Daniel Weiss of Weiss Digital Audio, whose Minerva Firewire DAC has a devoted following among audiophiles with high-end computer-based music servers, has agreed to an exclusive interview with the Ultra High-End Audio and Home Theater Forum to discuss his company and its cutting edge digital products.
Thank you for agreeing to be interviewed for the Ultra High-End Audio and Home Theater Forum and welcome. Weiss is only beginning to become more widely-known the U.S high-end arena. Please tell us about you background and the history of the company.
I founded Weiss Engineering Ltd. In 1985, after I was employed with the famous Studer/Revox company for 5 years. At Studer I worked in the PCM Lab on an asynchronous sampling rate converter and on digital tape recorders. Before Studer I got a BSEE in Switzerland. Back then at Weiss we concentrated on the pro audio market. With the advent of the CD there was a huge demand for digital signal processing equipment for CD mastering. So we ended up designing and manufacturing a modular processing system capable of doing equalization, dynamics processing, sample rate conversion, interfacing, mixing etc. Today we still are active in pro audio with the Mastering Studios being our main clientele. We have an excellent reputation in the mastering industry worldwide.
In 2001 we started our activities in High-End Audio. With the very demanding Mastering Engineer clientele we thought we could do just right for audiophiles. So we started of with a top quality D/A Converter, the Medea, which we still sell.
Although perhaps your most well know product in the U.S. is the Weiss Minerva Firewire DAC, about which we will have more to say in a moment, let’s focus first on some of your other high-end components, starting with the Weiss Castor monoblocks power amplifier. I understand that is a high-wattage Class-D design. What exactly is a Class D design, and what are its advantages and drawbacks, if any.
There are many different approaches to a class D amplifier. It is not like e.g. a Class A design where once you know it is class A, you know about what you can expect in terms of performance. Class D is still an emerging technology. The main idea is to pulse-width modulate a high frequency square wave. Imagine a square wave with a 50% duty-cycle, i.e. the wave is at positive voltage the same duration as it is at the negative voltage. The average voltage of such a square wave is zero. Now if we modulate the pulse width of the square wave the average voltage changes. So if we take the audio signal to modulate the pulse width, the average voltage is that audio signal. The advantage is that the power transistors have only two operating points, i.e. on or off, for generating the square wave. Thus power loss in the transistors is minimized. That is why class D amplifiers come with very moderate heatsinks even at high output power capacity. One drawback is that the square wave has to be filtered out before the signal hits the speakers. The tweeter may not like a several hundred kilohertz full power signal. Usually a high power LC (coil/capacitor) filter is used to get rid of the square wave and leave the average (audio) voltage. Of course such an output filter brings up questions regarding the output impedance (damping factor) and / or the feedback topology of the amplifier.
The modulator for the square wave can work fully analog, with an analog input, but can also be a one with a digital input, working in the digital domain exclusively. The digital modulator has some inherent drawbacks because of delays involved.
Distortion figures of class D amplifiers are getting better and better, they may not be at Class A quality level yet, but that it is simply a question of time. The power wasting Class A amps are kind of an anachronism these days.
One unique feature of the Castor power amplifier is its 12-position input gain switch. Under what circumstances would you select anything other than maximum gain, and how do you determine the optimum level of gain for your system.
This of course is a difficult topic. Proper levels throughout the system are important in order to optimize the system’s performance. It does not make sense to crank up the level at one point (and likely introduce more distortion because of that) and then bring down the level again with the preamp volume setting. I guess the best approach is to have the power amp sensitivity set such that the preamp level control works in the upper range, with maximum level resulting in a sound pressure level you just may want to listen to occasionally. Decent preamplifiers have gain trims at their inputs for alignment of the different sources. With such gain trims and with a gain trim at the power amp things get more complicated. In such cases one has to check the specs of the equipment involved and make sure the levels are exercising the equipment according to the specs, i.e., not too low (noise) and not too high (distortion).
Weiss also offers its own line of Chiron cables including AES/EBU digital and analog interconnects a glass optical digital cable, and speaker cables. Neither the digital nor analog interconnects are shielded. Why did you elect to design your cables without shielding? Isn’t shielding necessary for rejection of EMI and RFI interference and for rejection of hum over long runs?
We have chosen to use non-shielded cables in order to optimize for low capacitance. Wherever possible I recommend using balanced cables, where shielding is not that important. Our cables have a geometry with tight tolerances, i.e. the influence of common mode signals (hum, etc.) on a balanced cable is minimized. We also do unbalanced cables, unshielded. That works fine if the source impedance of the connected equipment is low and the cable is not too long. Of course it can’t be used for e.g. a turntable pick-up. The best for digital transmission is the optical cable of course. No interference at all.
Would you describe the unique features of the Weiss Jason CD transport?
It can do all you expect from a CD transport, plus:[/b]
- It can upsample the signal to 88.2 or 176.4. This allows the user to choose for his favourite sonics depending on the D/A converter connected.
