Make FET like Triode: Trioderization

Trioderizer 將FET三極管化

Structurally, vacuum tube can be categorized into triode, pentode, etc. The famous tubes, like power tube WE 300B and signal tube ECC88, are triode tubes. The curve of triode, pentode are very different. After transistor is introduced to the world, we realized that FETs are more similar to pentode, instead of triode.

There are exceptions, V-FET that are invented by Japanese have triode-like curves; nevertheless, V-FETs are now obsolete and extremely rare plus expensive. V-FETs are produced in late 1970's, and are soon discontinued due to many reasons. Anyway, part of people love its sound, they think V-FET sounds "tubey", and this is the most attractive reason of getting V-FET---A transistor with vaccum sound and as powerful as common transistor!

A 300B amplifier can generate something like, say, 10 watts or so. But transistor amplifier including V-FET style, can easily generate several hundred watts.

You can call V-FET as a power tube without filament heat, that the truth.

Since V-FETs are now gone, then it comes to the question: Can we make transistors like triode?

The key to the answer might be trioderizer or solid state tube.

我們知道，真空管有三極管(TRIODE)、五極管(PENTODE)等構造區別，常見的300B、ECC88這些都是TRIODE。TRIODE的曲線特徵是曲線具有凹向上性；而PENTODE則是有平坦的高原區，到終端具有平直的特性。
電晶體包含JFET、MOSFET等，曲線都跟五極管相近，而如果要原生就有三極管的特性，那就得選用停產又珍貴的V-FET=SIT這類型的場效應電晶體。
也就是說，真空管可以隨意選擇三五極管，但是電晶體大多數只能具備五極管特性。



這邊先不討論三五極管誰優誰劣，而是試著提出一個問題:
"能不能把FET的曲線三極管化?"


為什麼把FET曲線三極管化值得討論呢?因為這代表著，如果可行，那我們可以用FET來取代12AU7、ECC88，或者在功率級用諸如2SK351來取代300B。
VFET可以直接達成這樣的需求，而FET需要額外的線路來完成。


這樣的線路有幾個好處，低壓、低廢熱、大瓦數、不須輸出變壓器/級間變壓器。
而VFET相當昂貴，因此"Trioderizer "就是一個值得探討的方向，因為它是利用FET搭配外圍的主被動元件來達成曲線類似三極管，這樣避開了VFET難尋且昂貴的問題。

我試著使用2SK330來做三極管，發現效果不錯，2SK209也有相同的表現。反倒是2SK170並不好做。而如果想要直接在高壓線路上使用FET取代真空管，則可以使用LND150這類的高壓FET。

其實這些線路就是"固態真空管"相似的研究領域，固態真空管就是想要利用電晶體元件來直接取代真空三極管。


而如果不要求直接能在高壓線路上替換，那就是Trioderizer這類的範疇了，藉由2SK330這類的低壓JFET搭配被動元件，來"模擬"低壓的真空三極管。
如果不想用Trioderizer，也不想買VFET，那目前剩下的唯一可行性就是Korg的NuTube了，利用VFD物理上的特性來直接達成三極管曲線特徵。

Follwing the troderizer idea, I did the simulation, using a JFET 2SK330 and passive parts, and it did work like triode! That means a low-voltage solid state tube without the need of filaments! Wow.

Figure: The 2SK330 JFET trioderized and its triode-like curves, 

Figure: A DHT (Direct-heated Tube) and its transistor mimic. The Walter's idea is more tube-like, the filament is also implemented in the circuitry, using a photo resistor and 6V supply to replace the real filament. Genius!

Figure: Small signal VFET Sony 2SK63 curve. Exactly showing triode properties.

A Japanese website, call it "Pseudo-triode", which is triodization of FET

Current mirror feedback

Fig. 1 circuit




Now, the Id-Vd characteristic of this is as follows.

Fig. 2 Id-Vd characteristic of the circuit of Fig. 1

Well, why is this happening? Actually, the key is in Q2, but let's take a look.


Fig. 3 Current mirror If

you write the currents of V1 and R1 of the current mirror in Fig. 3, it will be like this.


Fig. 4 

V1 and Relationship with the voltage of R1 

The point is that the slope increases due to this voltage. When
this is put in Q1, it is enlarged by the characteristics of Fig. 5 and the graph is as shown in Fig. 2.



Fig. 5 MOS-FET

I experimented with 2SK213 to test the Vg-Id characteristics . 

I used 2SA1015-GR and 2SK30A-GR. DC fluctuations stabilize immediately. Output impedance is 137Ω (at load open 1.6Vpp). did. 

