Sony vFET Illustrated build guide

6L6

Moderator
Joined 2010
Paid Member
Nelson Pass testing procedure and instructions for DIY Sony Vfet amp -

Print this out!! :)

http://www.firstwatt.com/pdf/art_diy_sony_vfet.pdf

Pass DIY Sony Vfet amplifier illustrated build guide.


The Pass DIY Sony Vfet is a power amplifier from the brilliant mind of our technical, spiritual, and menu advisor, Nelson Pass. Quite possibly the most generous design yet, as Nelson released his private stash of the out of production 1970's Vfet to the DIY community so we can build an amp that nobody else can build nor buy. :)

How the circuit works


The Front End

DIY_Vfet_Front-end_copy.jpg




The overall schematic of the VFET amp is somewhat complicated in spite of the simple gain path – this is due to the special DC regulation circuits to cater to the fussy VFET transistors in the output stage.

Fortunately the front end voltage gain stage of the amplifier is pretty simple.

The Vfet DIY Front-end is basically a mini-F5. It serves to give us enough voltage and current gain to drive the Vfets so we can wire the Vfet as followers (known as “Common Drain operation, where no voltage gain is acquired, only current).

The input signal starts at R1 (2,2K), which acts as a “gate-stopper” resistor for the input Jfets, as well as a current limiter in case the attached preamp has a big turn-on or -off thump. R2, (47K) is from the input of the Jfet to ground, this gives the Jfet gates a reference to ground, and also sets the input impedance, which is about 50K.

The signal goes through the Jfets (Q1,2) and the gain currents at their Drain outputs go through the cascode transistors (Q3,4) and then to drive the Gates of the voltage amplification (Vas) Mosfets (Q5,6). The cascode transistors, Q3 and Q4, act as a voltage umbrellas, protecting the Jfets from the voltages appearing at T11 and T13 so that their Drains experience a fraction of the DC and AC voltage T11 and T13 as determined by R7, R9 and R8, R10 driving Q3 and Q4.

To quote Mr. Pass, “Cascoding is a very useful technique where the gain transistor (in this case each of the Jfets) is coupled to another transistor operated in Common-Gate or Common-Base mode. This cascode transistor contributes very little of its own characteristic to the amplification, but it acts like a voltage umbrella, shielding the gain transistor from the high DC voltage and noise of the power supply rails.”

The Drain pins of Q5 and Q6 deliver the output voltage (but not enough current) to drive loudspeakers. This point is found at test node T18.

Mosfet Q5 and Q6 need some voltage applied to their gates in order to turn on (bias) and amplify signal, and this is provided by the voltage appearing across R35 + P3 and R34 +P4. P3 and P4 are there to adjust this voltage so as to set the right amount of current through Q4 and Q6. This right amount of current is defined by 1.5 V DC appearing across resistors R5 and R6 AND the output of the front end (T18) near 0 volts DC. So P3 and P4 not only adjust the idle current through the voltage gain transistors Q5 and Q6, but they also set the DC output that appears on their Drain pins.

Adjusting the Front End bias

Bias is set by measuring the DC voltage across R5, (test nodes T6-T7), and R6 (test nodes T8-T9). Proper bias is when you have 1.5V across each of these resistors, at the same time also having 0 volts DC offset, which is the voltage at T18 to GND.

Adjusting P3 and P4 will make changes in all three of those readings, so in my opinion, you must have three DC voltmeters to make this an easy job. Even the very inexpensive Chinese DMM will work fine for this, just make sure they have good, fresh batteries in them.

A single voltmeter will do that job, but it is more tedious. In any case you want to adjust each pot in small steps, going between the pots and voltage measurements in rotation.

You start with both potentiometers at full counterclockwise, which is the lowest bias setting. At first you are simply looking to get the voltages across R5 and R6 up to 1.5 in small steps. You turn P3 just a little bit, measuring a small change across R5, and then you do the same for P4 and R6. And then you go back to P3/R6 and repeat this process, raising the DC voltages slowly and in equal amounts.

You also want to check the DC voltage from T18 to ground as you do this – you want to keep it near 0V if possible.

I suggest you pause at around 1 V on R5 and R6 – as the Mosfets heat up, the current will rise on its own. Eventually you will find yourself with “warm” settings which give about 1.5V DC bias and close to 0 V DC output at T18.

How good is good enough? 1.3 V to 1.7V is OK for bias. +/-200 mV is good enough for offset at this point. We will tweak these values after the completed amplifier has run for a bit.

