Universal Audio Apollo Audio Interface + UAD 2 DSP

[Anderton]

Originally Posted by JCSoundSystem

Also, could you possibly take a screenshot of the available ins/outs that are presented within a DAW? Thank you very much

Let's look at how Apollo interacts with your Mac and with various programs. We'll start at the most basic level, Audio/MIDI setup. Then we'll look at other programs like MOTU Digital Performer, Ableton Live, and Apple Logic.

When you go to the Audio window for Audio/MIDI Setup, here's what you see when you click the Input tab. I figured this also provided a good excuse to show the various sample rates that Apollo supports. (And just out of curiosity, does anyone reading this use 176.4 or 192kHz sample rates? Just wondering...)

Note that you can see all the inputs listed, but the volume controls have no effect here because of course, that's all handled in Apollo's Mixer application, which gives you far more control anyway.



This next screen shot shows what happens when you click the Output tab. This example shows the Clock Source menu, which can be S/PDIF, ADAT, Word Clock, or Internal.



Now let's look at what shows up in a DAW.

[Anderton]

Originally Posted by UniversalAudio

Apollo is, first and foremost, an interface for recording audio. For really nitty-gritty FOH guys who need a ton of I/O and a Pandora's box of flexibility, this might not be right tool.

I certainly agree, but for solo acts and duos, Apollo is pretty interesting. With Apollo and a laptop, you don't have to bring mic pres, external processors (except for a guitar multieffects so you can do good distortion - hint, hint), or mixer - just send the stereo outs to your PA.

The only potential drawback is the lack of tactile response with Apollo's mixer, as I don't see any way to use a control surface with the faders. For my act, that's not significant as the mixing action is done within Ableton Live, and Apollo would just be set and forget anyway. However for an act like a duet with a couple mics and direct ins, it would be inconvenient to do mousing around with a laptop while the crowd gets restless. So definitely, you need the right tool for the right job...but I think Apollo would indeed be the right tool for certain live performance jobs, especially as more and more people take laptops to a gig anyway.

[ParisMinzer]

Great review so far!

Quick question:

Let's say you *do* want to use an outboard piece of processing in the chain before you hit the A/D section.

Is it easy enough to hit the Mic Pre, then route it back out and into the virtual console?

[Anderton]

Just to be clear - so what you're asking is whether there's an ability to do something similar to the insert jacks on traditional console channel strips?

[Syncamorea]

Originally Posted by Anderton

(And just out of curiosity, does anyone reading this use 176.4 or 192kHz sample rates? Just wondering...)

I once recorded a short bit of drums at 192, just for the sake of doing it. I typically record at 44.1/24 although I am underway with some 88 and 96kHz comparisons because of all of the reports of improved plug-in performance at higher frequency and to try to come to my own conclusions about 44 vs. 96, etc.

I'm going to go back and check out the beginning of this thread - I can't remember if you mentioned what you are using as your 8 channel ADAT input. I'm anxiously awaiting delivery of a 4-710d that will be my ADAT in, fully realizing that, along with and Apollo, a 6176 and LA-610 in the rack, my rig will look like a UA advertisement. No, I'm not a UA homer, in fact, the rack has other pres and channel strips NOT made by UA!

EDIT: OK, I did a quick scan of the thread and did not see a mention of what you are using for your ADAT input. Was that install straightforward? I assume you are using the Apollo as the master clock?

[Anderton]

 

Originally Posted by Syncamorea

EDIT: OK, I did a quick scan of the thread and did not see a mention of what you are using for your ADAT input. Was that install straightforward? I assume you are using the Apollo as the master clock?

 

I'm trying to rustle up a piece of gear with an ADAT out to check this out. Stay tuned.

[Anderton]

Originally Posted by ParisMinzer

Great review so far!

Quick question:

Let's say you *do* want to use an outboard piece of processing in the chain before you hit the A/D section.

Is it easy enough to hit the Mic Pre, then route it back out and into the virtual console?

Assuming I understand your question correctly, there are no hardware insert jacks per se between the mic preamp and A/D input, nor can I find any way to route the mic preamp out separately before it gets converted to digital. However, there are two ways to accomplish what you want to do.

The first option is what we’ve been doing for years (and continue to do) with DAWs when using external gear, except you can do this within Apollo itself, and not have the same latency issues. Check out the screen shot...



The basic concept is to send the input signal to an output, then patch it back into another input. Here the mic is going into channel 1, which is feeding one of the aux buses. Note that the aux out can be assigned to any of the analog or digital outputs; in this case, analog outs 7+8. These patch into a stereo processor’s inputs, and the processor outputs patch into inputs 3+4. The meters in 3+4 reflect the signal going to the aux bus, then returning to those inputs.

