(Modification B: 2 inputs 6 outputs,
24bit 192kHz record/playback) 

Does ultimate perfection exist? It's unlikely if we are speaking about sound quality. Any production audio system is always a compromise between production costs and sound quality. And if it is a mass product, a manufacturer has a stronger temptation to save on the prime cost as here every cent spent brings thousands of dollars. And considering that it is capital, rather than the genius of engineers, that rules the industry, the products of the mainstream market follow the principle that a good advertising campaign is the most important thing. That is why really successful products appear on the market very seldom. And that is why we shouldn't expect superb sound from mass products.

Now let's turn to professional sound cards. All mass products within the range from $200 to $1000, except crippled and old cards, have approximately equal sound quality as they use the same components. They can differ in the number of inputs/outputs and various additional niceties like external units with connectors and mic preamps. If you are looking for a card with the extremely good converters and minimal other capabilities it won't be simple to do. No chances to find such thing among the mass products as well.

Today we are testing the Lynx Two sound card from the American company of Lynx Studio Technology. The distinguishing features of this solution are the reference quality of sound and an a high price ($1000 to $1200 depending on the modification).

The Lynx Two came replacing the Lynx One card which supported playback at the maximum of 24 bit 44.1 kHz. The first model had no unnecessary frills and was considered one of the best solutions in its price category for a long time (sub $600). By the way, the Lynx One is still selling and being supported by the manufacturers.

• model A: 4 inputs / 4 outputs; 
• model B: 2 inputs / 6 outputs; 
• model C: 6 inputs / 2 outputs. 

And even with such real characteristics Lynx Studio Technology does have something to boast of:

Analog-in parameters
24 bit 44.1 kHz mode

Frequency Response
20 - 20 kHz, +- 0.05 dB
Dynamic Range 117 dB A
Signal-to-Noise 116 dB A
Channel Crosstalk
<-120 dB, 1 kHz signal @ -1dBFS
THD+N
-108 dB (0.0004%) @ -1 dBFS
-104 dB (0.0006%) @ -8 dBFS
1 kHz signal, 22Hz - 22kHz BW 

Analog-out parameters
24 bit 44.1 kHz mode

Frequency Response
20 - 20 kHz, +- 0.05 dB
Dynamic Range 117 dB A
Signal-to-Noise 117 dB A
Channel Crosstalk
<-120 dB, 1 kHz signal @ -1dBFS
THD+N
-97 dB (0.0014%) @ -1 dBFS
-104 dB (0.0006%) @ -8 dBFS
1kHz signal, 22Hz - 22kHz BW

The data are given for the frequency of 44.1 kHz as the data obtained in this mode look best of all. The lower the sampling frequency in modern converters using oversampling and filtering, the greater part can be filtered out. But the difference between the modes with different frequencies is not great here.

The specs of the converters give better characteristics: 123 dB A for ADC and 120 dB A for DAC. In our tests we obtained 115 dB A and 113 dB A (for SNR and DR) when connected the input to the output which coincide with the characteristics they gave (total characteristics of the I/O section are always a bit worse than separate characteristics of an input and an output).

When we tested the Waveterminal 192X card (with the same ADC chip as the Lynx Two has) which boasted of 123 dB in its specs (i.e. the figure is just taken from the codec's specs) with the help of the Lynx Two we got much more modest scores: DR = 107 dB A and SNR = 105 dB A. There was no wonder even when we established balanced connected using a short 20cm mic cable (in case of the unbalanced connection the scores would be lower by 1 dB) and a high signal level of +4 dBu. But remember that this card is 4 times cheaper than the other.

The card supports both balanced and unbalanced connection of analog interfaces. But what are both modes for in a professional card? Unbalanced connection should be used if cables are quite short and there are no problems with common ground, for example, in a small audio studio based on a computer. In this case sound quality can be better than even in the balanced mode since it lacks an additional operation of subtraction of signals of opposite polarity. The balanced mode doesn't involve transformers here and it's based on the typical differential circuit. The card doesn't support manual selection of the mode for the analog interfaces, but in the documentation you can find different methods of soldering the connectors for different ways of coupling of balanced and unbalanced devices.

The card incorporates stereo DACs and stereo ADCs of the highest brand (Crystal CS4396 and AKM AK5394). Reportedly the ADC is priced at $30 wholesale, and 3 stereo DACs are available at $60. So, $90 are spent only for the converters. 6-channel DACs used for home audio systems are usually not dearer than $10, and AC'97 codecs used for multimedia cards are priced at about $1.5.

The Lynx Two sports a reprogrammable controller which can fix bugs, improve internal processing algorithms and add support of new extension cards released for Lynx Two (there are several such cards already). One of the firmware updates has the ASIO support improved and the error in the hardware dithering algorithm in channel mixing corrected.

Here you can select independently a reference level of signals for inputs and outputs (+4 dBu / -10 dBV). The digital interface is not neglected either - all flags of S/PDIF and AES/EBU DIF are displayed. The problem of stability of a master clock and synchronization is of utmost importance for professional equipment. That is why the Lynx Two can indicate a frequency of the master clock and supports synchronization for different sources including a special unit of synchronization with a video signal.

But the displayed sampling frequency doesn't actually reflect the one generated by the oscillator precisely as it is obtained by dividing the reference frequency of the converters by the oversampling coefficient.

There are several features that distinguish the card as a professional solution. For example, there are several SRC modes (sample rate conversion).

Or you can select a record dither type: no dithering, triangular, triangular with noise shaping, rectangular. As you know, dithering is adding of pseudo noise of the minimal power to the signal to prevent correlation of the signal with quantization noise. The dither types take their names from a shape of a random distribution frequency curve. Dithering is used only for recording because a reproduced signal does contain one of the dither types.

The mixer certainly has a level indication marked with decibels and a clipping indication. But it's impossible to enable a limiter, though it's justified given to the conception of noninterference into the signal.

Testing in RMAA

The tests in the RMAA 4.2 give a very clear idea that the Lynx Two scores the best results among all professional sound cards we've ever tested before. And while in the 16bit mode most cards have reached the format's upper limit, in the 24bit mode the Lynx Two leaves the other audio devices far behind.
 

General performance: Excellent (for all the tests)

Remember that these tests are synthetic. These figures can't determine the sound quality entirely, but the test results agree with the listening tests to a very great extent.

Conclusion

Some time ago it was assumed that high-quality sound could not be obtained at all on a computer system because of magnetic pickups from a video card and a processor into the sound card's circuit of the printed-circuit board, terrible power supply from a pulse power supply unit, jitter in converters, inadmissible usage of transfer electrolytic capacitors, low-quality connectors, cheap components of the analog section.

But if the approach is chosen properly it's quite possible to get much better sound from a more expensive card than from a cheaper solution on a computer.

Well, modern sound cards keep on improving their sound quality against all the odds.