12B4a Cathode Follower

[snoopyma]

I read with interests the concept of 12B4a liine stage using LM317 ccs at the cathode.

I want to try a line stage having the output stage of using 12B4a as cathode follower. My preliminary idea of this cf stage is:

(a) Shunt regulated B+ of 175 volt to the plate of 12B4A;

(b) LM317 current regulated using 48 ohm resistor (25ma) connected between 12B4a cathode and the ground with 220uf (+.1uf) bypass cap;

(c) 12B4a grid tied to ground with 1M ohm resistor.

Does this work? Is this so simple?

I don't know how to figure out the resistance of the LM317 ccs and the grid voltage here. Do I need to put another resistor between the LM317 ccs (317 with 48 ohm reisstor) and ground in order to bias the 12B4a correctly.

Any help and suggestion is appreciated. Thanks.

 

[barretter]

I have no idea about the technical side of your query, but why do you want to use the 12B4 as a cathode follower? I first used it 7 or 8 years ago as a line stage because it had low gain, low output impedance and reasonable linearity. What you want in a cathode follower is a combination of high gain and high transconductance to get a very low output impedance. I have used both EC91s and EC88s for this (both single triodes) and a lot of people use ECC88s in a White cathode follower circuit for even lower output impedance and good cable-driving qualities.

 

[Rtop84]

It's wasting very good tube. Use Gary Pimm CCS and mu-follower configuration for low output impedance.

 

[jeapel]

hi Uchi

Maybe can you help me with Gary Pimm ccs

If i understand in both case there is a buffer
12B4a(cathode follower) or irf820(mu-follower)

But what is the best sound and in your test when
you switch from just ccs (plate output) to
mu-follower output how is the sound because:

1- high mosfet Cin
2- no linear Cin with mosfet

thanks

 

[Brian Beck]

Caution about using 12B4 as CF

Usually you look for both high gm and moderately high mu for a CF. This is because the voltage gain of a CF is limited to mu/(mu+1). If mu is, say, 30 or higher (as in a 6DJ8) this factor is very close to one (about 0.97). With a low mu tube like the 12B4 (one of my favorites for common cathode use, BTW), this factor becomes 6.5/7.5 = 0.867. While this 1.2 dB loss may not be a problem, it does open the possibility for more second harmonic distortion since mu is now "more in the picture". Said another way, there is less negative feedback available for the CF. Whether this is sonically a problem or not is arguable. The 12B4 cathode output impedance (ignoring Rk for the moment) is about 1030/7.5 = about 140 ohms under data sheet bias conditions, and probably closer to 200 ohms under real world conditions. This is not bad, but a 6DJ8 can muster half of this value. As a side, if using a current source in the CF cathode circuit, I would NOT use a naked LM317. It's OK bypassed for common cathode service, as I showed earlier. You can use an LM317 with a simple cascode element (BJT, FET) above it to raise the output Z of the CCS at all frequencies. This gives the stability and predictability of the LM317 current setting based on a band-gap reference voltage and the cascode’s multiplication of the output Z.

 

[Rtop84]

If i understand in both case there is a buffer
12B4a(cathode follower) or irf820(mu-follower)

No, mu-follower configuration isn't buffer- gain isn't unity, but stage will have gain = tube mu.
In my preamp change from CCS to mu-follower CCS has given better dynamic and sound impact, it's caused by lower output impedance.
In Gary's CCS disadventages of mosfets(high and nonlinear Cin) are minimalized by cascode mosfet configuration with bias.

PS. Generally I don't like cathode follower sound, for me mu-follower with Gary Pimm's CCS is much better(more clear and more transparent).

 

[jeapel]

"In my preamp change from CCS to mu-follower CCS"

Exactly like me Uchi i want to change my C4S ccs to
Gary mu-follower ccs to speed up my preamp.

Maybe not a buffer but am i completly wrong or the sound pass through the irf820 .

thanks for sound feedback on Gary mu-follower

 

[snoopyma]

Thank you for your advice.

I am currently using a single-stage WE437a WOT (Lundahl transformer connected as 4:1) line amp. For the speakers, I am using horn speaker system of double 15" woofers per side and the cable connecting line amp and power amplifier is around 25' long. Therefore, I need a line amp of output impedance as low as practical. In my opinion and my previous experimentation, lower impedance can drive my speakers with more dynamics and speed and can make the amplified sound follow more closely to the original input music signal.

Hence, I would like to try a simple two stage design with the first stage as a pure voltage gain stage and the 2nd stage as the current output stage. For the output stage, I would try the cf which is simple and easy to implement.

Though there are many sayings that the cf is not good, I do hear a few line amp. using the cf which are sounding very good. I think it is the implementation which tell the differences.

