducted the test and shared them, and found some very interesting results, and shared them.

      When evaluating the clips, what you are really trying to listen to is to observe the difference in the noise floor, and compare them between the clips.

      Of course, you can use something more sensitive than a dynamic microphone, and the difference in the signal-to-noise results in the tests will not be as great. But, not everybody has a $500 50mA/Pa ME66. Lots of people use maybe NTG2 with 15mV/Pa (which would be boomed further away than where the handheld dynamic mic was placed), or a lav with 5mV/Pa sensitivity, and even handheld dynamic mics. So, it’s a great test to compare signal-to-noise with real signal levels.

      “But the suggestion that recording with a middling condensor directly into a camera establishes your pre’s superiority to a dedicated recorder (let alone one doing 24/96) is snake oil.” – If the front end of your system is the long pole in the tent when it comes to signal-to-noise performance, and it does not matter if you’ve got a better A/D running 24/96. That’s the whole point of the test. Some recorders (SD 702 and DR680) had very good front ends and had very signal-to-noise results along with using the JuicedLink low-noise preamplifier directly into the DSLR. I didn’t say the juiced link was better, but I was very proud that these could be mentioned in the same sentence as performing very well. Other recorders that were running 24/96 resulted in very poor signal to noise performance in the tests (R44, DR100MKII, H4n).

      So, it’s not “snake oil” that you can achieve better signal to noise performance using the JuicedLink directly into the camera (compared to that set of recorders that performed poorly with a dynamic microphone), it’s a substantiated fact as illustrated by the tests …

      So, my point is if you’re thinking of one of the recorders like the R44, DR100MKII, H4n, perhaps you should consider instead the SD702 or DR680 or the juicedLink direct to camera. I make the case that there are production flow advantages to using the JuicedLink (no need to synchronize and post, small, light weight, power efficient, never forgot to record your audio in the separate device (by having to hit record … twice), instant boot time for eventing, plus it’s Riggy so you can easily mounted to your camera and mount accessories like mikes and wireless receivers).

      With equally good front ends, sure the 24/96 will be better. But, as I mentioned in my reply to Bove above, I can’t hear 144dB of dynamic range, can you? The percepticle difference in terms of signal-to-noise is a second/third order effect. At this point, what will make a bigger difference is the frequency response because of the smaller caps in the camera. The freq response for dialogue is just fine in a camera. Would I record a symphony in a camera? No, I’d use a SD701 (or a H4n with a juicedLink low-noise preamp front end attached to it).

      “But if I shot a live rock band in a small club, even with the gain at zero I’m going to be blowing out these preamps unless I attenuate the signal before it gets to the input.” – Here is what the preamp headroom is, referenced to its input:
      - MIC, HI GAIN: 35mV
      - MIC, LO GAIN: 350mV

      The ME66 is a super sensitive mic at 50mV/Pa (to use as an example to be the least favorable for headroom in the preamp). So, to reference this at the mic input in Pa:
      - MIC, HI GAIN: 0.7Pa
      - MIC, LO GAIN: 7Pa

      To put this in perspective, a jackhammer at 1m is about 2Pa. So, the preamp has the capability for gobs of headroom. Plenty to handle a recording of a rock band. If you need more, engage the MIC/LINE switch to LINE and you’ll have 700Pa of headroom (and your ears would be bleeding).

      Thanks again for the great discussion …!

      - Robert from juicedLink

      • The only thing the audio from your stress tests clearly establishes is that nobody should be using a cheap dynamic mic to record dialog because all of your sound samples sound uniformly terrible. Yes, I get the S/N point you’re trying to make, and you’re right: a high gain amp will get a better s/n ratio under those circumstances. But who in their right mind would buy a preamp based solely on s/n? Your videos are all described as “versus,” “shootout,” and “audio test review”, not “signal to noise comparisons,” so you can understand how this might be seen as misleading. How about letting people compare sound quality or the character of the preamp?

        When I listen to the audio from your NTG2 test, I hear a pretty steep slope from around 1khz down, removing the richness of your voice. My suspicion is that that’s the A/D converters, not your preamp, but it sounds thin and in need of some serious attention in post. And were I expecting to fix it, I’d much rather edit the 24/96 file than the 16/48.

        I remain unconvinced that the on camera A/D converters are great at their job, or better than a dedicated external. I’m fine with syncing sound – it’s not that hard, and everything goes out to PT anyway. But there are certainly situations where the convenience would make your pres a good choice. Over a bed of music that recording is going to be fine.

        Still curious which AES paper you were citing as well.

        Just it reiterate: I think there’s a definite market for your pres and they are a good thing. People should buy mic pres because they’ll help with the audio. Just… sell them for what they are, not based solely on s/n ratio or whatever highly skewed test you can come up with. You can do better. Spreading misinformation and misunderstanding of basic principles in a field where audio seems to be treated as a black art is a disservice. I think you know enough about what you’re doing that you could help a lot of people, and I don’t think you’re intentionally misleading anybody. But you’re making an unreasonable comparison here, and it’s frustrating.

