This review page is supported in part by the sponsors whose ad banners are displayed below

Interview
Q. Let’s jump right into some of your design philosophy and choices. You use a transformer for your balanced inputs. Did you consider a fully differential circuit?


A. [Michel] It was a possibility but I soon rejected it. From a common mode rejection perspective it's almost impossible to build a perfect electronically balanced circuit. Maximum immunity is around 50dB. The only way to exceed 50dB is to use opamps which are not acceptable for audio. You have to use ultra-precision resistors of .001% tolerance which are very expensive and perhaps good for a metallurgy lab but not audio. The problem is that you cannot have an exact replica of two circuits – it is almost impossible. Some companies have done it with good results but for me and Tenor this is not the way to go. And as to size, yes you could do it but to maintain quality we would need additional chassis, crossing over to impracticality even by Tenor standards.


I chose a transformer because of this and total galvanic insulation between output and input, eliminating potential ground loops and acting as free interference filter. Because the transformer has limited bandwidth, it won't pass RF. With the transformer you can get 120dB of common mode rejection. The trade-off is that with any transformer you get a specific type of distortion because you are transferring signal via magnetism and coils. You must choose a transformer with the best possible specifications. Our input transformers are custom-made for Tenor and we had to specify a very precise list of characteristics. The transformer is also externally compensated to not create ringing with a very flat response and no resonance. We use a physically large transformer designed to avoid any type of audible or measurable distortion. The distortion level for the transformer at 20Hz is .005% which is extraordinarily low compared to other transformers. And the bandwidth is wide extending past one megahertz. Also phase is within 2° at 20kHz.


Q. You mentioned bandwidth. You specify -3dB at 500kHz. Was that a design consideration for minimizing phase shift using the 10 times rule?


A. It was a consideration to have very wide bandwidth for phase and speed. If you go with lower bandwidth you'll end up with a very slow circuit. I believe that the rise time for the signal is about 1µs. If you go over that you start to slow down and you feel and sense that distortion at high frequencies. Bandwidth flat from 0 to 100kHz is perfect. What is important after that is that the roll-off be very very slow. Steep cut-offs will create the same phenomenon as found with anti-aliasing filters in CD players. If you go with a brickwall filter, you will get phase problems, ringing etc.


Q. Both your preamp and amplifiers are very wide bandwidth. Are you concerned about picking up interference?


A. No. First we use transformers in balanced mode. In unbalanced mode, yes bandwidth is there but interference is eliminated by how the physical grounding is laid out. We use a multi star system. We make little star points in multiple sections and then connect to a central star point so no loop can form inside. We also use interference suppressors like ferrite beads to reject the RF.


Q. In some ultra-wideband components I've heard, there was a touch of brightness, maybe a little bit of hardness that is difficult to pin down – almost a feeling. How do you address this?

A. Some circuits can oscillate at extremely high frequencies, meaning hundreds of megahertz. When we designed our circuit, we looked at an RF spectral analyzer to ensure that there is no oscillation anywhere inside our very wide bandwidth. There is no way that parasitic oscillation can interfere with the signal. If one was detected it would be suppressed as part of the design. And if it wasn’t possible to suppress it, we would reject the circuit and do something else. You won't see this with normal measurements. I was doing a circuit several years ago and got oscillations at 55MHz and 450MHz. This is difficult to detect with normal instrumentation. We had to get an ultra-wide spectrum analyzer. These oscillations must be killed or they will pollute the signal forever. They produce very odd and weird defects in the signal where you almost sense something is wrong but again it's very hard to detect with test gear.


Q. Let's talk about the microprocessor control. How is digital grunge isolated from the audio path?


A. The microprocessor board has its own power supply and is completely isolated from the audio board. There is also a separate power supply for all of the matrix relay switching within the system. We chose a microprocessor that is physically a very small chip – about one square centimeter. Everything is small and surface mount technology. The microprocessor board itself was designed with interference in mind. It’s a four-layer board. All the interference that can be created by interconnection is between two ground planes. Additionally we have a large amount of filtering devices around the microprocessor on the logic board that makes it very quiet and transmits virtually no noise. Also a full battery backup system powers the processor. Therefore even if some parasitic distortion were to be introduced, it will be blocked. The logic board power supply essentially charges the battery and the battery runs the processor. The screen is a vacuum fluorescent display. We initially had some interference which we then blocked with filters so no interference is transferred to the audio board. The display is mounted in the heavily shielded logic board that also shields interference from the display and both are physically separate from the audio board for total isolation.