Indeed I have tried experimenting cf using a 6EM7 (triode no. 2), EL504 (pentode) and even 6B4G (DHT) with the output transformer (20ma, 4:1) connected to their cathode as its load with very good results.

My own experience on achieving a good cf is:
(a) very tight and clean B+
(b) high gm and low impedance tube
(c) adequate dc current

This time I would like to try 12B4a, which is a high gm and low impedance triode. As a cf, it is better than ECC88 in theory. It is much cheaper too. As far as I know, output impedance of a cf is 1 / gm * u / (1+u), so it can achieve an output impedance of 130 ohm.

I do think that a LM317 ccs at its cathode can improve the 12B4a cf too. If it can work, I will probably try connecting a high quality output transformer to its cathode via a 4uf capacitor. As such, the output impedance can be much lower. It is similar to a parafeed line amp., but this time the output is taken from the cathode, not the plate.

This is what I am thinking of and what I want to try. I think the experiment is worthwhile as both 12B4a and LM317 are cheap.

So can anybody give me some TECHNICAL advice as to implementing a LM317 ccs at the cathode of 12B4a as a cf? If it can work properly, I think it can be applied to other similar tubes that I have. Thanks in advance.

Just my cents. Please feel free to correct me if I am wrong.

 

[snoopyma]

Re: Caution about using 12B4 as CF

Usually you look for both high gm and moderately high mu for a CF. This is because the voltage gain of a CF is limited to mu/(mu+1). If mu is, say, 30 or higher (as in a 6DJ8) this factor is very close to one (about 0.97). With a low mu tube like the 12B4 (one of my favorites for common cathode use, BTW), this factor becomes 6.5/7.5 = 0.867. While this 1.2 dB loss may not be a problem, it does open the possibility for more second harmonic distortion since mu is now "more in the picture". Said another way, there is less negative feedback available for the CF. Whether this is sonically a problem or not is arguable. The 12B4 cathode output impedance (ignoring Rk for the moment) is about 1030/7.5 = about 140 ohms under data sheet bias conditions, and probably closer to 200 ohms under real world conditions. This is not bad, but a 6DJ8 can muster half of this value. As a side, if using a current source in the CF cathode circuit, I would NOT use a naked LM317. It's OK bypassed for common cathode service, as I showed earlier. You can use an LM317 with a simple cascode element (BJT, FET) above it to raise the output Z of the CCS at all frequencies. This gives the stability and predictability of the LM317 current setting based on a band-gap reference voltage and the cascode's multiplication of the output Z.

Hi Brian,

Thank you for your informative reply. You are correct, as a cf, ECC88 will be better than a 12B4a. I will look for a high gm and moderately high mu tube for a cf.

I am very interested in what you have proposed in the current source in the cf cathode circuit. As I am a layman to the solid state component, can you elaborate more (such as what cascode element) and sketch me the circuit if possible if I want 20 to 30 ma current.

Thanks.

TC MA

 

[tube-lover]

hi TC,

why not U try the 5687??

looks if U like to try the CF 5687 will be a good choice. I had some experience about the CF with different tubes.

MF with OPT.

EC8010, 6c45,437a,417a/5842, EC900,EC86,EC88.

double triode I was try the ECC88,ECC86,5687& 5965. for CF stage. But I guess that 5687 & 5965 looks better than ECC88 & 86. Regarding running large current (approx 15ma~25ma) of the tubes. The damping looks quite good.

But the best performance that I prefer was use CF with the 600Ohm:600Ohm pre-amp OPT.

I was in Hong Kong. hope we can discuss more detail. I hope to share more experience that we had.

In my new DAc, the tube buffer was use EC900CF with the 600Ohm:600ohm. I choose EC900 beause this tube max only running in 11ma. I choose running in 150V, 7ma for the tube buffer with the 1541a DAC. So the pre-amp OPT will not too large.

If U try the parafeed had good result, pls let e know.

pls free to call me 93443443.

thx

thomas

 

[Rtop84]

Exactly like me Uchi i want to change my C4S ccs to
Gary mu-follower ccs to speed up my preamp.

You have understood me wrong- I changed Gary Pimm's CCS in normal CCS configuration to CCS mu-follower, not from C4S to Gary Pimm's CCS, but you can test and compare both solution(Batery Biased CCS is very simple and works great).

 

[Brian Beck]

I am very interested in what you have proposed in the current source in the cf cathode circuit. As I am a layman to the solid state component, can you elaborate more (such as what cascode element) and sketch me the circuit if possible if I want 20 to 30 ma current.

