The Birth of the TS2019-PP80 Amplifier

Written by Menno van der Veen

www.mennovanderveen.nl info@mennovanderveen.nl
2021: all rights reserved: ir. bureau Vanderveen & TubeSociety

This web-publication is a not-ready-yet publication.

Publication is mandatory for the continuity of our present TubeSociety activities.

The Dutch TubeSociety Academy designs challenging tube amps. The TS-2019 project (heavily delayed by
corona) focused on a 80 Watt push-pull amplifier, based on an earlier PR20HE all tube design. We improved it
with a new Fet-driver stage, applied new toroidal transformers plus a new clean MSCF power supply section.
Our research resulted in two versions, with the same components, but impressive differences in measurements
and subjective qualities. Both versions are explained in this publication, while additional information can be
found in the student reports about this project. Most text is English, some is in Dutch (use Google translate).

1. How all started

Around 2006 I designed the PR20HE: an all tube 70 Watt push-pull amplifier, with ECC81 driver
section and two output tubes ranging from KT88 to the newest KT150. Click the next link to find all
the details of this design.

Figure-1: PR20HE all tube push-pull amplifier

https://mennovanderveen.nl/cms/images/producten/PR20HE_DIY-kit/PR20HE-UK.zip

The essence of this ‘oldy’ is a 720 Vdc supply voltage, where the tubes energy is delivered into an 8
kOhm push-pull EI-core output transformer. Then you easily get around 70 Watt output power. This
amp was sold worldwide as a kit (without case) and received very positive subjective ratings. Its DIY-
construction-success-rate was 100 % (at least I heard no complains).

However, it’s by-Menno-designed EI-core transformers (Pow80 and Git80) ceased production around
2018. They needed improvement, resulting in the new toroidal versions Pow80-t and Git80-t; see the
links below.

https://mennovanderveen.nl/cms/index.php/nl/producten/other/vdv-git80-t-detail

https://mennovanderveen.nl/cms/index.php/nl/producten/valve-power-transformers/vdv-pow80-t-detail
For our TubeSociety-2019 project I proposed to my students:

“wouldn’t it be challenging to improve the driver circuit with a modern Fet driver stage (more gain,
better linearity), optimize this for the wider bandwidth Git80-t and make the power supply section
extremely clean with my MSCF-stage”?

They agreed and this resulted in the amplifier and power supply circuits shown below.

Figure-2: New TS2019-PP80 amplifier circuit with Fet-driver stage and toroidal OPT

Figure-3: New TS2019-PP80 power supply circuit with MSCF ripple rejection and soft startup
 2. Comparing PR20HE with TS2019-PP80

When I started the design of the new Fet driver stage, I took my old PR20HE amplifier, removed the
ECC81 and soldered the DN2540-fets directly on the pin side of the tube socket. Next I added 12 V
Zener diodes for switch-on protection and all functioned perfect. Bingo, it is that simple! But I
noticed differences in gain especially, while bandwidth and distortion staid rather unchanged. So,
before proceeding, lets discuss this shortly.

Gain: With the former ECC81 driver circuit the effective voltage gain from input to 4 Ohm loaded
output was 17 dB. Now, with the new fet-driver, this gain increased to 29 dB (R8 = 1 kOhm), or even
could be raised to 49 dB (R8 = 250 Ohm). This offers promising possibilities to try new advanced
negative feedback (NFB) techniques. See later for the improvements that we discovered.

Bandwidth: the F-3H frequency (- 3dB frequency range on the high frequency side) did not change
that much due to the dominant high output impedance of both tube- and fet- driver stages in
combination with the Miller capacitances of the power tubes. You hardly could see differences
between the Git80-EI and the new Git80-t. However, when we applied NFB techniques, the superior
specs of the toroidal OPT’s jumped to the foreground. Again, see later for the improvements we
discovered.

Distortion: there is little difference in THD between ECC81 and DN2540 drive circuits including their
spectral distribution. This is very promising for equal Tube-Fet sound quality; the older PR20HE was
liked for good reasons. Now, with fet-drivers we have promising tools available to maintain this liking
while improving further.

Subjective rating: in our years-end meeting (sept-11-2021) we subjectively evaluated all new amps in
direct comparison to the old PR20HE amplifier (which had a reference rating of 3.0 on a scale of five).
All new amps received a higher rating (environment 3.6 on the scale of five) indicating that
improvements clearly were noticed.

Conclusive: by changing tube-drive into fet-drive we won gain, kept bandwidth equal and maintained
the fine distortion behavior. Especially the larger gain offers advanced possibilities to bring the new
amplifier to real high-end qualities. But, as will be shown later, a lot of work had to be done to reach
that goal. In this article we discuss what we did to get equal to better results with fets plus the MSCF
updated power supply.