The disadvantage is that the signal has to go through an extra stage of D/A and A/D, but given the sound quality of the converters, it’s highly doubtful that would be the limiting factor in any setup. An advantage of this approach is that you can apply UA plug-ins after the mic pre but before hitting the external processor, and also apply UA plug-ins after the processor but before the signal goes to the DAW. However unlike going through a DAW, in this scenario—even with the two added stages of UA plug-ins—the signal never actually goes through the computer. As a result he only delay is that caused by the conversion process; that depends on the sample rate, but figure roughly around 1ms. Naturally, this is dwarfed by any latencies within the computer itself.

The second option is to use a DAW’s “insert” plug-in, as found in many programs (the screen shot shows Logic’s I/O plug-in).



These basically facilitate the process described above, with the difference being that the mic signal gets converted to digital, goes into the DAW, then hits the DAW’s insert where the plug-in resides. Ths plug-in then sends the signal to an unused interface output that patches into the processor, and the processor’s output patches into an unused interface input. This has the same issue of adding an extra stage of D/A and A/D conversion, but with the additional issue of monitoring through the computer, which does cause latency. Note that most of these “insert” plug-ins can ping the “effects loop” and do latency compensation.

Getting real-world for a second, though, the odds are most external processors you’d want to add would be those for which UA doesn’t make an equivalent plug-in. My main use for external processors is to insert weird stomp boxes that have no UA equivalent, and often don’t have any kind of equivalent. With these, adding an extra stage of D/A and A/D conversion is inconsequential to say the least. Also note that if you have a digital effect with S/PDIF, you can do a “digital effects loop” and not have to concern yourself with additional conversion.

[Anderton]

Originally Posted by JCSoundSystem

Also, could you possibly take a screenshot of the available ins/outs that are presented within a DAW? Thank you very much

As promised, here's more info.

This shot shows how the inputs look to an audio track in Digital Performer. In this case I chose a mono track, so you can see each individual Apollo input. Of course, these are just numbers; they're not broken down into analog, ADAT, and S/PDIF. So you'll sometime need to watch the meters move to remember which Apollo channel correlates to the numbers in DP.



But wait - if there are only 18 ins, why is DP showing 28 input streams? Because DP can also pick up the signal going to the master output, auxes, headphones, etc.

Meanwhile, here's what shows up as possible outputs. This time to add variety, the screen shot shows the selection of stereo outputs instead of mono.

[jweisbin]

Originally Posted by Anderton

The basic concept is to send the input signal to an output, then patch it back into another input. Here the mic is going into channel 1, which is feeding one of the aux buses. Note that the aux out can be assigned to any of the analog or digital outputs; in this case, analog outs 7+8. These patch into a stereo processor’s inputs, and the processor outputs patch into inputs 3+4. The meters in 3+4 reflect the signal going to the aux bus, then returning to those inputs.

The disadvantage is that the signal has to go through an extra stage of D/A and A/D, but given the sound quality of the converters, it’s highly doubtful that would be the limiting factor in any setup. An advantage of this approach is that you can apply UA plug-ins after the mic pre but before hitting the external processor, and also apply UA plug-ins after the processor but before the signal goes to the DAW. However unlike going through a DAW, in this scenario—even with the two added stages of UA plug-ins—the signal never actually goes through the computer. As a result he only delay is that caused by the conversion process; that depends on the sample rate, but figure roughly around 1ms. Naturally, this is dwarfed by any latencies within the computer itself.

The second option is to use a DAW’s “insert” plug-in, as found in many programs (the screen shot shows Logic’s I/O plug-in).

Hey, this is a great article. (I've been following you since "Polyphony" magazine). I just came to this part so please forgive me if this is discussed later on in the article. I've had my Apollo quad for about 2 weeks now, and I'm loving it. This is from UA's FAQ:

"I understand that I can plug microphones, guitars, and line-level devices into Apollo and get low-latency performance, but what about using virtual instruments with UAD plug-ins in real time?"

For that workflow, we recommend using the ADAT or S/PDIF digital I/O as a “loop” (ADAT OUT to ADAT IN). For example, route your virtual instrument to an ADAT path and you will see it show up in the UAD Console application. Then, you can add UAD plug-ins on the Console ADAT input, and you will hear how latency becomes inaudible. You can also use analog I/O as a loop, but using ADAT or S/PDIF does not incur generational loss."

