Internet latency problems

[SteveW]

Hello Hans

Yes, the latency (using a Behringer UM2 with an Apple MacBook) is 11.5ms. I don't know if anyone with a Mac is getting as low as 2-4, because the CoreAudio has some inherent latency. Certainly another band member is getting around 7ms with a Focusrite interface on a MacBook. That would be a worthwhile latency improvement at a reasonable price. I'll have a good look through the discussion thread on recommended interfaces to see what else definitely works and which gives a good cost/latency compromise.

Regards

Steve

[Zlartibartfast]

My Focusrite Saffire is connected via Firewire, which by nature has lower latency than USB 2. In JamKazam's Latency Loopback test it returns 3.14ms ; in a session I see 7~8ms reported.

Since the May 29 JK client update, I've seen my total latency averaging 30ms when in session with my regular jam partner.

Getting Better All The Time....

[SteveW]

Hi Zlartibartfast

That sounds good, although my MacBook hasn't got a Firewire connector--only Thunderbolts. One of my jamming partners is using a Fosusrite Scarlett with a Mac and getting similar 7-8ms in-session latency via a Thunderbolt/USB dongle, so that may be the way to go.

But this is all way better than the the latency I was getting a couple of weeks ago, when it was varying wildly and often going over 100ms. Like you, I have to admit it's getting better...

[KennyO]

Thanks for the reply, Stuart.

I have tried changing the frame size to 1ms but the audio just breaks up. 2ms is the lowest I can get with the Behringer interface. I am not using ASIO drivers as it is an Apple MacBook, so just using the default Apple drivers, which are not as highly tuned as the ASIO drivers would be (though I'm not sure the UM2 interface has got any specific ASIO drivers, anyway). I am currently using a sample rate of 44.1kHz, which was chosen because that's what other band members are using--I think there should be a speed advantage in all using the same sample rate. Maybe we could improve the latency a little by all using a higher rate. But I imagine that is going to shave off a few milliseconds, rather than cure the big problem of latency fluctuating from 20 to over 100ms.

Very good point about the ping tests. I had forgotten about ICMP.

I have tried Test.Net, as you suggest. This tells me my download speed is 16.5Mbit/s and upload 1.1Mbit/s. These are well below the quoted speeds of my ISP, so I think I am being short-changed by them at the moment. Will give them a call today.

We could give a click track a try. We have tried it in the past for other reasons and didn't get on very well with it, but I guess its a matter of practice and just getting used to it. Also we are using the lockdown to learn some new material, including building band arrangements of songs written by one of the band members, who is a prolific writer. Previously we found that the best results with a click track were when we knew the material very well and could really concentrate on the tightness of our performance. I think it would be least effective where we are trying out new ideas, improvising parts, trying different tempos, etc.

I am not sure my understanding of JamKazam is correct, but I am assuming that, if I set my outgoing bitrate to 256kbit/s, this is the rate for each audio channel, and that I have a separate audio channel for guitar and chat mic to each of the other participants in the session. (Is this correct?) If so, I would need an internet upstream bandwidth of 2 x 3 x 256 = 1536kbit/s. Add to this the overheads of TCP/IP and I am well over the capacity of my upstream connection. So, if this calculation is right, what I have seen as latency may just be the saturation of my internet connection. That would be consistent with what I have experienced: lowering my bitrate made an improvement and, in fact, the band's most successful practice was a day when, for reasons that remain a mystery, I could not get the instrument channel to work on JamKazam and ended up using the chat mic to pick up the guitar sound from from a combo amp, rather than having a dedicated guitar channel. So, on that occasion, I was only setting up a single mono channel to each of the band members.

Will report back after next session with the band.

Steve

I think the outgoing but rate encompasses the sum total of all your audio inputs. But your 1.1mb uplink speed is marginal but workable I think. I forgot that you can't use ASIO on the Mac since it uses Core Audio. But if you can adjust the recording frequency you should play with that to get the lowest latency you can. You and your session partners don't have to use the same. Playing with a drum machine or click track is mandatory in our experience in this environment. It's also good musical practice even though most of us would never do it when we were all in a room playing together.

Stuart.

Hi Steve,

I'm not sure you've got the math quite right. I'm not sure I've got it right either but here's what I think...

First of all if you take a look at Manage>Audio Booster in the very top left corner you will find the actual bit rate including overhead. For example with the Maximum Outgoing Music bitrate set to 256Kbps, the outgoing Audio Bit Rate is 377.6 Kbps (Music Audio 256.0 + Overhead 121.6) assuming you are not using chat or video.

Both your input audio channels (instrument & mic) are mixed down to one stream - not 2 separate streams.

Since it's peer to peer you are sending a separate audio stream to each of the other 3 members, and receiving an audio stream from each of them as well.

So the way I figure it you are using 377.6 x 3 = 1.13Mbps upload and download.

Your download speed of 37Mbps is not a problem. But if your upload speed was 1.8 or 1.1Mbps you were cutting it pretty close, especially if there was other traffic from your home network. It may explain why you saw an improvement  when your provider set you to 3.1Mbps.

---

Hi Zlartibartfast

That sounds good, although my MacBook hasn't got a Firewire connector--only Thunderbolts. One of my jamming partners is using a Fosusrite Scarlett with a Mac and getting similar 7-8ms in-session latency via a Thunderbolt/USB dongle, so that may be the way to go.

But this is all way better than the the latency I was getting a couple of weeks ago, when it was varying wildly and often going over 100ms. Like you, I have to admit it's getting better...

Steve - when I first started on JK I was using a Mackie 8 channel mixer with a USB output as my interface to a Mac. The audio interface latency was something like 11-12ms. I have fibre right up to my desk @65/11MBPS so a pretty good connection to the Internet.

The Mackie worked ok when I was playing with local musicians located nearby (within 100km).  I was able to jam with people further away, but it depended on them having low audio interface latency and high speed internet. Otherwise latency-ency-ency.

I bought a used Thunderbolt interface (Resident Audio T4) and the audio interface latency is now 2ms, which is a big drop. So yes - Thunderbolt is the way to go but there are limited products and they can be fairly expensive. The Thunderbolt interface just gives me a reasonable chance to jam with folks a lot further away.

Cheers

[SteveW]

Hi Kenny,
Yes a 2ms latency would be great but Thunderbolt interfaces are all a bit pricey. I mostly just play music and never really do any high quality recording, so the only motivation for getting a better interface is for JamKazam use. So the balance of cost and latency is all-important.

Cheers
Steve

[Zlartibartfast]

For any Mac users reading this thread, and still shopping for an interface, it's worth noting that the thunderbolt 2 -> FW 800 adapter works great (except in my case, it did not work under Catalina). If you find a good deal on a firewire interface but don't have a firewire port, the adapter will get you up and running.