Optical fiber is my preference but it can be interrupted. Last year I had no connexion for about 6 weeks and now a wildfire has destroyed the line and nobody knows when it will be repaired. This is the reason why I have Starlink as a backup. While the bandwidth can be quite impressive, via satellite there is considerable jitter. A higher buffer size can compensate for that in exchange for more latency. To keep latency in check I tried a higher sampling rate (96 kHz) and here is what I get recording of a jazz lick in my JT studio London
When the main aim is to have fun making music with others the audio quality is almost good enough and the latency can be tolerated.
Thanks very much Thomas! This is an interesting report on using Starlink for JackTrip sessions from your studio in London. Please keep us posted on further findings.
finally I got an opportunity to try Starlink with someone else. In the video the trumpet is on a copper line and I connect to the internet via satellite (Starlink). We met in a studio set at 96 kHz and the audio quality slider was manually set to 26 (whatever that means) as the minimum acceptable audio level.
The backing track was on the server. Syncing issues are most obvious with video but the audio turned out ok. There is certainly room for improvement through additional rehearsing and agreeing on a a certain type of articulation for unison playing. Most importantly the experience was enjoyable. Have a look Starlink track on server 96kHz Belgium - Recording 2025-08-12 | JackTrip Labs
This is of course good news for all those who do not have access to optical fiber.
As many recent audio interfaces can handle even higher rates such as 192 kHz, it might be interesting to have this possibility also at studio level. With higher sampling rates one can increase the buffer size to compensate for jitter with poorer connections.
wow yes, with additional attention to the musical considerations you mentioned this is really encouraging usage of Starlink for JackTrip sessions. Bravo! Thank you Thomas for keeping us up to date with your endeavors with JackTrip.
Setting any audio quality differences aside and focusing solely on the latency side of things, 96 Khz with 128 buffer size is going to give the same results as 48 Khz with 64 buffer size. They will both create the same number of audio packets (slices of time) per second.
Higher packets per second also means that you computer is performing more audio→digital and digital→audio conversions per second, which means it demands more performance from both your audio interface and computer. The latest interfaces and computers can handle settings that are more aggressive than our defaults, but many cannot.
Mike, you are right and your reasoning could be pushed further by setting the latency aside. Latency is much less of an issue as previously thought. Many musical genres accommodate significant amounts of latency and there is no reason why this should be any different online. What can be be quite annoying are sonic artefacts like gargle noise and distortion which can be mitigated with bigger buffers.
The idea is to use higher sampling rates combined with bigger buffers to cope with poor internet connections. Modern equipment can even handle higher rates than 96 kHz but the JackTrip Studios currently do not allow for higher than 96 kHz. If we could get 192 kHz, I suppose one could achieve results with Starlink or DSL connections that come closer to what we have with optical fiber with 48 kHz.
What are your thoughts on that, Mike?
It’s an interesting topic, for sure! JackTrip’s Audio Quality slider has the biggest control over latency and buffer sizes. You should be able to use that to eliminate any sonic artifacts regardless of sample rate. If not, it warrants further investigation and may be a bug.
Higher sample rates (or lower audio buffer sizes, i.e. 128 vs 64) reduces the slice of time represented by each audio packet transmitted over the network. If that were the only parameter involved, one would think raising sample rate would cause more sonic artifacts. But in JackTrip, it’s a bit more complicated because of how we handle buffering..
JackTrip’s buffering (with the newer loss concealment “Regulator” strategy) is now more time based versus packet based. It automatically sets a minimum baseline (in ms) for each person based on the observed jitter present in their network connection, and then lets you control an additional time-based buffer based on your group’s tolerance for sonic artifacts. That is what the Audio Quality slider is for. The number you set it to is literally a number of milliseconds of extra buffer added on top of that baseline for each person who is connected.
Having worked with thousands of musicians over the past 5 years, we’ve found that everyone feels very differently about this. Some are happy to tolerate a little bit of the sonic artifacts if it provides better latency, and others want a perfectly clear sound. JackTrip is perhaps the only thing out there that gives you one simple control to express this preference, and handles everything else for you. And it’s perhaps the only thing that’s using advanced machine learning to mask the artifacts to the point of being inaudible, allowing you to achieve lower true latency for the same level of quality.
I think the impact of using smaller time slices per packet on all of this is interesting because it results in smaller gaps in the lost audio. Let’s assume your connection loses 2% of packets at latency X with buffer size Y, but 4% at latency X with buffer size Y/2 (2x the number of packets per second). Since the lost packets are smaller in duration, the audible loss could be the same, or even better. We’d be filling in more gaps but the predictions may produce a waveform that is closer to the truth. There may even be a “sweet spot” where loss tolerance provides minimal audible loss, and lowering it further doesn’t really change much. This seems like it would warrant further research..
Absolutely, we need more research! I do not expect that any time soon everybody will have fiber. If higher sampling rates together with bigger buffers can indeed compensate for poorer connections that would be good news for a number of my friends.
Just to reiterate: moving the “Audio Quality” slider to the right alone should already compensate for poorer connections. If you find that is not the case, I’d like to investigate and understand why.
The audio quality slider does its job in improving audio quality. At value 15 audio quality is acceptable without excessive latency at 96 kHz. At 48 kHz the latency is higher.