11-23-14: MinkwelderSeconding George's comment about impedance compatibility, I suspect that a significant contributor to the divergent opinions about in-line passive attenuators is the differing impedances of the components people use with them.
I have read some very negative things about the attenuators, but others seemed very enthusiastic about them, which leads me to be very cautious about going that route.
Based on measurements I once made of the Rothwell 10 db attenuators (which I used successfully, btw), the measurement having been performed with a not particularly accurate analog multimeter, they consist of a series resistor of roughly 22K, and a shunt resistor at their output of roughly 11K. Which together with a power amp input impedance of say 50K would amount to a load seen by the preamp of about 31K, or roughly 28K to 30K in the case of most power amps having significantly lower input impedances. Those values will be no problem for most preamps to drive, but might verge on causing perceptible sonic issues with some, especially some tube preamps.
Also, those values would result in the power amp seeing a source impedance (the impedance it sees looking back toward the preamp) in the rough vicinity of 7K. Conceivably that could sometimes have audibly significant consequences, especially in the case of solid state power amps having relatively low input impedances, and especially if that input impedance is not constant across the audible spectrum and beyond.
This assumes that the attenuator is placed at the input of a power amp. Similar considerations apply if it is placed at the input of the preamp, except that it would be the input impedance of the preamp and the output impedance of the source component that are relevant.
So I would say that in general the in-line attenuators stand the greatest chance of producing good results when used at the input of a tube-based component that is driven by a solid state component. And in general they can probably be expected to have the worst chance of producing good results when used at the input of a solid-state component that is driven by a tube-based component. Of course, there will always be exceptions.
Finally, I would not totally rule out the possibility that a few of the reports of really negative experiences with these devices may have been the result of installing them in the wrong place or in the wrong direction. They should always be inserted directly into the input jacks of the destination component, with no intervening cable. Inserting them into the output jacks of the component supplying the signal, for example, would generally result in a sonic disaster (although it would not cause any damage).
Regards,
-- Al