I tried it, and you can definitely hear the effect by sending it out to an ADAT channel with an effect setup on the console. Using Logic's "I/O" plugin on the track, I can send through the effect and bounce (but only in real time AFAIK). Of course, I lose the use of my ADAT's, so it is a pity that they only gave us 8 channels of ADAT and not 16 (in fact, when I ordered the unit, I thought it was going to have 16 because of the 4 light pipe connectors). I'm already using my S/PDIF to bring in the signal from my Audient Mico preamp, so I used the ADAT's (plus that way I get 8 channels to play with). There is some speculation that future firmware revisions might have some more flexible routing capabilities, which seems to be the most prevalent criticism of the unit I've seen. In particular, a loop back function like RME's Total Mix would be great.

[Anderton]

 

Originally Posted by jweisbin

Using Logic's "I/O" plugin on the track, I can send through the effect and bounce (but only in real time AFAIK).

 

As soon as you leave the computer, you pretty much have to do real-time bouncing. This also includes external FireWire DSP units like the TC PowerCore, SSL Duende, etc.

[Bill lorentzen]

I also am an engineer and do a lot of gigs, mostly singles and duos, so it's very intriguing for me to be able to have the great uad plugs in my live rig, even if it does mean another piece if gear to carry.
How much do I like my UAD plugs? I sold my prized 1176 LN after A/B ing them!
However I am extremely skeptical about working with any significant latency. Admittedly, none of my interfaces ever let me do much work below 4 ms (except Paris), but I can always hear it, especially for vocalists.
What exactly are the latency specs direct out of the unit, and have you used it yourself in a direct monitoring scenario with a few plugs inserted?
Thanks for this interesting review!

Addendum: I checked UA website and they quote 1.1ms with 4 serial plugs at 96k - presumably this would not change as you add channels until you reach system overhead limits. That seems like it would be unnoticeable in a live setup since I don't use in-ears. Not sure about vocals on the studio. Id have to hear it. Anyone got real world experience on this?

[meeble]

Any news/updates about the thunderbolt card availability for the Apollo?

[Phil O'Keefe]

 

Originally Posted by meeble

Any news/updates about the thunderbolt card availability for the Apollo?

 

As a new MacBook Pro owner, this is something that I'm interested in too.

[UniversalAudio]

Thunderbolt for Apollo is currently scheduled for release late summer.

[Anderton]

Originally Posted by Bill lorentzen

Addendum: I checked UA website and they quote 1.1ms with 4 serial plugs at 96k - presumably this would not change as you add channels until you reach system overhead limits. That seems like it would be unnoticeable in a live setup since I don't use in-ears. Not sure about vocals on the studio. Id have to hear it. Anyone got real world experience on this?

I'll talk about the live monitoring thing later, I just got back from traveling and I make it a rule never to do mastering, mixing, or any kind of critical listening with 24 hours of being on a plane.

Remember that even if you had the world's fastest computer, the process of A/D and D/A conversion introduces latency...I seem to recall something like 600 microseconds for A/D at 44.1, although I'm sure the Smart People at UA can give a more authoritative response.

Also remember that 1ms is how long it takes for sound to travel 1 foot, so in a typical situation with monitor speaker wedges, if you're 6' tall there's 6ms latency before you hear yourself even in a purely analog system. I see guitar players who stand 10 feet away from their amp, but then say that 5ms latency in an amp sim is "unacceptable" so I think some of the issue is psychological, too.

It's a little different for drummers, because of the precision with which good ones can hit notes. Drummers often can hear 5ms of latency but still cope with it, while 10 seconds drives them insane. I've certainly noticed issues when doing vocals and monitoring through computers in the past, although with fast enough computers, I don't find that to be a problem any more.

But, I'll give a fuller report either later tonight, or tomorrow. (PARIS really was a good system, wasn't it? Definitely ahead of its time.)

[Phil O'Keefe]

Originally Posted by UniversalAudio

Thunderbolt for Apollo is currently scheduled for release late summer.

Thank you.

Is the price still TBD?

Also, is it possible to cascade two Apollos on one computer?

[Anderton]

Originally Posted by Bill lorentzen

What exactly are the latency specs direct out of the unit, and have you used it yourself in a direct monitoring scenario with a few plugs inserted?

First of all, let me explain something about latency figures: They're estimates, and by nature, fictitious. For example, on my Roland Octa-Capture, a setting of 64 samples means that Roland can honestly and legitimately say there is a delay of 1.5ms caused by the sample buffers. In fact this is the ONLY spec they can give that's truthful, because the interface works in conjunction with other elements like additional hardware latency, conversion delays, and delays by going through USB itself (unlike Apollo, the Octa-Capture is a USB 2.0 device). These will vary depending on multiple factors.

However, many programs can also measure these ancillary delays, and display the ASIO reported latencies including buffer and hardware latencies. For example, Sonar lists an input latency of 6.0ms (265 samples), output latency of 3.4ms (152) samples, and therefore, a total round-trip latency of 9.5ms, or 417 samples. However, note that those figures are what ASIO is reporting, and that figure is not always accurate, either and there can be considerably variations among interface reporting accuracy. Ableton Live gives the same latencies as Sonar, but then again, they're also based on the same reported figures.

So, this is a roundabout way of saying that while you had mentioned that your interfaces don't let you work much below 4ms, that figure may be solely the delay caused by the sample buffers, and not all the other delays that need to be taken into account. I find it almost impossible to believe 4ms round-trip can be achieved, because given the other typical latencies, you'd need something like under 28 samples in your buffer. I've never met an interface that could go that low, let alone play audio with that kind of setting. If you're hearing a delay when you're seeing 4ms, I'd bet you're getting more latency than the 4ms would indicate.

You can prove this to yourself. Take a pair of stereo monitors, put one 4 feet behind the other, and turn up the volume somewhat on the further one to compensate for the distance. I'm certain you won't hear slapback echo in a room with decent acoustics (i.e., you're not hearing primarily reflections).

[Anderton]

Now, back to the subject at hand. I started off by singing into Apollo with headphones on, and noticed no latency. But hey - this is a pro review, so we go the extra mile and I figured out a methodology to measure the delay.

I plugged a mic into Apollo and headphones into the HP1 output. I then set up two mics. One was clamped inside the headphones so it was isolated from noise, and this became the "post Apollo" signal. The other mic was strapped to the mic feeding Apollo, so it picked up my original vocal sound. Therefore, the time difference between an output showing up on the straight mic vs. the post-Apollo mic would indicate the amount of latency through Apollo while monitoring. Bear in mind the computer was not involved in this, so there weren't any latencies involved in going through something like Digital Performer or Live.

Both mic outputs went into the aforementioned Octa-Capture, and I recorded the signals on individual tracks within Sonar, as it can give a readout of clip times down to microseconds or individual samples. This simplified calculating the differential between the time the different signals started. I labeled each clip so I'd remember what it represented.

The following screen shot shows the basic idea - you can see the clips (top track direct, lower track through Apollo), the difference in start times between the two clips, and the readout on the left that shows the length (in this case, 3.129ms). Although these views are zoomed way in, I zoomed in pretty close to max when taking the actual measurements.



Here are some of the figures I obtained, going from worse-case to best-case. Interestingly, it seemed that monitoring off the HP1 bus setting added another ms of latency or so compared to just using the Mon setting.

44.1kHz

4.5ms: No plug-ins, HP1 bus, no input compensation (which I believe would likely not be necessary when monitoring)
3.1ms: No plug-ins, Mon bus, no input compensation
3.6ms: LA1176 and Pultec Pro plug-ins active, Mon bus, no input compensation
3.7ms: LA1176 and Pultec Pro plug-ins active, Mon bus, medium input compensation

96kHz

96kHz gives less latency, as the audio moves through the system at a faster rate.

2.3ms: LA1176 and Pultec Pro plug-ins active, HP1 bus, no input compensation (so even at 96kHz, the HP1 bus seems to add a slight additional delay)
1.5ms: LA1176 and Pultec Pro plug-ins active, Mon bus, no input compensation

It goes without saying (but I'll say it anyway) that 1.5ms is a phenomenal spec. Remember, these represent total delay through all elements of the Apollo monitoring path - input A/D conversion, plug-ins if included, any calculations involved in the console, and output D/A conversion - so you can consider this true, worst-case "round trip" latency. To put it in perspective, if you're playing guitar and monitoring the output at 44.1kHz through the two plug-ins mentioned, it's equivalent to the delay caused by being about 3.5 feet from a speaker (which interestingly enough is pretty much the standard for near-field monitoring). FWIW, I've yet to meet a guitar player standing 3.5 in front of an amp and complaining that the delay in hearing the amp was intolerable

[Anderton]

Just for kicks, I decided to do a test to see what kind of delay would be noticeable when singing. So, I cleared out the signal path, inserted a DM-1 delay, and started dialing in different delays. At least for me, with vocals (which of course aren't percussive), I didn't start noticing any kind of delay until it went above 10ms; as the delay approached 20ms, it started to become objectionable. (All the delay figures quoted here are from the DM-1, so tack on another 2ms because I was monitoring at 96kHz).

With percussive "click" sounds (e.g., clave) I could notice latency above 5ms, but the sound was so tight as to be essentially meaningless. This became more obvious around 7-10ms.

