(Page 2 of 3 pages for this article  <  1 2 3 >)

Filed under: *VIDEO* •Audio •Cameras •compression •Distribution •Editing •Hardware •Post Production •Production •Software •Vendor Channels •

NewTek’s TriCaster TCXD300 ignites 3rd multicam revolution, now in HD

Differences between “locked” USA version, versus unlocked multi-standard version

Every time you create a new project with the TriCaster TCXD300, you decide the project’s spatial resolution. With the “locked” USA version, you get three options: SD, 720p59.94, or 1080i59.94. With the unlocked multi-standard version, you also get the options of PAL SD, 720p50, and 1080i50. If you are absolutely sure that you are only producing for the web, then you can probably get away with the “locked” USA version no matter where you live or where you plan to deliver, since the web doesn’t care about the framerate, and even in Europe, HDTV sets accept any framerate. (The opposite is unfortunately not always the case in the USA.) However, if you are also planning to record for traditional television delivery (and that might be a 50Hz delivery for at least some of your projects), you will certainly want to buy the unlocked multi-standard version, despite the US$3,000 increase in price. It is much better to purchase the unlocked multi-standard version from the beginning, since upgrading it later can be cumbersome.

About the TriCaster TCXD300’s webstreaming capability

Currently, the TriCaster TCXD300 encodes your HD signal as H.264 (my current favorite códec for distribution) and embeds it into a Flash wrapper. Despite my general concern about Flash’s non-compatibility with many mobile devices, like the iPad, iPhone, and iPod Touch, this is really a non-issue since you only send the biggest (720p) signal to your multicast service. Then your multicast service can send out multiple versions and sizes, including unwrapped (non-Flash) versions for these mobile devices.

According to Jorge Dighero —who gave me my extended demonstration at Pro Audio & Video Systems in Miami— the TCXD300 is also field upgradable to handle other web códecs which may become popular down the line.

I applaud NewTek for choosing to do live webstreaming with 720p, regardless of whether a given project is set for 720p or 1080i. In another recent article, I explained all of the reasons why presently, 720p is ideal for the web in almost all cases, together with the very uncommon situations where 1080p for the web should be considered today.

Internal video processing

Since I know some digital video mixers (“switchers”) unfortunately subsample the incoming video, I was concerned do find out what was really happening inside the TriCaster TXCD300. These are the questions I sent to NewTek:

• If a true live 10-bit camera HD source feeds the TCXD300, while it is full frame, does the signal ever get reduced below 10-bit internally?
• If that same source is blown-up (expanded), does the signal ever get reduced below 10-bit internally, or does it go to floating point at better than 10-bit?
• If one imports a graphic —including a color ramp— that was created at 10-bit to avoid banding, does it remain 10-bit, or does it ever get reduced to 8-bit?
• Do the TCXD300’s internal graphic creation tools create color ramps at 8-bit or 10-bit?

Here is what Andrew Cross —the VP of engineering at NewTek— responded:

This question is actually very hard to answer completely without getting into a lot of details simply because we do not work internally quite like other switchers. I will give a simplified explanation. The switcher itself does not work with just native frame-buffers of one format. Most traditional switchers work as follows:

Input -> (inside switcher) color convert if necessary -> (inside switcher) scale to switcher res -> switcher processing.

We do not do this, rather every input to the switcher has an arbitrary buffer on every input that can be in a large variety of formats (RGBA, YCbCrA 4:4:4:4, YCbCr 4:2:2, YCbCr+A 4:2:2:4, YCbCr 4:2:0 - this has big benefits, etc…) at a variety of different bit depths (FP, 16bit UI, 8bit UI), fielding formats (fielded, progressive), resolutions and aspect ratios. The processing engine of the switcher can use any of these buffers and combine them natively into the output without intermediate resizing or color conversion steps. In addition it analytically combines all color space transforms to avoid any intermediate steps, a simple example:

Color conversion RGB ->YCbCr is a linear transform that can be represented by a matrix. Similarly procamp color correction can be represented by a matrix. Normally one would first apply color conversion (which might clamp to the color space if we were not using float) and then apply color correction; instead we fold these two operations into a single mathematic transform that can go from the source RGB color space to the color corrected YCbCR space without any intermediate steps.

The reason I mention all of the above is that it is very hard to give a single answer to all of the questions below because we preserve image fidelity at its source format until it needs to be converted to another. Some relatively imprecise answers below ...

If a true live 10-bit camera HD source feeds the TCXD300, while it is full frame, does it ever get reduced below 10-bit internally?

We do not reduce precision on any source (including DDRs) until a precision loss is actually required by the processing. (e.g. conversion to 8bit for recording to disk or streaming.)

If that same source is blown-up (expanded), does it ever get reduced below 10-bit internally, or does it go to floating point at better than 10-bit?

Yes. Source sampling from images is done in floating point. 4:2:2 and 4:2:0 buffers have interpolation on the missing chroma to avoid artifacts.

If one imports a graphic —including a color ramp— that was created at 10-bit to avoid banding, does it remain 10-bit, or does it ever get reduced to 8-bit?

We keep files read off disk in their native format through final composition.