Turn the resistance of the source of 2SK30 and when the drain is 10V at the output 0.5Vrms, the distortion is the lowest, and there are many secondary components such as 1KHz 0.20% 2nd 0.20% 3rd 0.03%. Input is 70mV did. 

The frequency response was roughly flat from 20 to 150KHz . The LM317P is made by STMicro.

Fig. 6 Experimental circuit


Fig. 7 The completed headphone amplifier

VR1 is adjusted so that the drain of 2SK213 is 8.4V. It seems that 1K is too small.

It is a headphone amplifier that can be used even though it has a high gain and has a sea sea noise. In

the latest circuit, R1 has two 10Ω parameters. 1K is sufficient for VR1.


Fig. 7-1 Harmonic distortion characteristic Load 66Ω 0.86Vrms output

This data is measured at 6600uF, which is C3 with two 3300uF, and is currently modified as such.

Fig. 7-1-1 Frequency characteristics



Fig. 7-1-2 DF



Figure 7-1-3 distortion property



diagram 7-1-4 harmonic characteristic

noise level 840uV 150uV (JIS-a hearing-corrected)

　　
　　　　　　1 KHz 10 KHz


　　　　　100 KHz





Now, a little more, let explore.


characteristic measurement of FIG. 7A IRF530 Circuit Now, if

you touch V2 and trace the curve, you will get familiar characteristics.


Fig. 8 IRF530 Id-Vd characteristics

, let's apply DG feedback.

Fig. 9 DG feedback circuit
, Id-Vd curve you will do this.

Figure 10 DG Id-Vd characteristics of IRF530 that the feedback

Thus, this is the same principle as the ultra-3 Yui Kamijo teacher. in other words, I always was a semiconductor ultra-three binding.

Well, in, What kind of characteristics would you like?

Fig. 11 2N3055 circuit
This curve looks like this.


Fig. 12 Fig. 11 Ic-Vce characteristics

This looks good, but if you draw a 10Ω load line that connects VCE = 30V, IC = 0 and VCE = 0, IC = 3A, it will intersect red at about 17V 1.75A. , The voltage utilization rate is (30V-17V) / 30V = 43%. 0 to 17V is wasted.
Here, from the slope of the curve, the internal resistance can be seen as about 5Ω.



Fig. 13 Improved circuit

Fig. 14 

In the circuit characteristics of Fig. 13 , the same load line of 30V to 10Ω is used. 

When subtracted, it intersects at about 3V, 2.6A.
Therefore, the voltage utilization rate is (30V-3V) / 30V = 90%. Therefore, the output is much larger than before. 

In other words, the curve of the upper triode has the same voltage utilization as the triode and the output cannot be obtained, but the characteristics of the pentode are like this.

This curve, which combines the characteristics of the and triode, has the advantages of both. In other
words, the internal resistance is low, the voltage utilization rate is high, and the output can be obtained.

Here, the internal resistance is about 5Ω due to the inclination of the curve . As you can see.



Figure 15 improved their 2

characteristic of the circuit of Figure 16 Figure 15

the voltage utilization rate has been little bad.
in this case, the slope of the curve, the internal resistance is seen at about 2 [Omega.


in FIG. 17 2SK3163

of this In this case,


the characteristics of the circuit shown in Fig. 18 and Fig. 17

The

internal resistance can be read as 2Ω near 2A from the inclination of the curve.

Therefore, I built an experimental circuit shown below:

 Fig. 19 Experimental circuit (passcon is omitted. )

The Zener is a strange one in the circuit diagram, but it is the one of 12V1W sold by Akizuki Denshi. It is a
class A single. As will be described later, DF = 4.

The distortion is 6% at 4W, 3W5%, 2W4%, 3.3% at 1W, which is almost secondary. The high range is flat up to 70KHz and -1dB at 150KHz. In the simulation of


Fig. 20　
, the output impedance was 1.4Ω, but the actual measurement was 2Ω. did.
In addition, since it becomes 40W at 25V 1.6A, it is cooled by a fan. Overshoot occurs at


　　　　　1KHz

1KHz. We are currently pursuing the cause.

Transistor CB feedback

Fig. 4 The relationship between the voltage of V1 and R1 shows that this is CB feedback, and Fig. 9 DG feedback circuit shows that it is DG feedback of MOS-FET.

Let's consider the CB return of 2SC1815.

2SC1815-GR

Figure 21 2SC1815 CB return

Fig. 22 Curve trace (simulation) of Fig. 21

The drive voltage is 1V step from -5V to 5V.





Fig. 22-1 Actual measurement of the circuit in Fig. 21

Measured with the curve tracer of Analog Discovery 2. The drive voltage traced 0-2V in 10 steps.
The vertical axis measures the current with 10Ω, so 400mV is 40mA.