REMEMBER – BABY STEPS ON ALL ADJUSTMENTS.

If for some reason you can’t achieve full bias with zero offset, please troubleshoot in the build guide thread, and do tell us how much bias you can get with zero offset.


Vfet_Schematic.jpg


Please refer to the above schematic for component names

The output stage -





The Vfets themselves are wired as a complementary pair in source-follower mode. These add current gain to the signal sent to their Gate pins (to be able to drive speakers with a low output impedance) but have no voltage gain.

That’s it - the voltage gain has happened in the Front-End, and the Vfets just add current capability. :)

 But it’s not quite that simple… is it?

Well, it actually is, until you realize that the DC powering up the Vfets must be regulated for a stable DC voltage and low noise, the same with the bias voltage. Also, both the rail voltage and the bias voltage have to power-up and power-down in a specific order to keep the Vfets from trying to draw a ton of current when the circuit is powered up.



What’s the proof of Nelson’s genius in this circuit? In how simple (and quite effective) the regulator sections have been designed. Let’s talk about that now.



Regulators



Each rail of each channel has 2 regulators, for a total of 4. Each of the two rails has a bias regulator and an output stage regulator. For clarity’s sake, I’m now going to talk about the positive rail’s regs - the negative rails have the same thing, just in a mirror image.



Both regulators use a TL431, a nice little 3-pin shunt reg that can have it’s output voltage set by 2 resistors. 

Looking at the schematic around Q11, you can see the unregulated voltage coming from the rail through a simple RC filter (R29 and C7) then through R13 (2.21K) and then the set resistors R11 (10K) and P1 (5K pot).

Because one of the two set resistors is a potentiometer wired as a variable resistor, changing that pot value changes the set voltage which provides Vfet voltage bias. It is very important that the bias voltage is place when power is applied to the Drain of the VFET, so this regulator circuit (bias reg) is designed to charge up much faster than the regulated voltage powering the Drain. (The output stage regulator)



Remember how it was mentioned earlier that the Sony Vfets are “Depletion” mode? They are normally on, letting current flow, and require a negative voltage applied to the gate to turn them off (this for the N type – The P channel devices require a positive voltage), or to otherwise set the value of the current (bias) through the VFET.

Looking at the regulated voltage at T19, (+14V) you want to note that it crosses over and connects to the gate of the negative P-channel Vfet Q10. The positive N-channel Vfet Q9, is likewise biased from the opposite rail. This is the opposite of what you are normally going to see with complementary Mosfet followers.



Moving on to the output regulators, you can see the circuitry near Q13 looks similar to the other reg, the exception being the voltage setting resistors are both fixed resistors. (R15, R17) These resistors and the TL431 make a voltage of 24V, seen at T2. This fixed voltage is controls the gate of a power Mosfet, Q7. This Mosfet is slaved to the regulator and lets the tiny TO-92 package regulator control many amps of current. Wonderful! This is necessary because the regulator needs to regulate the power going to the output stage which drives your speakers.

However there is no free lunch, and the Mosfet eats up about 4.2V (the Mosfet’s Vgs) in payment for doing it’s job. This is why you see a regulated rail voltage presented to the drains of the*Vfets of 19.8V (or so). An important part of the rail regulators is C1 (220uF), there to delay the arrival of this 19.8V to give the bias regulator time to ramp up. More about this here;



Power-up and power-down sequencing.



Back to depletion mode devices - The Vfets are normally on, right? We need to apply a bias voltage to control them. If the bias voltage isn’t there controlling current flow when the rail regulators power up, the Vfets will conduct an excessive amount of current and the 40-year old, completely irreplaceable smoke made from vintage Japanese un-obtainium, ground unicorn horn and pixie dust will escape. A sad result which will also smell bad.


So we must have the bias regulator working before the rail voltage is turned on. 

How did Nelson do it? A simple application of the time constant to charge 2 differently sized capacitors in the regulator circuits. The regulator with the smaller capacitor will turn on before the one with the larger cap. These caps in the race are C3 (47uF) on the bias reg and C1 (220uF) on the output regulator. It’s a 4.7-to-1 ratio and C3 will be charged about 4.7 times sooner.



The TL431 are shunt regulators, so they do not conduct current until the voltage rises to the regulation value, and then they conduct so as to hold that value.