Do the TCXD300’s internal graphic creation tools create color ramps at 8-bit or 10-bit?

The titling tools actually generate 8-bit RGBA images (that is the format all on screen imagery would be in.) This obviously also applies to images placed in the DDR (PNGs, etc…) In all cases where we get 8-bit data in this form we perform full 2D error diffusion conversion to YCbCr instead of 8bit integer conversion (which is what every other product I know does.) This makes a phenomenal difference to the images and removes any banding in the color conversion process with no added noise (how most apps avoid banding.)

I am impressed with the way NewTek has decided to handle the internal processing in the TCXD300 from camera input to output, especially with all of the features it offers at its price point. Andrew’s response makes a lot of sense and explains from a technological perspective why everything in Jorge’s demo looked so clean on the high-end JVC HD-SDI program monitor he had connected. It seems that the only minor compromises have been with the DDR and internal graphic creation, since they are only 8-bit… but then again, so are most HD camera códecs and most HD tape formats are 8-bit too. As far as recording, I have already explained that it can be done externally at 10-bit if required, and if the budget permits it. More details ahead.

TriCaster TCXD300’s own audio/video recording

As stated earlier, all HD internal recording in the TriCaster TCXD300 is done with MPEG2, 8-bit i-frame 100 megabit/second recording, either 4:2:0 standard profile or 4:2:2 high profile. In addition to embedding audio in the MPEG2 file, there is the option to have the system record audio separately as an MP3 file, in order to make the signal more compatible with certain Mac applications which demand the audio be separate. I’d like to see a further option to have that separate audio recording be selectable among AAC (M4A), MP3, WAVE, or AIFF.

By default, the internal recording is sent to the TCXD300’s internal RAID0. However, if you require sneakernet capability (to be able to grab a drive and run just after a recording) and/or RAID5 redundancy (because you are afraid that the only recording might disappear due to a drive failure), Jorge Dighero confirmed to me that you can directly connect an external hardware RAID via eSATA, since the TCXD300 fortunately does have such a port. Then you can choose to direct the recordings to the external RAID. I would strongly consider the Promise DS4600 for that purpose, especially because of its cost, reliability, and ability to do RAID5 without a controller card. And if you want to facilitate sneakernetting of the recording to a Mac system, I would (after Jorge Dighero’s blessing to the idea) install MacDrive in the TCXD300, and leave the RAID5 formatted as HFS+ as opposed to NTFS. Jorge also clarified that the TCXD300’s own recording is either on the internal RAID0, or external, but not both.

If you must have 10-bit 4:2:2 recording, and/or a simultaneous recording both on the internal RAID0 and externally, then consider adding an AJA KiPro recorder to the TriCaster TCXD300, which would connect via HD-SDI. The KiPro is already HFS+. The KiPro does not currently offer redundant recording by itself (RAID1 or RAID5). However, in this case you could have redundancy by having one recording inside the TCXD300 (8-bit) and one in the KiPro (10-bit).

Are genlockable cameras required? What difference does it make?

NewTek highly recommends using genlockable cameras with the TriCaster TCXD300, and sending sync (either black or tri-sync) to the cameras and to the TCXD300, using an external device, like Blackmagic’s US$295 Sync Generator.

Blackmagic’s Sync Generator is part of their Mini Converter family and provides six simultaneous outputs, either black or tri-sync.

However, using genlockable cameras is not obligatory with the TCXD300. If you don’t send sync to the TCXD300, then it will automatically activate the frame synchronizers on each input. In any system, the use of frame synchronizers always means additional delay. As Jorge Dighero explained to me during his demonstration, the TCXD300 has inboard audio delays which compensate for the video delays in the frame synchronizers (when they are active), so there will be no lip sync issue in the recorded signal, or the broadcast or webstreamed signal. However, monitoring the the delayed output can be quite uncomfortable unless the control room is 100% acoustically isolated from the studio, since you would see a lip sync issue on the monitor (versus the live, acoustic audio). This is also troublesome in live applications where the output of the TCXD300 is connected to video projectors. That’s why it is recommended to use genlockable cameras, although some budget situations may have to go without genlock.

Non-genlockable HD-SDI cameras for TriCaster TCXD300

If you’re on a very tight budget and can stand the lip sync issue on your local program monitor (or have acoustically isolated your control room from your studio), you might consider the Sony EVI-HD1. This camera has direct HD-SDI output, is multi-standard to the max (including 720p25, 720p29.97, 720p50, 720p59.94, 1080p25, 1080p29.97, 1080i50, and 1080i59.94) and has a list price of US$4,028.

Sony’s EVI-HD1 robotic HD camera eliminates the cost of a tripod, lens controls, intercom, and potentially reduces the number of camera operators from three to one, for a 3-camera system.

The EVI-HD1 is robotic, and several can be controlled by Videobotics’ CosmosHD1 software which actually acts a a software based CCU for multiple cameras simultaneously for only US$1,199.95 plus a laptop and cabling. As of the publication date of this article, I have never seen the image from the EVI-HD1, but Jorge Dighero says that it works great with a TriCaster and has a very good image for its price. Also, Videobotics sells packages including the EVI-HD1 cameras with their CosmosHD1 software.

