SemiSouth SJEP120R100 Single Ended Follower Amplifier, with Cascode and Choke Load

[Ben Mah]

SemiSouth SJEP120R100 Single Ended Follower Amplifier, with Cascode and Choke Load

Always looking for a new diy project, I decided to do something with the two SJEP120R100 that I have. They started out in a diy stripboard version of a First Watt F2J. I really liked its sound, with its lower distortion, compared to the single ended SIT common source amps that I was listening to at the time. Now, I prefer the lower distortion level of common drain SIT amps, having built a few of them. So I decided to try a single ended SJEP120R100 follower amp.

The plan is to house it in a chassis that currently has two channels of a Sony 2SJ28 choke loaded follower. This chassis started out as a 2SJ28 choked loaded L'Amp (single ended choked loaded common source), and then FW F2J before its current iteration. The power supply is CLC, good for 23~24 VDC. The audio choke is Hammond 159ZC, 60mH, 0.7 Ohm DCR, and 2A DC current capacity.

So the plan is to make full use of the power supply voltage and choke current capacities. After some simulations, LTSpice showed that it may be possible for a choke loaded follower at 24V and 2A to output about 15W into 8 Ohm. However the dissipation may be pushing it a bit a around 45W, as the SJEP120R100 is a 114W device. To keep the same power output but with lower dissipation, I decided to add a cascode to the circuit.

The lower dissipation of the 2SJEP120R100 with the cascode in the circuit will also help in another way. When I had the F2J up and running, one of the 2SJEP120R100 had quite high gate leakage current. So running them with the higher dissipation would probably not work as the higher dissipation would increase the leakage current even more.

The schematic along with simulation results are shown below. The results are promising so the next step is to build the amplifier. Hopefully the amplifier will test as good as the simulated circuit.

 

[Zen Mod]

feel free to increase impedance of cascode potential divider

also cap across 10ktrimmer2

 

[Ben Mah]

Hi Zen Mod,

Thanks for your advice. Being a diyer who is learning as I go, I am trying to understand the issues. This is my first cascode circuit.

You mention "impedance" rather than resistance of the cascode potential divider so I am thinking, is there an AC issue? I assume the cascode potential divider is the 20k trimmer. As for DC, the current through the divider is only about 1mA. So in LTSpice I increased the trimmer value to 200k and compared the level of the signal at the mosfet drain V+. The level was the same with a 20k trimmer and a 200k trimmer. I checked the FFTs and there did not appear to any noticeable difference. It is very possible that I am not understanding how to increase the impedance of the cascode potential divider.

The bias 10k trimmer voltage comes from the V+ at the mosfet drain so the signal in that voltage would also appear in the bias voltage. I assume that the cap across the bottom of the bias trimmer would shunt the signal to ground, to keep it from the SJEP gate. I put in a capacitor in LTSpice and the signal that is in bias voltage definitely decreased drastically. So I believe I understand this issue.

 

[pinholer]

If the 20k ohm trimmer is adjusted so the resistance to ground is low, it provides a low impedance path for the signal from the SJEP source to ground. I don't know how low the pot will be adjusted realistically but I guess it's possible. A large value pot will keep the impedance to ground higher for the same DC voltage.

 

[Zen Mod]

I assume the cascode potential divider is the 20k trimmer.

take my "impedance" as resistance; make that "100K trimpot"

 

[Ben Mah]

The trimmer will end up at about half way as shown in the schematic, so if 20k trimmer, it would be approximately 10k to ground, and with an 100k trimmer it would be approximately 50k to ground. Perhaps the 20k trimmer is enough but it is an easy switch to a 100k trimmer.

As a diyer, most of my knowledge has been acquired from the diy electronics magazines in the old days and now in the age of the internet, from internet articles and diyA forums. I study various schematics and have modelled many circuits in LTSpice to understand them. As mentioned earlier I had previously built a version of the First Watt F2J. The SJEP bias voltage circuit in my circuit is electronically similar to that of the F2J. The main difference is the voltage in the F2J is from the SJEP drain and in my circuit the voltage is from the cascode drain. In both cases there is an 8k gate leak resistor to impede AC to the gate. The F2J has no capacitor from wiper to ground.