So what about power down? The rail voltage needs to be turned off before the bias voltage. (the opposite of power-up) BUT, the big cap/small cap ratio will NOT work in our favor when shutting down. It would work exactly how you’d expect, where the small cap will discharge before the big one.

So what? The VFETs will drain the supply quickly enough, but when the supply voltage drops down quickly we may find the Gate to Source voltage (Vgs) of the Mosfet at a high enough voltage to stress the transistor into failure. It's not that likely, but everyone knows that Pass wears both a belt and suspenders to hold up his pants, and this is reflected in his design.

The fix is a simple 2-cent diode. On power down, D1 will be a very attractive and very low-resistance discharge path for C1, discharging it rapidly and turning Q7 (and Q13) off, and therefore the rail voltage to the Vfets. Because there is no special path for C3 in the bias regulator to discharge, it will operate for a longer time, which is perfectly ok.



The stuff in the middle.

Ok, so far everything has been explained except for the wiggles in the center of the schematic near T16 and T17. Remember how the bias voltage regulator is set by the divider made from R11 and P1? This is a DC value but we also need to let the AC have it’s say in the circuit, as the AC is our music. :) The Front-End output signal was all big and powerful the last time we left it at T18. Now this signal needs to get to the Gates of the Vfets to move our speakers, and that signal (even though it’s AC) can get pretty darn big. Big enough that it’s voltage could effect the current flowing through the regulator set resistors. The network of C5, C6 and R21 through R24 prevent the music signal from having much influence on the bias regulators. Stable bias makes for happy Vfets, so this is a good thing.



What’s left… Um… R25 R26 (221ohm) are gatestoppers for the Vfet, R33 (22.1K) adjusts the brightness of the LED (which must be blue, this is a Pass amp, after all…) and R32 (0.1ohm) gives a place to measure the output stage current. (Bias)



Global feedback is applied from the output stage from the Vfets to the Jfets through R3 (2.21K), and then R4 (332ohm) to ground.


Cool! That's the basic overview of how it works. Let's build!!

:D :D :D





.
 
Last edited:
  • Like
Reactions: 1 user

6L6

Moderator
Joined 2010
Paid Member
Assembly of the Vfet amp PCB


Nelson's notes for testing and power up - http://www.firstwatt.com/pdf/art_diy_sony_vfet.pdf

Vfet_Schematic.jpg


Schematic.

Note that Q3 is NPN the kit includes a ZTX450 for this position, Q4 is PNP and the ZTX550.



VfetPSU.png

Example PSU Schematic



Please take a look at this photo -

IMG_1339.jpeg


I would like to take this opportunity to strongly suggest that you print out the schematic of the circuit and tape it to a spot where you can see it during the entire process. Does it need to be as big as shown? Well, no, but it certainly doesn’t hurt.


After printing out the appropriate schematics, build the PSU. Detailed notes are in another post in this thread. You’ll need the PSU operative for part 2 of the amp PCB test. Of course, those of you with bipolar bench PSU can use those for the testing, but most builders will need to use the amp’s PSU for that.



Begin by stuffing only the Resistors and Potentiometers.

IMG_1337.jpeg


How the PCB should appear for Testing Part 1 - Testing the resistors.

Please note, at this point the PCB has only the resistors and potentiometers, no semiconductors (including diodes) nor capacitors.

This test is very straightforward - print out the page of Nelson’s notes with the chart and measure as instructed. During this test a number of resistances are measured to verify that 1) the proper values are stuffed in the proper locations, and 2) to see if there are any gross soldering errors so far. If you see any measured values much different than on the chart, check the value of the component and check your solder joints. If everything measures incorrectly, replace the battery in your multimeter.

IMG_1336.jpeg


Did the test go well? If so, move on. If not, look it over, find the error, and correct it.


There are certain things that need to next be mounted to the PCB in conjunction with the heatsinks and the T-brackets. Let’s go over that now.

The first step is to mount the T-brackets to the heatsinks.

IMG_1343.jpeg


IMG_1358.jpeg


Apply heatsink goop to the bracket and very firmly press it into the heatsink, to spread it into the irregularities of the aluminum surfaces. On my first attempt I didn’t put nearly enough down and it looked like this.

IMG_1361.jpeg


More was applied, the surfaces pressed and smeared to get good coverage, but there were still gaps.

IMG_1362.jpeg


I realized that mere pressure wasn’t enough, so a thin layer was spread around both sides of the interface. That looks better.