- It has a volume control in the digital domain. One which is implemented properly, i.e. it employs dithering in order to avoid quantization distortion. On our website I have published a paper on this very topic, including listening examples.
- It allows setting the output wordlength to 16, 20 or 24 bits depending on the D/A converter connected. This quantizer is also dithered.
- It has a feature which we call DAC++. This is useful for certain D/A converters which may have less than ideal jitter specs. DAC++, when engaged, generates a wideband jitter on the output signals in order to cover up discrete jitter frequencies in less than stellar DACs.
- It has a plethora of outputs (XLR, RCA, ST optical, Toslink) including dual wire AES/EBU.
What are the advantages of a top-loading mechanism like the one used on the Jason transport over drawer and slot-loading mechanisms?
We have chosen the top loading variant because it was simpler to design, mechanically.
What special efforts have you made with the Jason transport to eliminate jitter?
There is a local crystal oscillator to generate all sampling rates etc. including the main clock for the CD transport module. This keeps jitter very low. In addition there are separate regulators for different power rails in order to minimize crosstalk from the transport servo / motor to the crystal oscillator.
Would you explain your upsampling methodology? In addition, why is upsampling implemented at the transport stage rather than in the DAC?
Usually the DACs upsample as well, so with the upsampling in the Jason the user has the choice to have upsampling (or part of it) happening in the transport and the rest in the DAC. It is a matter of the user’s taste and of the upsampler quality in the DAC.
We have done sampling rate conversion for more than 20 years now. We use a linear phase polyphase filter for interpolation. Upsampling always is a compromise among several parameters, like filter ripple, pre-echo, stopband attenuation, transition bandwidth etc. Finding the sonically right compromise is the challenge.
I noted that upsampling is not available on the Toslink or S/PDIF outputs of the Jason transport. Why is that the case? Perhaps this would be a good time to explain the difference between the AES/EBU and S/PDIF digital interface as well. I understand that the difference extends beyond simply cable topology and cable resistance.
Actually there are two RCA outputs, one is fixed at 44.1, the other allows for upsampling. The Toslink is fixed at 44.1. Toslink should not be used at higher rates – actually it should not be used at all if possible. Jitter performance is very bad.
AES/EBU and S/PDIF are different in cable and signal levels / impedance as you mention. Plus the Channel Status (CS) data bits are different. CS bits are additional information embedded in the datastream. Information on, for example, sampling rate, emphasis on/off, audio channel usage, wordlength information, etc. The receiving end, e.g. a DAC, can take the information of importance out of the CS bits. E.g. a DAC could check whether the signal is a dual wire signal or a single wire and treat the signal accordingly. Or it could check the sampling rate and switch its clocks accordingly. Of course it can also measure the sampling rate of the signal and do not care about what is signalled in the CS data. Or it can switch on the de-emphasis filter if the signal is flagged emphasized in the CS data. One bit in the CS data also tells whether it is an AES/EBU or a S/PDIF link. The S/PDIF type CS data does not support many of the bits available in the AES/EBU format. So actually all the outputs on the Jason transport carry an AES/EBU signal when it comes to the CS data.
Would you describe the unique features of the Weiss Medea digital-to-analog converter?
With the Medea we tried to accomplish the ideal D/A converter, i.e., not sensitive to jitter, high signal to noise ratio, all common sampling rates supported, low distortion, low output impedance.
“Not sensitive to jitter” is one thing often ignored in other converters. That is why all the discussions are going on with jitter in transports, cables etc. Jitter matters at the stages where analog signals become time discrete (A/D Converter) or where time discrete signals become continuous again (D/A converter). And at those stages the designers have to take care that jitter has (almost) no influence. So the Medea employs a very slow (sub hertz range) phase locked loop. That means that the Medea clock is almost running independently from the input clock. This in order to isolate the incoming jitter from the D/A converter sampling clock.
The it has a discrete Class A output stage with virtually a zero Ohm output impedance. This to be able to drive any cable or subsequent input stage. The range of output level is huge, it can drive a power amplifier directly. Sampling rates between 44.1 and 192 kHz are supported, via single or double wire schemes.
With 44.1 or 48 kHz input signal the Medea engages a separate Weiss upsampler algorithm running on a DSP chip. This upsamples by a factor of 2. Going from 44.1 or 48 up is the most critical step in upsampling. Going from 88.2 or 96 up is much less demanding.
Would you describe the benefits of upsampling? Are there any circumstances under which you would choose not to upsample? When upsampling a standard Redbook CD, is there any reason to upsample only to 88.2kHz when upsampling to 176.4kHz is available?
That all depends on the D/A converter. Most modern D/A converters do upsample because they are delta-sigma noise shaping designs. So if an external upsampler is employed, part or all of the upsampling going on in the DAC chip is bypassed. Upsampling on DAC chips often is compromised because of limited signal processing resources on the chip. Hence it may be of advantage to have a separate upsampler process.
As alluded to earlier, with the increasing interest in the computer playback of high resolution audio files, the Weiss Minerva Firewire DAC is currently perhaps your best known product in the U.S. Would you describe its functions and features?
