Webinar: Peltrix takes Dante over distance with Blue Note Entertainment Group | Audinate | Dante Pro AV Networking

Webinar: Peltrix takes Dante over distance with Blue Note Entertainment Group

Over the last year, Audinate (makers of Dante), the Blue Note Entertainment Group (of the iconic Blue Note Jazz Clubs) and Peltrix (an integration firm) collaborated on an industry-first demonstration for long-distance, networked audio-video links. In June of 2021, they demonstrated their solution linking three locations up to 750 miles apart with multiple streams of Dante audio and Dante AV video links.

In this demonstration, system latency was so low, musicians were able to comfortably play together as if they were in the same room. No click tracks or guides were used, and timing was natural for performers. Dante AV played an important part in the demo, allowing musicians to see each other, taking visual cues, read body language and engaging with each other seamlessly. The products used are all commercially available today.


Questions and Answers

AV Engineer Perspective - Full Production Access

Q: Were multiple audio and video channels sent between locations, or was it just a composite feed?  

Every site received individual channels for every instrument mic, talkback mic, tech intercom feed, submixes and multiple camera streams.  This is why we showed Dante Controller, so you could see everything was available to anyone.

Q: What is the maximum number of audio and video streams you can send?

In our Dante Level 1 Certification course, we offer these numbers to help people budget bandwidth (and put their mind at ease).  We don’t change anything for distance connections, so bandwidth calculation would be exactly the same:

• Audio: a 4-ch, grouping of 24-bit/48kHz audio will use about 6Mbps.

• Video: a 1080p/30 Dante AV stream is about 100Mbps.

• Keep some bandwidth in reserve.  20%-30% is a wise amount.

Q: What is the minimum network up and down speed connections?

In the webinar, we mentioned an iHeart radio link running on a 10Mbps connection.  Above, you can see an audio stream would use about 6Mbps in each direction, so that would probably be a good baseline… it just depends on the number of channels and streams you want.

Q: How much bandwidth did you consume per site?

Our link usage varied by site and by event.  It is safe to say some distance links were using at least 50% of capacity, but our bandwidth planning ensured we wouldn’t exceed 70% of capacity.

We never had to deny a link to economize.  In the planning phase, we sized our distant network connection for the equipment we planned to support.  That worked very well.

Musical Questions

Q: Will full performances be posted somewhere?

For copyright reasons, we can’t publicly post the entire performance.  We are authorized for the 20-second clips we used just as a demonstration of what we achieved.

Q: Did it take a while for musicians to get used to the system?

Musicians were always impressed with the system, they adapted right away.

There was a period where studios started adding isolation rooms to reduce mic bleed.  For instance, they’d want to reduce the sound of the drums in the lead vocal mic.  In a way, this is the same situation - you just can’t give your team member a high five in the control room after a good take.  They’re now 750 miles away.

It is natural for musicians to shoot the breeze before and after the performance.  That helped them warm up and cool down together, restoring some of that comradery

Also, from the first performance, Robert Glasper mentioned that he knew the drummer wasn’t in the room because the physical impact on his body wasn’t there.  Once we added a speaker in the room, musicians felt more connected.  

We were impressed how little SPL it took to restore that impression, too.  The engineers started at a modest level, seeing where the musicians wanted it.  Everyone asked for it to be turned down a touch.  But they loved it!

Q: Is there a maximum distance the performers can be apart?

Obviously, latency will grow as the distance expands.  Dante can probably establish a connection with outstanding conversational latency anywhere on the globe.  But for music, latency requirements will be tighter.

It is hard to say what the maximum distance is, because the reach is about latency.  Latency is a factor of fiber paths and data centers.

But, suffice it to say a coast-to-coast connection might be possible in certain circumstances.  Shorter reaches are more forgiving.  This is the type of thing we would need to discuss on a case-by-case basis.

Forward Looking & Planning Questions

Q: When can we replicate this in the field?

• Audio: This can be done right now.

• Video: Local networks today, distance networks by the end of the year.

• Training: Dante Certification Level 3 (2nd Edition) complete in Late October.

For audio, everything used shipping products and shipping firmware, so this is available today.

