Ring of two CCS for differential triode amp - stability?

[ray_moth]

Dear All,

I'm interested in using a 'ring of two' constant current sink in the tail of a differential amp. Ring of two seems a nice, simple solution but I have read that such a CCS can oscillate. Does anyone know of a way to ensure that this doesn't happen, such as a strategically placed cap?

[Geek]

I have yet to have this configuration oscillate, by chosing transistors with a hfe of ~100-250.

[wrenchone]

The only time I had a ring of two CCS oscillate was when I included series base resistors in an attempt to make the thing more stable. These apparently recuced the bandwith of the circuit and introduced enough phase shift to cause trouble. When I removed the extra resistors and hooked the thing back together the usual way things worked just fine. I've never tinkered with the basic circuit since then, and have never had a problem.

[ray_moth]

To continue with this thread: first of all, thanks for the answers. I'm encouraged! Now, I'd like to ask more about the ring of two (and TIA for any responses).

For instance, in Morgan Jones's Valve Amplfiers, he mentions the ring of two, as well as several other CCS designs, but in his example circuits he doesn't use it. He seems to like the cascoded transistor CCS but offers no comparison with the ring of two, so I'm in the dark about the relative merits of each type.

On the face of it, the ring of two would seem to be a good solution, because it's very simple and doesn't need a
(potentially noisy) separate voltage reference LED or zener, relying instead on the B-E junction voltage of the lower transistor. What sort of temperature stability, impedance and capacitance it has I don't know, nor how it would compare in these respects with a CCS made, for instance, from a small-signal pentode or a transistor cascode.

Any help in understanding the ring of two will be most appreciated!

[gingertube]

Ray,
A quick rave regards my experience with CCS.

My original "Baby Huey" started with a "Ring of Two" CCS in the diff amp tail.
I changed that to a LED referenced single transistor CCS without noticing much difference, I thought it was perhaps marginally better.
I then changed it to a cascode transistor CCS and immediately noticed a more refined top end and better detail. This has to do with a higher output impedance and in particular maintaining the high impedance of the CCS out to higher frequencies. I later modelled these circuits and the higher Zout extending to higher frequencies (of the cascode transistor circuit) stood out like the proverbial canine testicles. Don't use a bigger device than you need at the top of the CCS. If a TO92 or an E-line will do the job then use it. If you need more power or voltage handling then go for a TO-126 like an MJE340 but it will not be quite as good due to higher device capacitances.
This experience is why I now use cascode transistor CCS on all diff amps and reserve "Ring of Two" and things built around LM317 etc. for places where output capacitance and maintaining the CCS output impedance to higher frequencies is not as critical (eg in the cathodes of output tubes). Some of those IXS CCS devices which lots of folk here seem to think are wonderful are not too flash with regard to output capacitance either and I won't use them.

BTW - after I fitted the cascode transistor CCS I could even notice a difference due to lead dress from the CCS to the cathode pins on the tube socket - No Kidding!!!! - You have to keep that capacitance at the output of the CCS, including stray C from wiring, low.

To actually address your thread title - I've never had a "Ring of Two" oscillate on me but since Mr Jones says its possible its something I check for on the rare occassions I use one.

Cheers,
Ian

[Serge66]

Hi Guys,
I think that a LM329DZ (6.9V low noise Zener) would be a suitable candidate to use in a CCS.
I modeled a CCS with a 6.2V zener and a LT1021-10 (10V ref). The zener (with 120nF across) CCS is superior. I have not built the circuits though. That will be a Winter project.

[Serge66]

Check these articles.

http://www.audioxpress.com/magsdirx/...a/jung2779.pdf

http://www.audioxpress.com/magsdirx/...a/jung2778.pdf

They have been posted here sometime.
Have fun.

[ray_moth]

Ian, thanks very much for that detailed and intersting reply. The answer seems clear regarding SS current sources/sinks for low current applicaqtions (like a CCS tail), namely, the cascoded transistor approach is superior to the ring of two.

Serge, thanks for those article links, which I will study.

[Serge66]

Ray, you are welcome.
Morgan Jones'book offers also good articles on CCS. He states that ring of two can oscillate.

Ian, what type of transistors are you using in your CCS? Bipolars? Mosfets? Jfets?

[gingertube]

serge66,
I use simple garden variety bipolars.
BC547B for up to 60V
2N5551 for up to 150V
MJE340 if I need up to 300V although the MPSA93 (check this part number, it might be the MPSA92) would be better if you don't need the power capability of the MJE340.

The moral of the story - smaller devices have lower device capacitances and WILL work better to higher frequencies.

Another hint:
The higher the reference voltage in your CCS the better for 2 reasons. The current set resistor in the emitter of the lower transistor is higher and it is the value of this resistor multiplied by the Beta of the 2 transistors which is approximately the value of the CCS output impedance which you want as high as possible. This is offset to some degree by "Early " effect in the transistors. Using a high reference voltage mitigates this effect.

And another hint:
For low current CCS you do not need huge currents in the reference chain. A really good approximately 6 to 7 volt reference can be made
by reverse connecting a small transistor (eg BC547B) and leaving the base open. Its becomes quite an amazing low noise zener which has practically 0 temperature dependence. Despite being reverse connected, the thing will not blow up as there is not enough current through it to damage it. Its good for up to a couple of mA.

For an example see the diffamp CCS in the schematic at post #54 here
http://www.diyaudio.com/forums/attac...amp=1141187034

Cheers,
Ian