Q. You have paid specific attention to grounding, including the flexibility to adjust grounds in the left and right channels individually.


A. The audio ground is generally connected physically to the chassis which becomes the  mechanical ground. In my opinion that's the wrong thing to do because then it’s connected to the wall. This is very bad. The audio ground has to be separated from the electrical ground of the house due to the potential for parasitic currents flowing from the chassis into the audio signal. This also creates the infamous ground loop problems. To address this we decided first to separate the grounds totally from the left and right channels so they are not connected. In the case of balanced cable the ground can be lifted. Again this ground should not be connected directly to the chassis. With the Tenor control feature, the audio ground can be connected to the chassis ground completely, partially or not at all. In partial mode the audio ground connects to chassis ground with a low value resistor. If you have a small ground loop, the potential will be dissipated in the resistor. Large amounts of hum can be treated by lifting the ground completely, either in one channel or both. This makes the Tenor applicable in a wide variety of systems that have problems with ground noise or cables with insufficient shielding. For systems without a problem we recommend the resistor and normal setting.


Q.  Many manufacturers use a large torrid transformer; however Tenor uses several smaller ones. What is your philosophy on that?


A. The power supply section uses separate power supplies for each channel in one chassis. For each channel there is a low-voltage and high-voltage transformer. The low voltage takes care of the filaments, the relay power supply and the stepper motor power supply. Two separate power transformers for each channel are designed to eliminate any interference one to the other. Contrary to other designs we do not use toroids. Yes there is a little less magnetic stray field coming from a toroid but the problem with them is wide bandwidth. Everything from the wall is dumped into the circuit. Another problem is their inherent magnetization curve. Toroids are very sensitive to DC voltages. The transformer will block DC voltage but will also saturate. And that generates a large amount of harmonics which pollute the power supply.


Tenor uses a conventional custom EI transformer. The inherent design of the toroid allows parasitic high-frequency currents present in your home power supply to enter the circuit. The laminated design of an EI transformer has much narrower bandwidth with higher parasitic capacitance and diminished bandwidth. The windings are made to narrow the possible bandwidth. The toroid has a low magnetic stray field which is very good, the EI has more. One way then is to diminish the magnetization of the primary winding by the addition of a magnetic shield. Look at the graphs. Although a toroid is more efficient, the design of the Tenor transformer allows it to be driven to its limits without any saturation. Plus additional electrostatic filters in the design of the transformer allow it to be spectacularly clean. Again these transformers are designed exclusively for us. If after all of this some interference still exists, the transformer is enclosed in steel, which as a magnetic material will short the remainder of the interference. The net result is that the stray magnetic field is virtually immeasurable. Finally the transformers are mounted on a steel plate which is isolated by rubber grommets to remove any possibility of vibration transmission to the chassis.

Q. After transformers we get to rectification and filtering…


A. Most of the power supply is conventional except for the high voltage which must be extremely clean. For the high voltage we use an ultra-fast F.R.E.D diode (fast recovery epitaxial diode) which generates very low switching residues with very smooth recovery generating much fewer harmonics and less  noise. Then additional filtering is applied. We then go into a 4th-order filter with two inductors and two banks of capacitors. A very smooth very narrow bandwidth waveform is presented to the first regulation stage that adds additional electronic filtering. After power is sent from the power module to the preamp module, the power is regulated a second time inside the preamp. For maximal efficiency the regulator should be in close proximity to the circuit it powers. This also allows temperature control. The second regulation is for the high voltage, the 500V for the tubes and the ±45V for the buffer stage. This also cleans up the power if there's any possible interference picked up in the cable between power supply and control unit. This regulator circuit developed by Tenor has no sonic signature. As with the audio circuit, the regulator circuit does not use feedback, which would create the same kind of problems. In reality you are listening to your power supply which is modulated by the amplification system. If the power supply is not perfect you will not have good sound.


Q. What is next for Tenor?


A. [Jim] We have a few things in the works. We will be introducing our phono preamp around year’s end. The design and sound will be based on the design of the Line1/Power1. Also Michel has some fascinating concepts for a state-of-the-art DAC. We have not made a decision on that one but it could be very interesting…
Enlarge!