Conrad Johnson ACT2 against Audio Research Ref 3

Excellent points John. I will call Leonard at ARC when I have a minute. I will post his answers here.

Simply put......The deficiency that Guido heard was a human error based on his activating SE and not balanced outputs on the Ref 3. It has nothing to do with a fully balanced system. Having said this however I rest assured that Guido is going to love his Ref 3 as I do.

Just spoke to leonard at ARC:

Ref 3 is a truly differential design. No splitters/inverters are used to simulate balanced operations. Two 6H30 tubes are used for the positive signal, and two for the return signal. If the Ref 3 is switched to SE operation through the remote control, a relay switch is activate which turns off the return signal on the XLR connectors and takes the 6H30 tubes responsible for the return signal out of the active circuit. Leonard suggests that if single ended operations are desired, the user should employ RCA ICs instead of balanced ones. However, he confirms that turning off the balanced signal on XLR, or using RCA connectors will still result in a 'good' sound with a predictable loss of 6DB of gain. This 'good' sound will however be noisier, slightly flat, and pinched, if compared to the sound of Ref 3 in a fully balanced configuration.

One further note, according to Leonard, the 6550C tube employed on Ref 3 is in fact an original SED Wing 6550 manufactured in the St. Petersburg plant, rather than the lesser pseudo-Svetlana equivalent.

Please note that in the previous post I should have been more specific: instead of 'sound' I should have specified 'Ref 3 sound signature'. The reason of course is that the actual sound is a function of all components. As such, we can't make a categorical statement about how best to run the Ref 3--single ended or balanced. It will all depend on the rest of the system.

Guido: Thanks for the update. Sounds very much like my scenario 2 on the input except the "-" phase is deactivated rather than simply grounding the input which results in the same .... and scenario 1 on the output. Good to know as both of these are the "purist" approach. Any word on work done here to trickle down to a new LS model?