30.3 Latency and Monitoring
The term Zero Latency Monitoring has been introduced by RME in 1998 for the DIGI96 series of audio cards. It stands for the ability to pass-through the computer's input signal at the inter- face directly to the output. Since then, the idea behind has become one of the most important features of modern hard disk recording. In the year 2000, RME published two ground-breaking Tech Infos on the topics Low Latency Background, which are still up-to-date: Monitoring, ZLM and ASIO, and Buffer and Latency Jitter, both found on the RME Driver CD and the RME web- site.
How much Zero is Zero?
From a technical view there is no zero. Even the analog pass-through is subject to phase er- rors, equalling a delay between input and output. However, delays below certain values can subjectively be claimed to be a zero-latency. This applies to analog routing and mixing, and in our opinion also to RME's Zero Latency Monitoring. The term describes the digital path of the audio data from the input of the interface to its output. The digital receiver of the HDSP MADI can't operate un-buffered, and together with TotalMix and the output via the transmitter, it causes a typical delay of 3 samples. At 44.1 kHz this equals about 68 µs (0.000068 s). In Dou- ble Speed mode, the delay doubles to 6 samples, for both ADAT and SPDIF.
Oversampling
While the delays of digital interfaces can be disregarded altogether, the analog inputs and out- puts do cause a significant delay. Modern converter chips operate with 64 or 128 times over- sampling plus digital filtering, in order to move the error-prone analog filters away from the au- dible frequency range as far as possible. This typically generates a delay of one millisecond. A playback and re-record of the same signal via DA and AD (loopback) then causes an offset of the newly recorded track of about 2 ms. The following table lists the delays of the HDSP MADI's DA-converter for the headphones output:
Sample frequency kHz | 44.1 | 48 | 88.2 | 96 |
DA (43.4 x 1/fs) ms | 0.98 | 0.9 | | |
| | | | |
DA (87.5 x 1/fs) ms | | | 0.99 | 0.91 |
| | | | |
DA (176.8 x 1/fs) ms | | | | |
| | | | |
Buffer Size (Latency)
Windows: This option found in the Settings dialog defines the size of the buffers for the audio data used in ASIO and GSIF (see chapter 13 and 14).
Mac OS X: The buffer size is defined within the application. Only some do not offer any setting. For example iTunes is fixed to 512 samples.
General: A setting of 64 samples at 44.1 kHz causes a latency of 1.5 ms, for record and play- back each. But when performing a digital loopback test no latency/offset can be detected. The reason is that the software naturally knows the size of the buffers, therefore is able to position the newly recorded data at a place equalling a latency-free system.
AD/DA Offset under ASIO and OS X: ASIO (Windows) and Core Audio (Mac OS X) allow for the signalling of an offset value to correct buffer independent delays, like AD- and DA-conversion or the Safety Buffer described below. An analog loopback test will then show no offset, because the application shifts the recorded data accordingly. Because in real world operation analog record and playback is unavoidable, the drivers include an offset value matching the HDSP MADI's converter delays.
Because the HDSP MADI is a completely digital interface, and the delays introduced by exter- nal AD/DA-converters or other digital interfaces are unknown to unit and driver, the drivers in- clude the digital offset values (3 / 6 samples). Therefore the delays caused by external convert- ers have to be taken care off in the record software, which usually means that the user has to enter specific offset values manually.
User's Guide HDSP MADI © RME | 71 |