Fig. 23 Ic-Vp characteristic simulation 0.2V step from the top to -1.6V at 2V

So I made an amplifier. A 3.3K load line is drawn in Figure 23. The circles are the operating points of the amplifier.



Fig. 24 The

actual measurement is shown below the circuit of the amplifier .



Fig. 25 Actual measurement waveform The

operating point is almost the same as the simulation.


Fig. 25-1 Frequency characteristics (actual measurement values)

These are the actual measurement values ​​of frequency characteristics.



Fig. 26 Distortion rate characteristics (measured value)

Fig. 27 Harmonic characteristics

Since the power supply voltage is 50V, the output voltage is limited to AC16V.

2SA1015-GR

Fig. 28 2SA1015 CB feedback circuit



Fig. 29 Ic-Vp characteristic simulation Load line is 3.3K Circles are operating points


Fig. 30 Amplifier operation (actual measurement) Circles are operating points Although

they are slightly off, they are generally correct .

J-FET 2SK170-GR

Fig. 31 

Vd-Ig characteristics of DG feedback of 2SK170-GR Feedback resistance 220K Drive resistance 10K Drive voltage is 0V to -1V.

The measured vertical axis is 100Ω, so 1V is 10mA.
It looks like a triode vacuum tube.

2SC2120-Y

Fig. 32 2SC2120-Y CB Feedback resistance 220KΩ Drive resistance 10KΩ

Measured with a curve tracer of Analog Discovery 2. The drive voltage traced 0-2V in 10 steps.
The vertical axis measures the current with 10Ω, so 400mV is 40mA.
It is almost the same as 2SC1815-GR.

MOS-FET BS170

BS170 has the same outer shape as 2SC1815. I bought it at Akizuki Denshi Tsusho.



Fig. 33 MOS-FET BS170 DG Feedback resistor 680K Drive resistor 47K

Measured with a curve tracer of Analog Discovery 2. The drive voltage traced 1-3V in 10 steps.
The vertical axis measures the current with 10Ω, so 1V is 100mA.



Fig. 34 

MOS-FET BS170 Simulation results are similar.



Fig. 35 Simulation circuit

I tried to simulate an amplifier. The drain voltage is 26.5629V in the simulation, but the measured value is 28.8V, which is a little different.
Well, I wonder if it's like this.



FIG. 36 The frequency characteristic (simulation)

gain is about 34 dB.


Fig. 36-1

Compared with the frequency characteristic (actual measurement) simulation, the actual gain is 31 dB, which is about 3 dB lower. The cutoffs are roughly correct.



Fig. 37 Output impedance (simulation)

Since it is a triode operation, the impedance is lower than the load resistance.


Figure 37-1 Output impedance (actual measurement) 

The output impedance is quite low.

2SA1015-GR

Figure 38 

2SA1015-GR CB feedback Feedback resistor 220K Drive 10K 0-2V

was measured with a 10-step Analog Discovery 2 curve tracer. The vertical axis is 400 mV and 40 mA.

When the load line is drawn, the current is small. That is, when the drive voltage is close to 0V, the interval is wide, and when the drive voltage is close to 2V, the interval is close.

2SA673AC

Figure 39 

2SA673AC CB feedback feedback resistor 220K drive 10K 0-2V

was measured with a 10-step Analog Discovery 2 curve tracer. The vertical axis is 400 mV and 40 mA.

When the load line is drawn, it is almost evenly spaced regardless of the drive voltage.

2SA950-Y

Figure 40 

2SA950-Y CB feedback feedback resistor 220K drive 10K 0-2V

was measured with a 10-step Analog Discovery 2 curve tracer. The vertical axis is 400 mV and 40 mA.

When the load line is drawn, the interval is tight when the drive voltage is close to 2V.

Principle of pseudo-triode tube circuit

The principle of the pseudo-triode circuit is based on NFB.



Fig. 41 When the Vc-Ic curve of the pseudo triode circuit and the load line

NFB are applied, the distortion becomes small and the intersections with the load line are evenly spaced.

This means that if you write a lot of load lines like this, you will have to have triode characteristics.
If the distance between the load lines is getting closer and closer, the minute front and back of a certain load line will have to be consistent, so it will have to become a triode characteristic.


Therefore, if feedback is applied, such as CB feedback or DG feedback, all will have triode characteristics.


If you think about it a little more, you can say that the transistors, FETs, and pentodes are in the final stage of the amplifier, and when NFB is applied, all of them have the Vc-Ic curve as a triode characteristic.
That's what NFB is when you think of it as a Vc-Ic curve.