DN2540 + LM317 cascode CCS?

[rongon]

Been doing a little experimenting recently and finally got around to putting a 1N5314 in the plate of a miniscule 6DJ8 line amp I have lying around. It's really cramped inside, and has only 120V B+ available, so it made sense to use as small a part as possible. I got Ea = 75V, Ec = 1.6V, Ia = 5mA (of course). The surprise was that it sounds very nice indeed, especially considering its complete lack of anything expensive in the signal path. So that got me to dreaming about changing out all the plate load resistors in my 'real' system for ccs's...

I was reading Walt Jung's AudioXpress articles about ccs. The version with DN2540 or 10M45S on top and an LM317 on the bottom looks to have excellent performance. (I think that's Fig. 13C.)

I've been searching the posts here and I've found this mentioned a couple of times, but only in passing. I haven't seen a real discussion of this particular topic.

I know the ccs using two DN2540's is the favorite, but what if I want a more repeatable, more easily adjusted version? Is putting the LM317 as the bottom half really going to do great damage to the sound of the circuit? It sure doesn't look like the performance suffers.

I have some little perfboards from Radio Shack to wire up. I'm hoping to get a couple of DN2540 + LM317 ccs's wired up this evening, and a couple more with dual DN2540's.

Do you think I'll hear a difference?

--

[SY]

Jung followed that up with a long letter to the editor showing the two DN2540 cascode- it handily outperformed the hybrid with a 317. Will you hear a difference? Unlikely.

[rongon]

Wow, ask and you shall receive. One view, one reply! Thanks, SY. I didn't know about the follow-up.

I guess I could use the 317 on the bottom for ease of setup, get things working in-circuit, then swap in the dual-DN2540 version to finish things up. I have a tendency to blow things up if I have to do too much tweaking, and I only have a few DN2540's.

[SY]

They're something like $1.50 from Mouser- you can get all you need in a day. It's actually really easy to set them up, easier (I think) than the 317.

[dsavitsk]

One advantage of the LM317 is that it's specs are a little more predictable than the DN250 which can be all over the place. This makes resistor selection easier. And you can get them at Radioshack. However, they are reputed to be prone to oscillate.

[wrenchone]

I have found that the DN2450s are remarkably consistent, at least within a given batch. However, if you don't use a stopper resistor while you're sorting, the results will wander all over the place due to oscillation. I had to retrofit my fet sorting jig with 1k stoppers, as I was getting wandering results even with medium-gm devices like the J110. My sorting jig uses variable current sources so I can select for Vgs for a given drain current.

[Sheldon]

Originally Posted by rongon

So that got me to dreaming about changing out all the plate load resistors in my 'real' system for ccs's...


I know the ccs using two DN2540's is the favorite, but what if I want a more repeatable, more easily adjusted version?

Consider a gyrator instead of a CCS - a couple extra resistors and a small cap. Adjust with a divider at the power supply. Also, the DN2540 is really useful, because it's a high voltage part. But the lower device doesn't have to be. Nor does it have to be depletion mode for the gyrator. You can use a low voltage JFet or MosFet in a TO92 package.

Sheldon