The Minerva can do several “conversions”:
- Firewire to analog
- Firewire to AES/EBU on XLR, RCA
- AES/EBU on XLR, RCA, Toslink to analog
- AES/EBU on XLR, RCA, Toslink to Firewire
In addition to being a D/A converter it can also take an AES/EBU signal from e.g. an A/D converter and feed that to the computer via Firewire for recording. E.g. it can work as an interface for an A/D converter to the computer for recording records with dedicated computer recording software. Or it can interface a computer via Firewire to any other DAC.
The Minerva handles sampling rates up to 192 kHz on all connectors, it can also do dual wire AES/EBU. The analog output stage uses a minimalistic approach with as few parts as possible between D/A chip and output.
A very nice feature is that the Minerva can be master clock when in Firewire mode, i.e. the computer is slaved to the Minerva.
For dejittering when the Minerva is slaved to an external source (e.g. via the AES/EBU input) a dual PLL approach is used, which results in a very high jitter attenuation. The output level can be set in 4 coarse steps (analog domain) to accommodate for the input sensitivity of the amplifier. In addition there is a digital level control feature which gives 0.5dB steps over a more than 120 dB range.
Another nice feature is the “insert mode” which allows inserting any digital device (e.g. a digital EQ) via the XLR or RCA AES/EBU connectors. E.g. one could do a chain like computer — Firewire to Minerva — external EQ via AES/EBU — back to Minerva to analog output.
For computer playback, what are the advantages of a Firewire connection over USB, coaxial and Toslink S/PDIF, and AES/EBU connections?
One main difference is that Firewire is a bus structure, while the other protocols mentioned are point to point connections. Firewire has several advantages, like:
- Our implementation uses the so called Isochronous mode of the Firewire link, which means that a defined bandwidth is allocated for a given connection via the bus. E.g. if on the same bus a hard disc is connected, it can not happen that the hard disk grabs all the bandwidth and the audio link chokes because of that. The audio link has its bandwidth reserved without any exception.
- The Firewire is very easy on the host computer CPU power, as most of the work is done by the Firewire interface in the computer.
- Several devices can be connected to a single Firewire bus. This allows building multi-channel systems with two-channel devices. That is why the Minerva has two Firewire connectors, i.e., to allow for daisy chaining devices.
- As mentioned above, the Minerva can be master clock to the computer, which makes it easier to maintain excellent jitter performance.
What are the differences, other than the faceplate and cost, between the Weiss Minerva and the Weiss DAC2, your professional Firewire digital to analog converter?
There isn’t any difference besides the faceplate and the feet. The price difference comes from differences in the distribution chains.
Weiss also offers the Weiss Vesta, a Firewire to digital interface. Does the Vesta offer the same features as the Minerva with the exception of digital to analog conversion?
Yes, that is right.
Would you describe the application and features of your professional Saracon sample rate conversion software? Does Saracon have any uses outside the professional arena? For example, would there be any benefit to converting the audio files on your computer from 16-bit/44.1kHZ to 24-bit/176.4kHz using Saracon rather than letting the Weiss Minerva handle the conversion. Do you still offer Saracon-Light for conversion of two-channel audio files only?
There are two versions of Saracon: Saracon and Saracon-DSD. The difference is that Saracon-DSD can also do DSD conversion, i.e., it can do PCM to DSD and DSD to PCM conversion. Note, it can not convert SACD discs, as the DSD stream on SACDs is encrypted for copy protection. The DSD version is mainly for studio use.
Saracon can be useful if the D/A converter at hand sounds better with a higher sampling rate. Today most of the D/A converters are of the delta-sigma variant, which means they do upsample the signal before conversion. Upsampling (or oversampling, which is the same) is not a trivial task. The most demanding stage in upsampling is the one going up from 44.1kHz or 48kHz. Going up from 88.2kHz or 96kHz is less demanding. The upsampling in the DAC is often times done in the DAC chip itself. And often that upsampling is (fairly) compromised in terms of performance. So doing a decent upsampling outside of the DAC can help in terms of sonic quality. Saracon can do that. The drawback of upsampling the files on the hard drive is that they take up more space after upsampling.
Best is to test whether the DAC at hand performs better at higher sampling rates. And if that is the case then check whether upsampling the audio files could be a viable option considering the required space on the hard disk.
In closing, is there anything else you would like to tell us about your company and its products?
As mentioned at the beginning, we have a range of pro audio products besides the high-end audio line. Some of the pro audio products can be interesting for the audiophile. Like e.g. the EQ1 equalizer family, the ADC2 A/D Converter and the already mentioned DAC2 D/A Converter and the Saracon sampling rate converter. Check them at http://www.weiss.ch via the link at the left hand audio side.
One other thing – on our high-end website there are the manuals of our products. They contain some basic information on various audio topics, like dithering, jitter, output stages etc. If you are interested give it a reading. http://www.weiss-highend.com.
Thank you for joining us today.
Thank you too for the opportunity to present our views and products.