For video, Dante AV works on a local network right now.  For distance, we used some new features not yet released to enable the long-distance link.  Some are in testing now, others (like Dante AV in Dante Domain Manager) will be released around the end of the year.

If you want to learn how to use Dante in a Layer 3 space, Dante Certification Level 3 (2nd Edition) is highly recommended.  This on-demand class is in production right now, and will be available at the end of October.

The second edition classes are substantially revamped with significantly more information in them.  If you have taken the older versions of Levels 1-2, it is worth taking the second edition so you are properly scaled up.  Level 2 in particular is significantly enhanced, and will better prepare you for this new class.

Dante Certification Level 3 (2nd Edition) provides a clear explanation of Layer 3 networks and the role for Dante Domain Manager in that environment.  Without a doubt, Dante Certification Level 2 (2nd Edition) is a prerequisite for the new Level 3 class.

Q: What market or applications are being targeted with this project?

The way people phrased this question, they usually think Audinate started this project.  In the webinar, we tried to make clear this was Amit’s project - we were just a technology provider.

From Amit’s perspective, he has many ideas.  The initial goals may be to bring in guest artists from other regions for live performance or recording studio use.  It could even be a pre-production tool, allowing remote artists to participate in rehearsals even if their schedule requires them to be in another state.  There are many more, but for now we’ll stick with that.

From Audinate’s perspective, this project is more about stretching the technology in new ways - it is a learning experience.  We often say Dante is not just a protocol or some static solution.  We are constantly improving it.  Projects like this help us test the system with an application that hasn’t been tried before.  

In this project, I don’t think Audinate saw anything unexpected from a networking perspective.  But it was really great to see the system meet this need so well.  And, going through the process certainly helped us prioritize a few features on our product roadmap.  

Video

Q: Do we have to use Dante AV cameras, or are there gateway devices to other formats?

Using a Dante AV native product can make for a clean installation and simplified control.  One CAT6 cable provides Dante AV, camera control and 802.3bt PoE power.  However, there are devices to get traditional AV devices on and off the Dante AV network.

The Patton FPX6000T (transmitter) and FPX6000D (decoder) are HDMI encoders and decoders for just this purpose.  The Bolin D10H decoder also works this way - you can mix encoders and decoders across brands, no problem.  

All encoder and decoder units are 4Kp60 capable.  If you already have cameras or have a brand preference, these units can bring them into the Dante AV fold, and the encoders and decoders from different brands can be mixed together without a problem.

The Bolin D220 camera we used is the 1080p60 model.  Since we were only looking to produce at 1080p30, that was all we needed.  However, if we wanted to run this at 4Kp60 and keep everything Dante AV native, the Bolin D412 is a 4Kp60 model.

All of these devices can operate off of 802.3bt PoE power.  (On this event, many did.)  This minimizes the cable clutter.

Q: Did the PTZ controller only affect local cameras, or all?  What protocol?

Bolin supports VISCA over IP or serial links.  We used VISCA over IP to control all Dante AV cameras on all locations from the single PTZ controller. It was able to traverse the Layer 3 network without issue..  

If you need to integrate with existing cameras or controllers that support VISCA or VISCA over IP, it is likely you can blend these Bolin cameras and/or controllers with your existing system.  So, you could use this controller to also control your existing cameras, or you could use your existing controller to control the Bolin Dante AV PTZ cameras.

Q: What other products are coming to the Dante AV platform?

Clearly, we have to leave it to our Dante AV licensees to choose the timing of their product announcements.  But yes, there are more devices coming.  I would bet many of the things people asked for are being worked on, as well as some things you haven’t even thought of, yet.  Stay tuned!

Q: What video frame rates were used?

Our original goal was to run at 1080p30 - the front cameras always ran that.  Part way through, we decided to try the side views (for musicians to see each other) at 720p60.  720p60 and 1080p30 are about the same bandwidth, so that meant we didn’t have to revisit bandwidth budgets.  This change was about shaving latency a bit more, just to see if musicians noticed anything.

Ultimately, the musicians didn’t feel a latency difference.  The musicians commented that the timing with audio was absolutely solid and the video links helped with communication.  Some of them watched the screens quite a bit!  We certainly felt the video link would be appropriate for use with an audience.

Q: Which displays have the least video lag?