        You can do better.

        • Colin,

          “The only thing the audio from your stress tests clearly establishes is that nobody should be using a cheap dynamic mic to record dialog because all of your sound samples sound uniformly terrible.”
          - No. What it establishes is that the front end matters when it comes to signal-to-noise. A 24/96 system with a poor front end will not perform as well as even a 16/24 system with a properly designed front end. If you’ve got a poor front end running 24/96, then you will be able to listen to the analog noise of the front end with all of the clarity and fidelity that 24/96 will provide you. the low-noise preamp fixes the front end. It’s called the “cascaded noise figure of amplifiers”. Just the same way that the juicedLink low-noise preamp makes the signal-to-noise performance of the 16/24 camera able to “be mentioned in the same sentance” as the SD 702, I can take the juicedLink low-noise preamp and fix the front end of the Zoom H4n as well.

          Regarding frequency response, I think I’ve already mentioned the cameras will not have as good frequency response because of the smaller coupling caps. A camera is not the right tool for recording a Symphony, but just fine for dialogue. Frequency response is not the same thing as signal-to-noise, that the videos are demonstrating.

          Most humans can not perceive the difference between 16/24 and 24/96. AES paper:
          “1. Meyer, E. Brad and David R. Moran. Audibility of a CD-Standard A/D/A Loop Inserted into a High-Resolution Audio Playback, Journal of the Audio Engineering Society, Sept. 2007, pp. 775-779. This reference is discussed in the article The Emperor’s New Sampling Rate”

          I’m happy to agree to disagree with you. But, I will NOT accept that I am in any way spreading misinformation …

          Best regards,

          Robert from juicedLink

          • Bove, Colin, and everyone … Thank you for taking the time to watch the videos, and for the engaging conversation. But, the direction of this conversation has gotten me a little “amped” (no pun intended). I feel I’ve adequately addressed questions, concerns, and how things work to my satisfaction. Even if it is not yet to your satisfaction, I’m sorry, but I need to step back, move on, and get some other work done. And so, it will just have to be …

            If you, or any other readers have further questions, you can always contact me through the juicedLink website.

            Best regards,

            Robert from juicedLink

          • I disagree with you, Robert, but I appreciate your taking the time to reply and hope other here do as well.

            I do consider it misinformation to call a s/n comparison anything broader than just that, and think you’re doing a disservice to your product. But I think we’ve both been clear on that – anyone who’s interested can make up their own mind.

            Regarding Meyer and Moran’s paper, that isn’t a discussion of A/D converters. It was a blind test of audio playback. While I basically agree with it’s results (if you played me music I was unfamiliar with one time, fading in and out of 16bit and 24bit playback, I probably couldn’t tell the difference), I can guarantee you that none of the authors of the paper would suggest recording in your output format if you want to do any editing. The same logic that leads people to shoot RAW suggests that recording at 16/44 or 16/48 is not the best idea. Whenever possible, I prefer to go the extra step and do things right.

          • Well … damn … if you didn’t catch me while I was still reading stuff on NFS …

            The “Recorder Shootout” video plainly has “Signal-to-Noise Challenge” labeled across the top for the duration of the slideshow during the video:
            http://juicedlink.com/blog/2012/09/audio-recorder-shootout-sound-devices-702-dr-100-mkii-h4n-dr-680-and-more/

            Shooting video in RAW is completely different. You need greater bit depth in capture because you are going to apply a non-linear gamma curve in post. Changing the frequency response in post for audio will linearly boost/attenuate the signal (at a given frequency). Greater bit depth for frequency response manipulation of an audio signal in post is of no benefit.

            … Checking out (for real) this time …

  • Christian Anderson on 11.14.12 @ 6:43PM

    The “JuicedLink” song at the end is amazing. :D

  • I’m sorry to hear that the Blackmagic audio right-out-of-the-box is looking grim. I have a Juicedlink that I use with my SLR. And while for the most part it does a great job, I had interference issues with it (a radio station coming through loud and clear) one day that killed me, and a Zoom came to my rescue (and was clean). So I’m pretty nervous about relying on the JuicedLink now. I look forward to the day that I can record top-notch multitrack audio right to the camera. With the image quality race careening towards “everything just looking great”, I think it’s those sorts of features that are going to help cameras stand out. Syncing sound in post is for cavemen.

  • Robert does good work, and I wouldn’t call anything he says misinformation, but he definitely has a bias that affects what he says, as any business owner would.

    I spent many hours considering his juicedlink preamps because he is very effective at marketing them to the DSLR/low cost market, but in the end bought a Sound Devices mixpre preamp used, for about the same cost as a juicedlink, and it is much more functional and better specd in almost all ways. I doubt you will find any tests on his site comparing his preamps to Sound Devices because they would lose bigtime, even against a Sound Device preamp feeding cheap recorders like the $100 DR-05.

    At any rate, information is worth something and Robert deserves respect for posting more detailed technical information on budget cameras than most if not all other companies in this business. I learned a lot from the video.