The CosmosHD1 software comes with a ShuttleProV2 which facilitates its use to control the cameras and recall presets and user-controlable macros. It can also be network-remote controlled (if you wish) from anywhere, including LAN or Internet.

If you’d prefer dedicated hardware instead of a laptop, Sony offers the BR-RM300 for US$1,575 plus cabling. One BR-RM300 can control several EVI-HD1 cameras, both the pan/tilt and mini-CCU functions.

On the other hand, you might consider Sony’s recently introduced HXR-NX5 (the first NXCAM), which I announced immediately after its embargo ended, and then was later covered in much more detail by Adam Wilt here, and then updated here. Since Adam and I each had very different emphasis in our articles, I recommend reading all three links offered in the previous sentence, in order to get as much information as possible.

Sony’s HXR-NX5 (the first NXCAM)

Sony’s RM-1000BP mini-CCU is compatible with the HXR-NX5, among many other Sony cameras.

The list price of the HXR-NX5 is US$4,950 (as opposed to US$4,028 for the EVI-HD1), so the difference per camera —at list price— is only US$922 per camera. However, you will pay over US$10,000 more for three HXR-N5 cameras once you add the cost of the tripods with lens controls, mini CCUs, and an intercom system, compared with the EVI-HD1 with remote/mini-CCU (not counting the savings in camera operators from three to one). Here is a comparison of these three cameras —at list price— to feed the TCXD300. I have left out the cabling, but that should easily be covered in either case by a discount you’ll get on the cameras.

3 EVI-HD1 HD robotic cameras US$4,028 x 3 = US$12,084

1 BR-RM300 remote control/mini-CCU (or ± Videobotics’ CosmosHD1 with a laptop) US$1,575

Total: US$13,659

3 HXR-NX5 US$4,950 x 3= US$14,850

3 RM-1000BP mini-CCU US$1,309 x 3 = US$3,927

3 tripods with lens controls at ±US$1,200 x 3 = US$3,600

1 Datavideo ITC-100 intercom system ±US$1,500

Total: US$23,877

So you would save about US$10,218 overall by choosing the EVI-HD1… or spend US$10,218 more if you choose the HXR-NX5. The HXR-NX5 probably has a slightly better picture, but for a studio situation, its design makes it much more costly than you might have first imagined, compared to the EVI-HD1.

Of course, if you choose to go without genlock, you could use even lower-cost cameras than the ones mentioned so far, as long as they at least have a live HDMI output. In that case, you’ll need a converter box that goes from HDMI to HD-SDI like the AJA HA5 (US$490 plus US$40 for power supply = US$530), Blackmagic HDMI-to-SDI (US$495 including power supply, per camera), Convergent Design nanoConnect (US$395 per camera, including power supply), or the Datavideo DAC-9 (US$400 per camera, including power supply). However, keep in mind that some cameras have a continuous, undefeatable menu superimposed over the live HDMI output, so make sure before committing to a particular camera.

Genlockable HD-SDI cameras for TriCaster TCXD300

If you’ve decided that you won’t be able to stand the lip-sync issue on your local program monitor and cannot isolate your control room acoustically from the studio… and are seeking the lowest price HD-SDI cameras with genlock on the market, then you might consider the following two categories:

Conventional HD cameras with included genlock and HD-SDI output, which will require mini-CCU/full CCU/triax, power supply, tripod with lens controls, camera operator, and studio viewfinder:

• JVC’s GY-HD250
• Panasonic’s AG-HPX300
• Sony’s PWW-EX3

Robotic HD cameras with genlock, their “gotcha” HD-SDI output card, and remote/mini-CCU:

• Panasonic’s AW-HE100
• Sony’s BRC-H700 and BRC-Z700

Videobotics also offers packages with the Sony BRC-H700 and BRC-Z700 and their CosmosHD software to control pan/tilt and mini-CCU functions.

Ahead in this article

On page 3 you’ll find:

• Virtual sets and virtual live cameras
• Character generator keyboard compatibility
• All about black level, setup, and pedestal
• Waveform monitor panic, and response
• Wish list
• Conclusions

Click here for page 3.

(Page 2 of 3 pages for this article  <  1 2 3 >)

WSmith,

I am very glad you liked the article. Yes, TriCaster is best used for live-to-disk, live-to-air, or live-to-web production. The live-to-disk is especially important when little or no post-production can be done.

If you re-read my review on KiPro, you will see ways to make it work even for Windows, by adding MacDrive to your Windows machine, and downloading the ProRes422 códec for Windows (read-only) from Apple.

Allan Tépper

Posted by Allan Tépper  on  02/23  at  12:14 PM

---

Thanks Allan,

How about that! I learn something new every day around here. I wasn’t aware that ProRes was available for Windows.

I’ve been a user and fan of Cineform so I really haven’t been paying much attention to ProRes.

Posted by wsmith  on  02/23  at  12:33 PM

---

Name:

Email:

Location:

URL:

Smileys

Remember my personal information

Notify me of follow-up comments?

Submit the word you see below: