Updated: February 17, 1999

The following tests were performed with Pulsar v1.1.

The Physical Connections

I connected Pulsar to an Alesis ADAT with both analog I/O and optical I/O.

The Pulsar Project

In each test step, I set up the Pulsar project to connect

The Test Procedure

In short:

In depth:

I created a very simple single-oscillator Pulsar Modular patch to generate a pure sine tone. I recorded this tone optically onto track one of the ADAT (configuring a Pulsar project which routes the Modular output to the Adat Dest #1). I recorded five short pulses of a middle-C. This way, I'd be able to observe the latency on five different instances to check the consistency of my work.

Then, I performed a series of tests which passed the audio from ADAT track one, thru the Pulsar and back out to record onto ADAT track two. I did this in various combinations of analog and digital, as you will see in the results chart.

Then, I recorded both ADAT tracks simultaneously back into Sound Forge, via the Pulsar optical connection. (Configure a Pulsar project with Adat Source 1-2 connected to the Wave Dest module, and then just tell Sound Forge to create a new 48k Stereo file, recording from the Pulsar In).

Now, in Sound Forge, I had a stereo file with which I could observe the time difference between the left and right channels (left being the original ADAT track one, and right being the processed ADAT track two).

Sound Forge allows you to zoom in on a waveform to the point that one screen pixel is equivalent to one sample. At this zoom level, I can simply count how many samples there are between the start of the left audio and the start of the right audio, and that is the total latency of that particular recording scenario. To count the number of samples, I click on the beginning of the left track's audio, and drag a selection which ends at the beginning of the right track's audio. In Sound Forge, the size of the selection is displayed in the window status bar, so Sound Forge does the counting for me. Note that this counting isn't always necessarily 100% accurate, as I'm only working with what I can see, and using the mouse isn't always perfect. Sometimes the Sound Forge doesn't display a signal if its amplitude is too low, and if I try to compensate by increasing the amplitude zoom (or by normalizing), it makes slopes steeper and harder to treat well...

The Chart

Explanations of the tests are below.

On the whole, I'd say Creamware's "Zero-latency" claim is pretty well founded. To see what happens when you are mixing multiple channels, see the results for the ADAT All-digital multi-channel mix test

A good rule of thumb is that if the latency is less than the latency you'd get by moving your head in the soundfield (or just turning your head), then it's ok! If you move your head by one foot in the soundfield, you'll get a change of about 1 ms (speed of sound is roughly 1300 feet/sec, so if you move your head by just 1.3 feet, you get a 1 ms change).

The All Analog Non-Pulsar Reference Bounce Test

In this test, I just physically patched (no Pulsar here) the ADAT analog output one, through my Tascam mixer, back to the ADAT analog input two, and recorded the test. This is to determine the latency of a direct bounce on the ADAT. The results are expected to be zero, as the only latency is introduced by the speed of the ADAT's own digital-to-analog and analog-to-digital converters. If you're a real geek, you can add in the latency of the electrical signal travelling at the speed of light along about 2 meters of Hosa cable and some circuit traces inside my Tascam mixer, which comes out to, let's say 3 meters at 3x10^8 m/sec = .00000001 seconds but this is so much less than the sampling frequency of the ADAT (48k or once every .02ms), that we can ignore it (duh!)

This test resulted in consistent latencies, as one would expect, of absolute zero. I had formerly reported here somewhat inconsistent non-zero results, which I thought must have been attributable to inconsistent D/A's in my old (blackface) ADAT. It turned out in fact to be the result of a forgotten compressor patched into my mixer's first channel, which even though in total bypass mode, was introducing some latency. After repeating the test, bypassing the mixer entirely, (and then repeating the test again through the mixer, but without the comrpessor insert), the latencies were consistently zero.

ADAT Analog Thru Pulsar to ADAT Digital

In this test, I patched the Pulsar Analog Source left to the ADAT Digital Dest #2. After recording the test, and bringing the stereo pair back into Sound Forge, I found the latencies to be 23, 36, 35, 37, 38. So, an average 0.7 ms latency.

ADAT All Digital Thru Pulsar

In this test I patched the Pulsar ADAT Source #1 directly to the ADAT Dest #2. This is just a bounce, passing through the Pulsar. In this case, I found the latencies, to be a rock solid, consistent 9 samples, or 0.2 ms.

ADAT All Digital Thru Pulsar With Mono 4-Pole Filter Insert

Same as previous test, but I inserted a 4-pole filter between the ADAT Source and Dest. This time, the latencies were consistently about 15 samples, or 0.3 ms.

ADAT Analog Thru Pulsar to ADAT Analog

In this test, I patched Pulsar Analog Source #1 over to Pulsar Analog Dest #2. This, therefore goes through both the ADAT's and the Pulsar's A/D's and D/A's. In this case, the latencies were 83, 62, 51, 49 and 49. So, an average 1.2 ms.

ADAT Analog Thru Pulsar to ADAT Analog with 4-Pole Filter Insert

Same as previous test, but I inserted a 4-pole filter between the Pulsar Analog in and out. This time, the latencies were an average 1.5 ms.

ADAT All-digital multi-channel mix

In this test, I recorded the pure sine tone simultaneously to six ADAT tracks. Then I wired the optical outs of ADAT 1-6 to Bix Mixer inputs, and the mix outputs of the Big Mixer over to ADAT optical inputs 7-8. This way, I could test mixing 6 channels to stereo, while adding various effects in the Big Mixer. Then to check latencies, I could optically pull ADAT tracks 6and 7 back into Sound Forge to compare initial start times.

Incidentally, I received precisely the same results with fewer channels (even one channel), so I'm assuming that adding channels will also result in consistent latencies. Excellent!

The only question is, since the latencies aren't fixed then there is a possibility of two halves of a stereo pair becoming out of phase with respect to each other. I'm adding a request for the ability to "phase-lock" a stereo pair in the big mixer, to ensure phase consistency on stereo pairs.

Compared to the Yamaha 02R

The Yamaha 02R is a relatively expensive (about US$7000) digital mixing console. The following Yamaha 02R latencies were reported by jhno in rec.audio.pro, as told to him by Steve Gunning at Yamaha.

According to Yamaha, the 02R latencies are fixed as follows:

From any analog input to the stereo analog output or aux analog output:
2.2ms when the console is operating at a 44.1K clock frequency, 2.0ms at 48K

From any analog input to digital stereo output OR 8-bus digital output:
0.75ms at 44.1K, 0.7ms at 48K

From any digital input to the stereo analog output:
1.8ms at 44K, 1.7ms at 48K

You can see that the Pulsar compares quite favorably to the Yamaha numbers.