Many people wanted our research on video displays.  We aren’t trying to be coy, it isn’t in deliverable form and products change frequently.  The useful part for you is the concept.

Amit mentioned in his research, commercial displays generally have the worst video lag performance.  One might assume designers of commercial displays are trying to bedazzle people with the best possible image - display lag is not their priority or possibly even a consideration.

For our event, we used consumer TV models that offered a "game" mode.  Effectively, this particular mode bypasses much of the internal video processing (sacrificing some picture quality) in favor of shorter display lag.  After all, if someone playing a video game sees the image (and hears cues) 70msec faster, then that is like improving their reflexes by 70msec, right?  The same principle applies to our use case.

We don’t recall the specific model number, but we know it was already discontinued when the events came around.  Consumer TVs tend to refresh 1-2 times a year, so this is not unusual.  Further, game mode is not present in every consumer TV, nor is it necessarily in every model of a TV line, nor is it necessarily consistent through product refreshes of the same product line.

When you need to narrow down some models for this use case, it might be good to think like a gamer.  Narrow down potential models by going through tech influencers on YouTube.  When they review displays they will usually mention (or even test) display lag.

Q: Can you recommend measurement tools for input lag?

In the engineering space, there are plenty of measurement tools for video.  But, for simpler, down-and-dirty testers, there are two models from Leo Bodnar Electronics that are interesting for field engineers.  

Here is a 1080p model that has been popular in the gamer community: https://www.leobodnar.com/shop/?main_page=product_info&products_id=212

Here is a YouTube influencer using the device: https://youtu.be/6UEavnn4ERA?t=134

There is also a newer 4K model:  http://www.leobodnar.com/shop/?main_page=product_info&products_id=317

According to the manufacturer, if you wanted to force a test at a particular resolution, you can choose that in the configuration software.

Q: Curious about the use of vMix - wouldn’t that add significant latency?  

While VMix was used on this project, I admit they were not a “partner”.  They didn’t turn us down, we just never had to ask them for help.  Their system was used because it was already in-place (or available).  So, I would not feel qualified to discuss the latency their system does or does not produce.

However, I can say the musicians were not watching the output of vMix.  The vMix stream was intended for a remote audience.  We did push the image to the projection screens on either side of the stage, just so the techs could keep track of the final output.  That was almost impossible to see from the stage, so it was not going to affect the performers.  

In general, those side monitors did feel like it might pass for an in-house audience.  If we fire up the PA in the venue, the further the audience is from the stage the more video catches up with audio.  After all, light is faster than sound.

Telco

Q: Did you use dark fiber or shared fiber?

If we did this with dark fiber, it would require a connection from each location to each other location.  At three locations, that means each location would have two connections.  At 5 locations, each location needs 4 connections.  This doesn’t scale well.  Using shared fiber means the system could scale better, only requiring one connection per location.

Q: Was the network layer 2 vs 3?

Since Dante Domain Manager allows a Dante system to manage everything with Layer 3, we went that way.  It simplified the design, avoiding encapsulation or other tricks.  The more efficient the design, the lower the latency.

Q: Was QoS used in the Telco link?

No.  QoS was a costly capability from the telco and it actually caused more problems than it solved.  So, we dropped it.  Besides demonstrating the point that Dante can traverse complicated networks with ease, this translates to lower costs for the telco service and its administration.

Q: What about Latency, Jitter and Packet Loss?

Shared fiber providers offer a performance guarantee on latency and jitter specifications.  Packet loss was not an issue, either.

Q: Have you tried Dante with DSL or cable modems and public internet?  

Actually, yes.  Dante can certainly make a connection across the internet.  However, different types of internet connections will have different latency performance.  In the webinar, I mentioned we made a connection from our offices in Sydney, Australia and Portland, Oregon.  That used business class internet service, and required 135msec.  That isn’t useful for musical collaboration, but it was better than I expected for such a long trip on a generic resource.

Internet connections are often sold based on bandwidth capabilities, not latency.  When Amit spoke to telcos about how fast their network was, they always discussed bandwidth, right?.  

