Over 7 years ago I made a run of a great circuit with a bit of a twist. The “Super-linear Cathode Follower” or SLCF is a great way of optimising the cathode follower stage and make it more linear by bootstrapping the anode and forcing the valve to operate in a vertical load line hence reducing the distortion significantly. And it does sound pretty good I have to say. If you aren’t familiar with this topology and want to learn a bit more, of course read Allen Wright’s amazing textbook or have a look at the following posts here.
Read more: SLCF to the rescue?The wrong valve for the job
I made a run of this circuit in a compact PCB which I ended up selling over time after people reached out to me for them despite I didn’t plan to sell them in the first place. Recently I was contacted by other builders who wanted the last few boards and needed to modify a Marantz 7 preamp and other similar gear to improve their output stage.
Well, you have to ask first what stage they have. I don’t want to get into any debate about the design quality of the Marantz preamp, however in my view the ECC83/12AX7 is the wrong valve for the job. Some wrongly say that the ECC83 is a poor valve, instead I say it behaves poorly in the wrong design! Its job is high-gain amplification for small signal, not to be a beefy cathode follower to drive cables. It has low current capability as well as transconductance. Not what you need for a cathode follower and indeed there are much better valves for the job. There were much better valves in the 50s, so not sure why the designers made such a poor compromise in the design, cost driven likely.
SLCF
Needless to say, the SLCF can improve the circuit to a degree. Let’s have a look:
The bottom circuit shows the standard stage with bias bootstrapped. The power supply noise rejection is as expected on this topology (i.e. 40dB) and relies on filtered power supply to keep a good SNR at the output. With a 100K load it does a respectable 0.005% THD at 2Vrms in theory. This THD creeps up significantly when you add the cable and input capacitances (at least 400pF if not more) and then if you are driving a lower impedance SS amp with 10K or less. The slew rate is closer to poor with just 3.5mA of current drive capability.
On top on the other hand is the SLCF implementation of the same stage. If this feels intimidating don’t worry. It may look complicated at the beginning but let me explain how we design a stage like this so you can do it too!
The design process
We need to start with the operating point we want the valve to work with. In this case, I wanted to keep as much as possible the original bias to minimise the changes to the preamp as per the owner’s expectation.
Looking at the anode characteristic curves we will start with a Vak=200V and about 4mA of anode current. For this we need Vgk=0.5V.
The top CCS formed by a pair of LND150 will ideally be biased at about 300/500uA for best temperature response (tempco).
So the current being sunk by the tail CCS is the sum of the cathode current and the top CCS bias current, so about 4.5mA.
With an Rk of 80 ohms (as per original design) we will get a bias of 360mV. For 0.5V we will need to raise it to 110 ohms. So we will go with the 360mV feedback despite the initial expectation was to place the bias point here:
We will use the simplified diagram (Fig 2 above) to calculate the components of this circuit. Let’s work out the voltages next. At least the bottom CCS will need 25V or more to operate properly, so let’s tart with the 25V. The cathode will be at 25V-0.36V but will ignore the small difference for practical purposes. If we want the valve to operate at 200V across anode-cathode, then the anode will be at 225V. Assuming the MOSFET T1 is an enhancement one NMOS, 4V will be at least the threshold Vgs. so the point at VFB will be 229V. RfB will need to manage 500uA of current and drop 204V across. That will give a theoretical value of 408K. As I have the common 390K resistor at hand, we will go with that and adjust the currents on the CCS which is going to be required regardless.
CFB forms a high-pass filter (HPF) with RFB so at least 100nF will be required for a 3dB corner at 4-5Hz. We have all the main components calculated!
Seems easy, however the practice is a tad different. There are 2 CCSs and a couple of DC feedback loops! Adjusting the stage is the key point to make this circuit operate as expected. Otherwise you will be disappointed with the result: a highly distorting cathode follower!
So, let’s get back to the detailed stage and look at the detail:
Now if we expand the details of the two CCS, the circuit becomes clearer. The bottom CCS is made out of a cascoded pair of BJT biased with a transistor array. In this case we don’t have a bipolar supply, just the +280V rail, so we bias the 4 LEDs with a 47K resistor. equally the CCS current is set with a multi turn trimpot (R5) which the value it’s roughly 3V/ICCS.
On the other hand the top CCS is formed of a pair of cascoded depletion FETs (LND150) and Rtrim sets the current. Given the currents at play we don’t have a need to add heatsinks to either Q1 or M1, so quite convenient.
I believe the rest of the circuit and component choices are straight-forward. If you have any questions, please shout!
A great advantage of this stage is the noise rejection from power supply which increases to a whooping 80dB.
The SLCF PCB configuration
The following diagram shows the final PCB configuration which all of its components. A few caveats:
- The STF3KN80K5 (my beloved MOSFET of choice) has included pair of protection Zeners.
- R7 is set as anode stopper to a degree and the paralleled LED isn’t used here given the low anode current.
- There is an output resistor R11 for stability as per cathode follower stages. You can tune this as needed.
- The only components not included on the boars are the coupling caps (Can and Cost) as well as the RL and the grid stopper (Rgs)
The resulting stage has 5 times lower distortion in the simulation and I’d expect this to be the case in practice. If you look at the tests I published before the THD was below 0.003% for 3Vrms. Excellent results on bandwidth as well.
This is how it looked when I built and abused of the board:
Stage calibration
Once built. My recommendation is to set the tail CCS to the target current of 4.5mA or whichever current you need as first step.
Then you proceed to calibrate the top CCS. You can measure the voltage across the valve but not at the VFB node which is the gate of the MOSFET. It will oscillate, guaranteed.
You can only calibrate this stage with some sort of FFT tool (PC or similar) where you can see the output harmonics and THD and tune the top CCS until you bring the harmonics to minimum and that will give you the best THD response:
Wrap up
The SLCF can improve greatly a normal cathode follower, yes. And it sounds great too. Mostly, its as clean as it gets and the improved performance generally generate further improvements on other parts of the system.
Will I pick up the ECC83 for this job? Hell, no. There are much better valves for this role for sure. I showed this already with the 6Z52P which is a beast for this job. You can use many high transconductance and with more current capability than this valve and get even better results.
Some time back, a fellow builder (Peter Hoenen) built an amazing D3a stage with these boards which he was very happy with, here is a view of them:
I hope some may find useful this longer explanation. Keen to hear your comments and to learn from your experience.