Well, 20 to 30 mA is a bit rich for a 6DJ8 which is max rated at 25 mA. 10 or 12 mA ought to be good enough for that triode. The LM317 is not rated below 10mA, IIRC, although I found most to work well at lower currents, if needed.

Here's the general diagram of a LM317 with cascode, although there are many, many variations. I haven't built the CCS exactly as drawn here, but it's pretty straightforward and should work fine. I like to put the heater supply to work to supply the CCS too. A BC549 would work fine for the NPN.

 

[snoopyma]

hi TC,


I was in Hong Kong. hope we can discuss more detail. I hope to share more experience that we had.
"
"
"
pls free to call me 93443443.

thx

thomas

Hi Thomas,

We spoke today.

I have to admit that I seldom come across a person in Hong Kong who possess such a broad spectrum of knowledge in diy audio covering digital, transformers as well as tubes. It must be a pity for me not to know you earlier.

I do think that you will be 'disturbed' by me frequently thru' the phone from now on. I hope that I can acquire your assistance and expertise in developing my next project: KR-T1610 single-ended amplfiers.

Feeling very glad to meet you here!!!

Best Regards,

T.C. MA

🙂

 

[snoopyma]

Well, 20 to 30 mA is a bit rich for a 6DJ8 which is max rated at 25 mA. 10 or 12 mA ought to be good enough for that triode. The LM317 is not rated below 10mA, IIRC, although I found most to work well at lower currents, if needed.

Here's the general diagram of a LM317 with cascode, although there are many, many variations. I haven't built the CCS exactly as drawn here, but it's pretty straightforward and should work fine. I like to put the heater supply to work to supply the CCS too. A BC549 would work fine for the NPN.

Hi Brian,

Thank you for your information. Before asking you further questions, I think I need to take some time to study your circuit as I am a slow learner.

Thanks again.

Best Regards,

T.C. MA

 

[Brian Beck]

For more headroom...

Snoopyma,

The headroom (before the CCS stops behaving like a good CCS) is limited in my prior posting above, if the triode is a 6DJ8 or any tube with a low cathode voltage (Vgk). If you bias a 6DJ8 at 100V and 10mA, the cathode voltage is about 2V to ground. In the circuit above, the maximum output voltage will be about 1.6Vrms before the CCS’s NPN saturates. This is enough for many power amps, but not all. If using a similarly biased 5687 or even a 12B4, the original circuit should work OK. So I’ve added three variations on a theme below for tubes like the 6DJ8, each of which provides for larger output signal swings that should drive any power amp.

Fig. A shows how to use two signal diodes or a “cheap red LED” (to follow the current fashion) to bias the NPN downward in voltage to allow the triode’s cathode more room to swing.

Fig. B is for tubes using a 12.6 VDC heater.

Fig. C is for when a negatively referenced DC heater supply is not desired or possible. A voltage divider string from B+ biases the NPN base (6+V) and also the grid (12+ V) through a 1M resistor. This option requires the addition of an input coupling capacitor.

As I said earlier, there are many more variations on this theme. One can make better and better CCSs with perhaps double cascodes and other tricks, but the circuits here can get you started relatively cheaply and easily and with decent performance. BTW, here’s a link to my servo CCS variation of SY’s “Heretical line stage” that may interest you.

Servo CCS link

 

[SY]

Just for grins, here's John Broskie's way to do it.

 

[Brian Beck]

Another way to skin the same cat. I really am remiss in not reading all of Broskie's prolific output. I must admit to getting a little restless early on slogging through all of his power supply noise rejection ideas - excellent though they are, and not staying up with his later writing. He is a clever "topologist" and deserves to be read. BTW, I notice here that Broskie has used two poles in his feedback path, first the integrator, and then the unspecified R and electrolytic cap, presumably to remove the opamp's residual signal output . I did something similar with my C2 in the posting linked above, but did so with some trepidation, and only at a relatively high 160Hz, to avoid underdamping the loop.

 

[SY]

Brian: This is not a previously-published circuit of Broskie's; he gave me permission to post it here.

The second pole was at my suggestion for exactly the reason you surmised. I just don't trust summing junctions. The pole can be pretty high; in my old preamp, I set the second pole at 50 Hz, figuring that even a crummy opamp can do very well at those frequencies.

 

[Brian Beck]

SY, I agree that 50 Hz - or perhaps even 160 Hz - is in the range where opamps can be trusted to behave themselves. As long as the two loop poles are spaced far apart there should be little bump in the bass.

 

[SY]

Ya know, it occurs to me that, Aikido-style, you could use the servo to null out power supply noise by returning its reference to the PS transistor's base...

Sorry, too OT. Back to the 12B4.