Spatial Audio Raven Preamp

@jc4659  I know Spatial runs the tubes at least 20 hours before they send things out, just to weed out any failures.   A 6SN7 is pretty well run in by 20 hours.  Yes, there are subtle changes in them out to 50-100 hours, but most of the sound is there in 20 or so.  What you will hear is the cathode bypass caps running in.   They are very large film caps and they are idling along at literally about 2% of their voltage rating.  So they take a while to settle down.  When they do you get that last 10-20%.   Where the real magic lives if the rest of your system can present it.

Unfortunately, opamps are limited in not being able to dissipate much heat due to the small package size. Very few opamps are designed to be used with heat sinks. So the only way to keep heat emission low is efficient Class AB output stages, relying on feedback to linearize the crossover region (opamps typically have very high feedback). Higher powered transistor and tube amps also use Class AB to keep heat emission to acceptable levels, at the expense of higher distortion in the Class AB transition region.

AB amplifiers have no problem with the crossover region and do not rely on feedback to sort this problem out. Its a matter of the output devices being properly biased rather than anything to do with feedback. Of course, feedback is helpful (if applied properly) to improve the overall linearity of an A or AB amplifier. As Norman Crowhurst indirectly pointed out, most of the time feedback is incorrectly applied.

@lynn_olson Since you are using an output transformer, there is an ideal load for which the transformer is optimized. As you know, transformers transform impedance and  this goes both ways, so if the load at the output is higher, the load on the output tubes is higher too. But transformers can ring if the load is too high and conversely, roll off if the load is too low. Traditionally, balanced line transformers are designed for low impedance operation (in the old days 600 Ohms, hence the dBm rating). Ampex provided a switch with a 600 Ohm resistor on the back of their tube tape electronics so if the electronics driving something with a high impedance input, the load could be provided to prevent the transformer ringing.

So I'm very curious how you handled this issue- did you design the transformer for a low impedance and simply installed a loading resistor?

I saw a comment from you about no servos in the context of noise which I assume was a typo since as you know, servos do not introduce noise. You might consider one, since your circuit is direct coupled from input to output before the transformer. Offsets (possibly at the input) can result in an sizable imbalance which introduces distortion, even if you run matched tubes. A servo can be easily used to correct this issue, allowing the output transformer do have its lowest distortion and they are inexpensive to implement.

lynn_olson

Unfortunately, opamps are limited in not being able to dissipate much heat due to the small package size.

That's why so many good balanced components use discrete circuits for the operational amplifiers.

I own a Geshelli DAC and their new balanced headphone amplifier. Although value-priced, both products use the Sparkos (made in Colorado!) discrete op-amps, with a very large Class A operating region. If you’re going solid-state, this is an attractive approach.

The Raven is kind of the opposite: Zero feedback, all-triode balanced, using some of the lowest distortion vacuum tubes ever made (6SN7). What makes it possible are the custom-designed input and output transformers ... before that, Don and I were limited to off-the-shelf transformers, which in turn limited the selection of vacuum tubes that would be compatible with the circuit. The original Raven, designed way back in 1998, used tubes in the 5687/7044 family. These are very good, but are not as linear as 6SN7 family, and the subjective difference is pretty noticeable.

The zero feedback approach brutally exposes the sonics of every single part in the signal path, right down to the sonics of the volume control. Our custom transformers have outstanding sonics (subjectively), and a flat response to 30 kHz in-circuit. The Raven has a signal path of: a discrete-resistor volume control, wire, a balanced 6SN7 dual triode, and a custom output transformer. The unbalanced RCA inputs have their own custom input transformer for phase-splitting and ground isolation.

Zero feedback vacuum tube circuits are not for everyone. If tubes make you nervous, I can recommend the Geshelli with Sparkos discrete op-amps (if the low price doesn’t offend you). If you are very price sensitive, there are lots of Shenzen-made products for $200 or less, with ESS converters and pretty decent op-amps.

It might sound odd, but both the Geshelli and the Raven have a similar parts cost to retail price structure. Solid-state isn’t really that expensive if you buy the parts in hundreds to thousands quantities. And the expensive, heavy case with the 1/4" front panel adds nothing to the sonics ... that’s 100% marketing.

Tube circuits, though, get expensive very fast. Top of the line transformers are NOT cheap. Precision regulated 300 to 450 volt supplies aren’t cheap either. Tubes, by themselves, are kind of mid-price, but again, the top-of-the-line models aren’t cheap either. Vacuum tubes have always been handmade, even in their heyday in the late Fifties. Same for transformers.

None of this should be surprising. The majority of the solid-state market use parts that are made in quantities of hundreds of thousands to millions. This drives down costs. By contrast, the tube sector uses parts that are made in quantities of hundreds to a few thousand, several orders of magnitude smaller than the solid-state sector.