3. Trim-Pots and Cathode Feedback

When you measure the specs of both triode sections of an ECC81, you can hope that both triodes are
nearly equal, which mostly was the case. With two unselected DN2540-fets, you can count on spec
differences. Therefore I added R8 to trim the DC-balance of the fets, and R7 to trim their AC-balance.
See the “Start-up” section where I will give guidelines for a correct start-up and trimming of the new
amplifier.

We maintained the cathode feedback technique around the power tubes, because its open and
detailed soundstage are features that should stay; no reason to implement any change there.
 4. Details of the new TS2019-PP80 amplifier

The complete fet driver-circuit is surrounded by a red-dotted-line, indicating that this circuit is placed
on a small PCB (designed by Erwin Reins), as shown in figure-4.

The two coupling capacitors 220nF/350V and grid stopper resistors R21 and R23 and negative bias
voltage resistors R20 and R22 and the screen-grid resistors R16 and R19 are not placed on this PCB,
but directly connected between PCB and tube socket pins. These components could have been on the
PCB, but my students love to use huge Russian paper in oil capacitors and then the dimensions of the
PCB become unpredictable. Remark: place R21 and R23 as close as possible to the G1-pins (pin-5) of
the tube-sockets to prevent oscillation. See figure-6 for all the details.

Figure-4: PCB of the fet driver stage of TS2019-PP80 (7,0 x 5,5 cm)

5. The improved Power Supply

The original PR20HE had a very basic power supply with the VDV-Pow80 EI-core transformer (sold
out) and rectifier diodes and huge electrolytic capacitors.

The new amplifier has an improved toroidal power transformer Pow80-t. Several years ago I
discovered and published my MSCF-circuit which has great advantages. See figure-3 where the actual
MSCF circuits are inside the red dotted squares. To name some MSCF advantages: at the supply
outputs you don’t need extra electrolytic capacitors because of the low output impedance of the
MSCF circuit. Also, at switch-on, the supply voltage raises gradually; a nice ‘soft-start’.

In the circuit of figure-3 two extra 3 Watt resistors of 100 kOhm are placed, indicated in red. For
warmth/heat reasons, they are not placed on the PCB. They create a permanent current demand of
about 4 mA per MSCF-section, to make the fets Fp1 and Fp2 always conducting. These two resistors
are not shown in figure-6 for clarity reasons. Also this PCB is designed by Erwin Reins.

The heat production of fet Fp1 is several Watts. So, the warning to mount it on a cooling surface
(chassis) is given with good reasons. This fet can be placed at some distance of the PCB, when an extra
1 kOhm resistor is soldered directly at the Gate pin, to prevent any oscillation.
The supply can be switched with jumpers into two independent 380 V supplies and the negative bias
voltage can be raised to 130 V with another jumper. However, such is not needed for this PP80
amplifier. See figure-6 for the correct jumper connections.

The 2 x 8V supply is not needed now, only comes in function the moment any opamps are added to
the amplifier circuit. See the sections Phase Splitters.

Figure-5: The 14 by 9 cm power-PCB of the TS2019-PP80 (design: Erwin Reins)

6. Connecting the different sections

In the next figure-6 the wiring of the audio-section-PCB plus power-PCB and tube sockets is shown.
You might place the different parts at other positions. However, please lay all wires to their
connecting points as indicated. This drawing careful takes all currents into account, preventing the
large power-tube currents to enter the sensitive driver-section. Please also notice: the blue wires are
audio-ground!

The wires to mains and from Pow80-t to power-PCB are not shown. These are self-explanatory. Also
the 6,3 filament wires (indicated with f) are not shown. Twist them and connect to pins 2 and 7 of
each power tube socket. Don’t forget to place the two 100k-3W resistors. See also figure-5 for more
details.

The primary winding of the VDV-Git80-t has two unused UL-taps. So, the Brown and Orange wires of
the primary winding need proper isolation. Don’t cut them; you might wish to use them later. Also,
applying CFB and UL-FB at the same time is not wise, it will not benefit the sound quality. Isn’t there
an old rule saying: “don’t apply feedback chains on top of one another; gone will be your micro
resolution”. In my practice I totally agree with this old wisdom.
 Figure-6: wiring of the audio-section with an RCA-input.
Note that pin-1 (connected to the metal-base of the KT-tubes) is connected to “chassis”.
The white lines on the power-PCB show the correct jumper positions.

