Linn Knekt Kivor hard-disk multizone music system Measurements

Sidebar 3: Measurements

To assess the Linn Oktal's performance, I ripped a CD-R containing all of my regular 16-bit CD-player test signals. The Tunboks drove the Oktal via a 1m Cat.5 AES/EBU link. I also fed the Oktal 24-bit digital data via one of its S/PDIF data inputs. Unfortunately, the processor would not lock to 96kHz-sampled data, though it did to an 88.2kHz datastream, after a few burps and hiccups.

At 1.942V, the maximum analog output level was inconsequentially below the specified 2V RMS, this sourced from a consistent 211 ohms across the audioband. The Oktal preserved correct absolute polarity. The frequency response (fig.1, bottom pair of traces above 1kHz) was flat within the audioband for both 44.1kHz and 88.2kHz sample rates, with a very slight top-octave rolloff continuing above the audioband at the higher sample rate. However, pre-emphasized data (fig.1, upper traces) were not decoded with the appropriate de-emphasis, presumably because the Tunboks does not preserve the relevant datastream subcode flag. Fortunately, pre-emphasized CDs are extremely rare these days. The Oktal did recognize HDCD data, however. Channel separation (not shown) was excellent below 600Hz, being buried in the noise floor, but deteriorated at 6dB/octave above that frequency to a still-good 80dB at 20kHz, due to the usual capacitive coupling.

Fig.1 Linn Kivor Oktal, frequency response at -12dBFS, without emphasis, at 44.1kHz and 88.2kHz sample rates (top); and with emphasis, at 44.1kHz (bottom). (Right channel dashed, 0.5dB/vertical div.)

Fig.2 shows 1/3-octave spectral analyses of the Oktal's analog output while it decoded 16- and 24-bit data representing a dithered 1kHz tone at -90dBFS. The increase in bit depth results in only a 2dB drop in the noise floor, suggesting that the Oktal really offers 16-bit performance. Which is fine, considering that the Tunboks can send it only 16-bit data. But it does mean that the processor will not get the best from sources with greater resolution. Slight spuriae can be seen at 200Hz, 2kHz, and 7.5kHz. I'm reasonably confident that the two lower-frequency spikes are due to the Audio Precision System One test set, but the 7.5kHz peak appears to be characteristic of the Oktal. This peak can also be seen in the spectral analyses of the Oktal's output while it decoded "digital black" (fig.3). Note the dramatic increase in noise above 20kHz, due to the noise-shaping used to maximize the DAC's audioband resolution.

Fig.2 Linn Kivor Oktal, 1/3-octave spectrum of dithered 1kHz tone at -90dBFS, with noise and spuriae (from top to bottom): 16-bit data, 24-bit data (right channel dashed).

Fig.3 Linn Kivor Oktal, 1/3-octave spectrum of "digital black," with noise and spuriae (from top to bottom below 20kHz): 16-bit data, 24-bit data (right channel dashed).

Linearity error (fig.4) was very low down to -100dBFS, below which noise made its presence known. This noise can also be seen in the Oktal's reproduction of an undithered 1kHz sinewave at -90.31dBFS (fig.5), where it obscures what should be three discrete voltage levels. As expected from fig.2, no improvement in the waveshape could be seen when the word length was increased to 24 bits.

Fig.4 Linn Kivor Oktal, departure from linearity, 16-bit CD data (right channel dashed, 2dB/vertical div.).

Fig.5 Linn Kivor Oktal, waveform of undithered 1kHz sinewave at -90.31dBFS, 16-bit data.

8787 Perimeter Park Boulevard
Jacksonville, FL 32216
(904) 645-5242

MikeMaine's picture

Or you can buy a Mac

CuteStudio's picture

... that you can run the SeeDeClip4 multiuser music server on a regular, noisy PC in the spare room and access and/or control the music using any modern gadget like a Chromebook, tablet, iPad etc.

This makes the choice of client easy - there's lots of cheap alternatives and an iPad can be hooked up to Toslink using an Apple TV or Airport Express etc.

The free version does a lot more than you'd think, it's a complete home audio solution.

JonGreen's picture

A little late to the table(!), but thanks for an excellent, well-balanced review.

I was the systems architect of the Imerge SoundServer, which was rebadged (with some enhancements) as Linn's Kivor. I also designed the XiVA-Link communications protocol, and worked with Linn's Alan Clark (designer of the iconic Sondek CD12) on the S/PDIF hardware and drivers: Alan did most of the hardware work; I assisted in some of the FPGA firmware, and write the drivers.

I can confirm the accuracy of just about everything reported here. During 2000, both SoundServer and Kivor were going through a series of rapid evolutions. Towards the end of 2001, the products were settling down.

I'm a little surprised that they were reported as being MP3-only, though. One of the key selling points for audiophiles was that both products were able to rip and play uncompressed audio. This is why SoundServer (and, I believe, Kivor) came with up to 1.1 TB of storage - a massive amount at the time - configured as eleven 100 GB drives. It ran hot and heavy (and, yes, a bit noisy), but had enough elbow room to accommodate a lot of raw audio.

It was true that we only had one genre allocated to a track or album. This was partly because of the limited information we received from Gracenote. I always felt that having more than one genre per item in the database would be a good thing, but I was over-ruled. Apart from anything else, it would have made genre-based searches substantially slower, for a bunch of technical reasons it's not worth going into here. I think that if we'd done it today, we'd have used a noSQL database such as MongoDB or Couchbase, so we could have had the flexibility to enhance with additional fields such as user-assigned genres or arbitrary tags.

Anyway, thanks again. Great memories, revisiting that part of my career!