Remember, there is such a thing as too much heatsink goop, it’s purpose is to fill in any air gaps, normal white goop is slightly thermally insulating, but it’s much more conductive than air… So a thin layer is better than a thick one.

However, if you have a heatsink compound that’s thermally conductive, like Arctic Silver, this is a perfect application for it. There is no electrical connection in this interface, just thermal, so use it if you have it.

Now secure the T-bracket to the heatsink, and observe the proper torque pattern below.


IMG_1357.jpeg

M3 screw, lock washer, washer.


The brackets need to be attached like a cylinder head attached to an engine block, you must start in the center and work your way out, in order to insure the bracket is as flat as possible on the heatsink. If you proceed as shown it will have the greatest likelihood of being flat. Just attaching the screws randomly will almost guarantee a gap under the T-bracket. A gap will be easy enough to see.

IMG_1363.jpg

Start with these two screws and torque them.

IMG_1364.jpg


Then the other two in the inside.

IMG_1365.jpg


One set of opposite corners

IMG_13661.jpg


And the final corners.



Messages_Image_2071078713_.png


The Torque order shown on one photo



IMG_1367.jpg


Please note that the mounting holes for the TO-3 package Vfet are not symmetrical. The pins (center holes) are slightly offset in one direction regarding the mounting screws. And the PCB mounts under the T-bracket. So make sure the holes in the PCB align with the holes in the bracket.

Also please see that there are no specific holes to mount the amp PCB to the T-bracket. The screws for the Mosfets and and the Vfets are sufficient to hold everything together.


IMG_1370.jpg


Capacitors and regulators installed.


IMG_1377.jpg

Stuff the capacitors, TO-220 package Mosfet with heatsinks, regulators, Jfets, BJT, and diodes (some of those are not shown in this photo)


IMG_1371.jpg


The small mosfets shown are in the TO-220FP package, so no insulator is required. You do need a bit of heatsink compound, so smear a bit on and try to achieve better coverage than in this photo. If your mosfets have an exposed metal tab then you'll need a small plastic sleeve for the holes and an insulator. These sleeves are supplied in the supplemental kit along with TO-220 mica insulators which require grease on each side.

IMG_13761.jpg

Use this hardware to attach the assembly to the board as shown below.

IMG_13751.jpg



IMG_13731.jpg




Attaching the power Mosfet will give sufficient hold to the PCB and the bracket.

IMG_1381.jpg


Bend the leads down, and apply the mica insulator and goop.


IMG_13801.jpg


Get this hardware from the parts kit,

IMG_13831.jpg


And attach as such. The small washer is under the PCB on the bottom, and the lock washer is under the nut above the fender washer on the top.



IMG_1382.jpg

Attach both devices

IMG_13842.jpg

And verify that the holes for the TO-3 Vfets align in the bracket properly.

If so, now you can tighten the power Mosfet screws, and solder the pins to the PCB.


IMG_1369.jpg

At this point all this should be stuffed on the PCB.
 
Last edited:
  • Like
Reactions: 1 user

6L6

Moderator
Joined 2010
Paid Member
APPLYING POWER TO THE FRONT END

In this section we will apply power to the amplifier PCB, one at a time and set the bias on the front end and check a few things in preparation for mounting the Vfet. If you have a bench PSU, you may use it.

Everything is on the amp PCB except the Vfet.


IMG_1395.jpg


IMG_1396.jpg


P3 and P4 are full counterclockwise

Reference the 2nd chart that Nelson made in his documentation and test for DC volts across the test points one at a time.

IMG_1401.jpg





A few notes on those values -

G-T18 This may initially read very different than 0. If so, make a small adjustment to P3 and set until it reads +/- 1.0v . There are tiny currents involved at this time, so small changes to the pot should make a big difference. Setting exactly to 0.0 is not important, just get near 0

G-T19 , G-T20, G-T17, G-T16 The suggested voltages are very rough. Within 3-4 volts is fine. The check is looking for shorts or opens. If you see nothing, or the rail, you have a problem, but if you see something close to the suggested value, it’s good.