7. Symmetrical or a-symmetrical inputs

Figure-6 shows how to create an a-symmetrical input, cinch or RCA or whatever indication you wish
to give it (in Dutch the name is ‘tulp’). When you wish to use a true symmetrical XLR-input, see the
sketch in figure-7. What do I mean with ‘true symmetrical’? That the signal voltage at pin2 equals the
inverse sign signal voltage at pin3 with respect to ground pin1. Equal in the sense of magnitude and
exact inverse phase at least from 20 Hz to 20 kHz, within 0,001 degrees. These conditions are very
hard to realize. At the end of this article I introduce a nice phase splitter solution which was designed
during the TS-course. This splitter meets the stringent demands as mentioned here. But for now, if
your signal source delivers truly balanced output, you safely can use the solution shown in figure-7.

Figure-7: true symmetrical input.

You could use a simple double pole switch to convert from symmetrical to a-symmetrical. Why don’t
I apply that solution? Because the optimal phase inverting (of the two-fet differential circuit that is
used in this amp) must be trimmed with R7 (AC-balance) for either a-symmetrical input or
symmetrical input. Meaning, with every input change from symmetrical to a-symmetrical, you have
to trim the amplifier again. That is not smart. So, before you start with the next chapter, first decide
what you wish to use: a symmetrical or an a-symmetrical input.

8. Start-up: how to trim the amplifier to optimal operation?

There are four trim potentiometers on the driver PCB. Their functions are indicated in figure-8.

Figure-8 shows the trim-pots and their function

Next the start-up of the amplifier and the correct trimming is discussed in detail. This procedure
assumes that you know all safety rules when working in a high-voltage environment. The most
important rules are: don’t put a hand or both hands inside the amplifier when it is switched on. After
switching mains off, wait several minutes until the capacitors on the power-PCB have discharged to
below 30 Vdc. Remove the mains-cord when you have to connect meters or to solder or to change
anything inside the amp. If possible, don’t work alone; have someone nearby who is instructed
where to find the mains-off switch (or the plug to be pulled out of the mains-socket). Never touch a
person under high-voltage stress.

There we go:

1) Remove Fuse-1 and Fuse-2 from the power supply PCB

2) Switch mains to ‘on’ and check the filaments of the power tubes for glowing; if OK: switch
mains off. If not OK: check the filament wires. Switch mains-off.
3) Turn R9 and R10 fully counter clockwise
4) Measure with a DC-voltmeter between audio ground and pin 5 of the left power tube. Switch
mains-on: after a while the reading should be environment -65 Vdc. Repeat this for the right
power tube. Switch mains-off. Wait a while for discharging !!!
5) Place fuse-2: switch main-on; check with a DC-voltmeter for about 380 Vdc between pin H10
= 0,5B+ and audio-ground. The MSCF-circuit creates a slow raise of this voltage, so, take your
time.
6) Check on the driver PCB with a DC-voltmeter between cap-left-side (= H6) and cap-right-side
(= H7) and read close to 0 Vdc. Trim with R8 = DC-balance to exact 0,0 Vdc. Switch mains-off.
7) Place fuse-1 also; connect a DC voltmeter in parallel with the cathode resistor R17 = 1 Ohm.
Connect a second DC voltmeter in parallel with R18 = 1 Ohm. Set the range of these two meters
at 100 mV maximum.
8) Switch mains-on and check the B+ voltage. It should read around 760 Vdc in respect to audio-
ground. MEANWHILE check the two meters over R17 and R18. They should read 0,0 mV.
9) Now slowly turn clockwise R9 and read the meter over R17 until the reading is 50,0 mV.
Repeat this for R18 while turning R10 slowly clockwise. You will notice that the readings on
R17 and R18 will influence each other. Repeat until the readings are stable at 50,0 mV.
10) Check the B+ voltage again and you will notice that it is around 720 Vdc. Also measure 0,5B+;
its reading should be close to 360 Vdc.
11) Eventually repeat step 6). Next switch mains-off and wait for discharging.

The next section explains how to optimally trim the AC-balance:

12) Place the fuses-1 and -2; connect the meters over the cathode resistors R17,18 as indicated in
7) above.
13) Connect an 1 kHz sinewave oscillator to the amplifiers input. Set its output voltage to zero (to
be turned upwards later).
14) Connect a 4 Ohm dummy load between the black and blue speaker terminals. Connect these
terminals also to the oscilloscope input. Warning: the BLUE-speaker connection should go to
scope-GND=black terminal and the BLACK-speaker connection should go to the scope-
input=red terminal. (This is the only disadvantage of using the secondary as CFB-winding, but
Audio-Research can live with it; so can we).
15) Switch the mains-on, wait for the two meters over R17 and R18 show 50 mVdc and slowly raise
the oscillator voltage. Check the sinewave on the scope screen. It should be a nice sine-wave.
16) Meanwhile check the meters over R17 and R18. At larger sine wave input voltage their
readings should raise equally. If that is not the case, than trim AC-balance = R7 till this condition
is met up till maximum output just below clipping (now your second harmonic distortion is
minimal at frequencies above 500 Hz).
17) If you don’t own an oscilloscope, then you can trim R7 by using your ears. Trim R7 to the
position where you softest hear the soft singing of the output transformer core.
 Next trimming is for those who own spectrum analysis equipment