To put this in perspective, suppose you needed to get an envelope from Los Angeles to New York.  Let’s suppose you ask a shipper how long it will take them to deliver the letter and they respond, “Oh, we can use trucks up to 52’ long.  We can carry letters, don’t worry.”  See the problem?  You’re asking about time for the journey (latency), they are discussing raw shipping capacity (bandwidth).

Now, in the ISP’s defense, they usually sell based on bandwidth because that is what the typical customer understands.  They don’t perceive the difference between 15 and 50msec.  However, with shipping, we are accustomed to considering shipping time.  We understand the difference between overnight and 7-10 business days.  

The more we simplify the network, the better the performance becomes.  We achieve higher sustained bandwidth, lower latency and lower jitter.  All are important.

The best performance would be dark fiber.  As Amit explained, if you only need to connect two locations, this point-to-point solution will deliver the best results.  Shared fiber options expand your reach to many locations, and will be a close second.  

Frankly, shared fiber is very much like shipping.  When you drop a package to FedEx.  They don’t hand-off the package to UPS, DHL or USPS for some leg of that journey.  FedEx manages delivery from any point in their shipping network, right to the doorstep of the destination.  

Likewise, with shared fiber, the same telco manages the data the whole distance.  That means they have total control of the network path, and they can then confidently make certain performance guarantees that become reasonable for distance performance.

Public internet services offer the cheapest connection and the widest reach, but it has the most volatility in performance and delays.  This problem is independent of the service you run across it - the network itself is just more complex and therefore takes longer.

Dante can cross the public internet for a professional link with longer, conversational latency.  It could be used for a producer to monitor a recording session at another studio.  The link would be pristine quality.  Public internet links might be appropriate for music lessons, where students don’t have that refined timing, yet or for shorter distances that can stay within musical timing.  

However, for musical interaction at a professional level over any real distance and for long-term, reliable uptime, the performance of public internet is likely too volatile.  The cost of shared fiber can be a bargain, compared to the time lost in maintenance of a generic internet link.

Sorry for the long explanation, but I hope that makes some sense.  

Bottom line, different connection types have different advantages and disadvantages.  You need to balance those options for your use.  As Cynthia said, you can also delegate that investigation to folks who have done it before.  Design consultants like Amit can add tremendous value and experience in solving these issues.

Q: Was a VPN used?  What about Encryption?

On our shared fiber link, it was not necessary to establish a VPN.  We felt the telco’s private network was enough.  If someone was to get past that, Dante Domain Manager was implementing a layer of access security.  Someone couldn’t just open Dante Controller and see our system - they would need log-in credentials.  We felt that was enough security for our use.

On our Sydney to Portland link, that did operate through a VPN.  It wasn’t required generically, but that doubled as a test for another customer project wanting to try a distance link in a particular VPN solution.  Because of Non-Disclosure Agreements, we can’t say more than that.

Audinate is investigating options for data encryption, but this is a totally different tool that could be used in combination with access security that Dante Domain Manager offers.  However, at this time, we have nothing to announce on encryption.

General Latency Capabilities

Q: How did you achieve 6.5msec or 14msec of latency when Dante only supports 5msec?  

Different Dante devices have different amounts of buffer memory.  Some devices can go as high as 40msec.  

If you are only using Dante Controller, we know you are on a Layer 2 network.  In that case, we only show you options up to 5msec.

When Dante Domain Manager gets involved, we anticipate you may be on a larger network, and we enable longer latency settings up to 40msec.

It is worth noting that our system used latency options between the offerings in Dante controller.  For instance, we normally have steps at 10msec and 20msec.  This application introduced a desire for more options, and that has always been possible under the hood.  So, we made that adjustment and were on our way.

Q: What is the maximum latency Dante can manage?

While this webinar was about a low latency application, Dante can tackle much longer latencies as well.  Various Dante interfaces will have different amounts of memory, but the Brooklyn II (our most popular chipset) has buffer memory for up to 40msec - though that isn’t the latency limit, either.

Effectively, the buffer only needs to capture the jitter of your network (and any safety margin you wish to apply beyond the requirement.)  So, it will allow you to capture the 40msec leading up to playout.  

This is how we made a connection from Sydney, Australia to Portland, Oregon.  Latency on that with business class internet was about 135msec.  Network packet arrival times easily stayed within 120msec-135msec.