Now look at your Vfets for the numbers hand-written on them. This will give a number that will define the starting value of Bias - (and this bias is set before installation of the Vfet)

My Vfets had the following:

J28 =79
K82=78

Take that number, place a decimal and get 7.9 and 7.8, and then

While reading DC volts from Ground to T16, adjust P1 until you read 8.9v (7.9 + 1.0v)

While reading DC volts from Ground to T17, adjust P2 until you read -8.8v (-7.8 - 1.0v)

Yes, the pot is on the opposite side of the PCB to the test points. And yes, the N-Channel device needs a negative bias voltage, proving that these devices are, in fact, depletion mode. :)

Take time to read and follow the section in Nelsons guide “Part 3 - Biasing up the Front End” as it is very clear and talks about the necessity for adjusting in small steps, setting the bias low until things warm up, and setting for minimum offset. Again, as we need to look at 3 voltages at one time, it’s a perfect excuse to buy 3 DMM and install clip leads to them. As I’ve said before, having 3 meters is incredibly helpful and although not strictly necessary, there is no good reason to use fewer. Cheap $5 meters work just fine.


IMG_1397.jpg



Ok, is the smoke still in? Are all your test voltages reading good? Next step,


INSTALLING THE VFETs

Assuming all the voltages in the part 3 procedure are good, the Vfets can be installed now.


IMG_1404.jpg



IMG_1405.jpg


Make sure the insulators are in placed in the bracket.

IMG_1407.jpg


The mica needs heatsink goop on both sides. Cover all of the mica, but try to not get it too thick.

Notice how the PCB labels for the TO-3 devices are covered up? Don't put the Vfet in the wrong location... The left side (with the mica in the above photo) is Q9, the 2SK82, and the right side is Q10, the 2SJ28.

IMG_1408.jpg


And a thin layer on the bottom of the Vfet as well. This is the right side Vfet, Q10 (2SJ28)

IMG_1410.jpeg


It is extremely important to understand that the case of the TO-3 Vfets is conductive, and the only connection of Drain to the circuit. Therefore the cases must be attached properly, and the proper washers used. Bolt from the top, small washer, lock washer, nut. Make sure everything is secure and well-aligned, then solder the pins.
 
Last edited:
  • Like
Reactions: 1 user

6L6

Moderator
Joined 2010
Paid Member
Last edited:
  • Like
Reactions: 1 user
Adjusting the Front End bias

Bias is set by measuring the DC voltage across R5, (test nodes T6-T7), and R6 (test nodes T8-T9). Proper bias is when you have 1.5V across each of these resistors, at the same time also having ) volts DC offset, which is the voltage at T18 to GND.

@just a typo, in case of bracket there should be a number?

happy tnx for your guide, just have a belt and need to buy suspenders before i built it, right ? : )
 
Measuring v-fets Vgs for starting bias point

Thank you Jim! The case for my vfet amp arrived today so I will be following closely :)

I do have one question: I know the article presupposes that one bought the full vfet kit ( which I did ) - but I also bought vfets long before the kits were released. How would I go about measuring them ( safely obtaining an accurate Vgs ) to set the right bias point if that was all I had? Guessing there will be others in that boat with vfets they cant afford to mistakenly burn.

Not looking a gift horse in the mouth. I will be forever grateful for your help on my BA3FE and F4 :)

Thank you much!

NP from he Sony Vfet Part 2 article:
"I matched the VFETs for a Vgs within 0.1 volts at 0.5 amp current at 20 volts, and 0.2V Vgs is good enough. Some of you won't have good matches or only enough VFETs for 1 pair per channel, but you can build the amplifier and it will work fine, with somewhat less power and higher distortion. I tested that with 1 amp bias current, and it did just under 20 watts into 8 ohms."
 
Member
Joined 2012
Paid Member
Thank you Jim! The case for my vfet amp arrived today so I will be following closely :)

I do have one question: I know the article presupposes that one bought the full vfet kit ( which I did ) - but I also bought vfets long before the kits were released. How would I go about measuring them ( safely obtaining an accurate Vgs ) to set the right bias point if that was all I had? Guessing there will be others in that boat with vfets they cant afford to mistakenly burn.

Not looking a gift horse in the mouth. I will be forever grateful for your help on my BA3FE and F4 :)

Thank you much!

NP from he Sony Vfet Part 2 article:
"I matched the VFETs for a Vgs within 0.1 volts at 0.5 amp current at 20 volts, and 0.2V Vgs is good enough. Some of you won't have good matches or only enough VFETs for 1 pair per channel, but you can build the amplifier and it will work fine, with somewhat less power and higher distortion. I tested that with 1 amp bias current, and it did just under 20 watts into 8 ohms."

This is what Nelson presented how to measure the VFETs
 

Attachments

  • VFET%20TEST.gif
    VFET%20TEST.gif
    51.9 KB · Views: 2,134