 

[Zen Mod]

do not mix from where voltage is and from where reference is

in both cases - F2J and your amp, refrence is form appropriate place to obtain and keep proper gate potential of part

if you make voltage divider for cascode voltage biasing, of higher resistance, your C5 will have lesser work, while recovery time etc still on bay

now, what I said for cap across trimer, SJEP gate is referenced to GND; when you add that cap, there is no way that ripple from positive rail sneak up on gate - voltage divider is decoupled

 

[Ben Mah]

Thank you, ZM, for the explanation. I understand your reasoning. It's better to do it right than to have it just work. 🙂

 

[Ben Mah]

I did a bit more LTSpice exploration as I have spare time and my mind was curious. I looked at the AC signal that was floating around in front of the SJEP gate. I did not include a capacitor at the bottom of the bias voltage trimmer so that any AC was not bypassed to ground. It seemed that there was enough impedance in the 8k resistor to the gate that the 2Vpp that was present in the bias DC voltage did not make it to the gate.

The simulation most likely does not consider possible high gate leakage current which may affect the impedance of the resistor, if my understanding of impedance is correct (actually I probably don't know what I am talking about). So the best thing to do is to include the capacitor.

But I do like to explore, and hopefully learn. 🤓

 

[Ben Mah]

So now I am delving into impedance - resistance, inductance, and capacitance. So much to learn and understand.

Since resistors are the only components and the connectors are short, the main component of impedance in this case is resistance. The signal riding on the bias voltage should drop drastically across the 8k resistor, and LTSpice does show that.

My education continues.

 

[Nelson Pass]

So many toys, so little time...

 

[EUVL]

If I have a pair of SJEPs to do a follower, which perhaps is a bit of a waste, this is what I have in Spice, since January.
The signal source only needs to drive 2x 2SK208 in parallel, which is a very easy load (also in capacitance).
The K208s will take care of any gate leakage current of the SJEP, shielding that from the input signal source.

The topology was first discussed here :
https://www.diyaudio.com/community/...g-headphone-buffer-with-adjustable-h2.376469/


Cheers,
Patrick

.

 

[Ben Mah]

Patrick, thank you for the schematic.

Originally I did consider a buffer at the input for biasing of the SJEP to avoid gate leakage current issues, but I decided against it because I wanted to keep it simple and not add a bipolar power supply for the buffer circuit.

 

[EUVL]

Using a FETlinton with the SJEP and having bipolar supply are completely different matters,
You can still do what I showed above with single supply and choke loading.

The advantage of the White follower vs choke load is that it is quasi-push pull.
Hence double the current driving capability.


Cheer,
Patrick

 

[Ben Mah]

The problem that I saw with a single supply on the input buffer was not able to output a reasonable low bias voltage to suit the choke loaded SJEP and still get enough voltage swing out of the buffer. Perhaps there is a way but I'm just a diyer so my electronics knowledge is limited.

I see that the White follower is quasi-push pull, but I will not be going that route. I also do not have four SJEPs.

 

[Ben Mah]

I finished the pcbs a while ago. There are separate pcbs for left and right channels. I sent the right channel to JLCPCB and I should get them any day now.

Here is the right channel:

 

[Ben Mah]

The right channel is done.

It is placed in the right side of the chassis that was a choke loaded 2SJ28 follower amplifier. It started up with no issues with a Dim Bulb Tester in place. No issues either with full AC power. The cascode Voltage and bias Voltage adjustments did their thing, and the Voltages and current were stable over time. I let it cook for a couple of hours with no issues. The Iq was set at 2A and the Voltage at the SJEP120R100 drain was set at 8.5V.

I did some FFT plots and the results were surprisingly in the same ballpark as the LTSpice simulations. FFT plots shown below.

I also hooked up a signal generator and oscilloscope and did a frequency sweep from 10Hz to 20kHz and beyond, and response looked flat.

Right now I am listening to this right channel together with my choked loaded 2SK180 follower, which has very similar distortion and power output, on the left channel. I am so used to the sound of single ended Tokin follower amplifiers now. This SJEP120R100 cascoded follower has the same sound as far as I can tell. It has the high frequency clarity and realism that I enjoy when combined with the DIY Front End 2022.

 

[Zen Mod]

so, it's not always just parts, something is also brain

 

[Ben Mah]

Well, the SJEP120R100 is a great part. I really like the harmonic distortion distribution at all power levels.

Since the right channel is working, the left channel pcb has been ordered, manufactured, and is awaiting transportation across the Pacific.

 

[Zen Mod]

considering what's format of pcb, small - so everything being sorta tight and centralized, you really don't need right and left

that's necessary with long pcbs, where by nature of things is best (or simply chosen) to have input on one end of pcb