Now, here's the bottom line. If you're monitoring through Apollo it seems that even with plug-ins, your latency is going to be well below what you would get by monitoring through even a very (very) fast computer. This is especially true when you take into account the fact that the computer will have more delay than the sample buffer figure would indicate. The story might be different with some of the really heavy-duty plug-ins like the Massive Passive, but I didn't test with those because realistically, while monitoring during tracking you're likely going to be using the "bread-and-butter" plug-ins to get a decent sound.

I presume that the only way to get lower latency would be to split the mic off into an analog mixer, feed your DAW outputs into the same mixer so you can hear the tracks, and use only analog effects inserted in the analog mixer because as soon as you inserted a digital effect, you'd have conversion latencies.

Standard caveat: I'm pretty sure this testing methodology makes sense, but if not, I'm sure one of the Big Brains at Universal Audio will set me straight

[Anderton]

 

Originally Posted by Phil O'Keefe

Is it possible to cascade two Apollos on one computer?

 

Yes, but I'm also going to see how it aggregates with a 4x4 MBox Pro running on USB 2.0. Should be interesting

[Bill lorentzen]

Craig, Thanks very much for your test results. It sounds like a proper method to me. I'm encouraged and excited because I'd love to run an LA2A,
Pultec, delay and EMT plate on my live vocals!
I should have said that my interfaces don't go below 4ms. I sure did not mean that they actually let me work at 4ms. The point is, by the time I'm ready to track vocals, I can't achieve an acceptable latency.
Just dreaming, but I'd love to hear a UA amp sim application (not Nigel). Their engineers could probably give the rest of the industry a run for it's money!

[NY2G]

Hey Craig,

I am a huge fan of your reviews and really appreciate the detail you are going into on the latency issue.

I have an MR816 and you did a great review of that unit as well. I am very familiar with that unit. What I am wondering is do units like the MR816, Motu, and RME have TRUE "0ms" latency? Is that what we are comparing here in going toward an Apollo? Or is that just market speak and only UA is being honest. Lol

Also are the latency figures different if you put fx on the aux bus as opposed to the direct channel? i.e. lower latency with reverb on an aux.

I am a vocalist and find that sometimes minute differences in latency aren't always easy to hear, but there is a vibe lost. For example, if I software monitor in Logic at a sample buffer of 32, it "feels" different to me than 64. Admittedly, I am very sensitive to latency in cans.

[Anderton]

Originally Posted by NY2G

Hey Craig,

I am a huge fan of your reviews and really appreciate the detail you are going into on the latency issue.

Or as I'm finding out...more like the lack thereof

I have an MR816 and you did a great review of that unit as well. I am very familiar with that unit. What I am wondering is do units like the MR816, Motu, and RME have TRUE "0ms" latency? Is that what we are comparing here in going toward an Apollo? Or is that just market speak and only UA is being honest. Lol

Good question. As soon as conversion to digital is involved, there will be latency. The only way to get true 0ms latency is with all analog gear. However, conversion latency is very low, and compared to the sample buffers in computers, appears subjectively to have no latency.

More companies are starting to call what they offer "near-zero latency" or "insignificant latency" so that's a good sign. But remember, even going through a digital floor box will add some latency. I just don't think it's all that important as a real-world issue.

Also are the latency figures different if you put fx on the aux bus as opposed to the direct channel? i.e. lower latency with reverb on an aux.

Hmmm....didn't think to try that. I suppose I should, now that I've taken apart my test setup

I am a vocalist and find that sometimes minute differences in latency aren't always easy to hear, but there is a vibe lost. For example, if I software monitor in Logic at a sample buffer of 32, it "feels" different to me than 64. Admittedly, I am very sensitive to latency in cans.

I know it matters to some people, but not to others. Not sure why. However, I spent many years playing on stage and still do from time to time, so I think I've been "trained" to accept, and compensate for, a few milliseconds of latency as it has always been a part of what I do...whether I want it to be or not

But really, what the specs show for the monitoring issue is that the amount of latency through Apollo is minuscule, especially at 96kHz, compared to what you could ever hope to achieve by monitoring through the effects in a computer. There is no way you'll get even close to 1.5ms of round-trip latency through any computer. Well, maybe one with 96-core plutonium processors and cryogenic cooling

[Anderton]

Hey! Just aggregated FireWire Apollo with a USB Avid Mbox Mini. Works great! If anyone wants details on how this works, let me know and I'll elaborate.

[furikuri]

Can you elaborate on live usage of the Apollo? Which plugs work, which don't?