18) Apply 1 kHz sine input and trim R7 to minimum D2 at 2 kHz

19) Apply 40 Hz sine input and trim R17 (or R18) to minimum D2 (= 80 Hz). This measurement
checks if the push-pull quiescent currents create absolutely equal inverse-sign magnetic fluxes
inside the core of the OPT. When equal, the net DC-flux inside the core is zero (as it should be).
20) Do tests 18 and 19 at 1 W in 4 Ohm load at + 6 dBV = 2 Vrms output level.

You might wonder: can’t trimming be more simple? Yes, but then you don’t get the high quality level
that we are aiming at. Then you have to rely on additional negative feedback (NFB) to remove all
faults. However, the quality that you reach with careful trimming by hand surpasses NFB by
remarkably better microdetail reproduction. Preventing faults really is better than healing afterwards
with NFB (this is a pure Menno-rule).

9. Measurement results

The output power of the older PR20HE amp was 73 Watt, while the new Fet-version delivers 91
Watt. This difference is not a surprise because the secondary voltages of the new toroidal power
transformer are larger. This has nothing to do with fet- or tube- drive.

The -3dB high frequency (f-3H) of the old and new amplifier version is about 35 kHz. This is mainly
caused by the Miller capacitances of the power tubes. In the fet version f-3H easily can be raised by
applying extra source followers, but subjectively I felt none desire to make such addition.

The next figure 9 show the spectrum measurement of an 1 kHz signal with magnitude of + 6 dBV in 4
Ohm load. This is the 1 Watt level where we mostly listen to at home. Please notice the small 50 and
100 and 150 Hz residual hum-levels, which underline proper MSCF-functioning. Around the basis of
the 1 kHz test signal, some 100 Hz residuals cause intermodulation products, however with
impressive smaller amplitude than in the earlier PR20HE amplifier. Again, this has nothing to do with
fets versus tubes. It tells all about the MSCF-circuit.

Figure-9: TS2019-PP80 with Fet drivers at 1 W in 8 Ohm

Figure-10 shows the harmonic distortion at 1 Watt level as function of the frequency. This is an older
tube driven PR20HE measurement. I would love to make the scales equal to the next measurement
in figure-11, but I can’t because the old 2006 measurement file got lost on my computer.

Figure-10: Harmonic distortions as function of the frequency of the tube-driven PR20HE

Figure-11: Harmonic distortions of the fet-driven TS2019-PP80

Now compare figure-10 with figure-11 which shows the results of the fet-driven amplifier. First have
a look at 20 Hz: the tube-version distorts almost 1 % while the fet-version distorts 0,1 %. This clearly
shows the improved Git80-t compared to its older EI-version. Also compare D2 at 10 kHz and D3 at 5
kHz. The results are almost equal in both amplifier versions, underlining the almost equal sound
character that we noticed.

The next two measurements show the output impedance of the PR20HE (figure-12) and the TS2019-
PP80 (figure-13) as function of the frequency.

Figure-12: output impedance (at the 4 Ohm output) of the PR20HE with Git80-EI

Figure-13: output impedance (at the 4 Ohm output) of the TS2019-PP80 with Git80-t

This comparison clearly shows that the low frequency behavior of the Git80-t has largely improved
compared to the earlier EI-version. At higher frequencies around 20 kHz the EI version clearly shows
larger leakage inductance, causing Zout to become larger. As is known, in the toroidal Git80-t the
leakage inductance Lsp is much smaller. These measurements prove this feature.
Machine Translated by Google

In the next text (in Dutch; please use Google Translate) in chapter-10 I introduce a simple phase
splitter-1 for the TS2019-PP80 amplifier, in chapter-11 followed by the results of a theoretical Mathcad research for
a new and better one: phase splitter-2.

10. External Phase Inverter-1

I noticed with the TS2019-PP80 project amplifier that the two DN2540 control Fets at the input have to
perform too many tasks at the same time. They must sufficiently amplify the signal from the input and
also deliver 2 perfect signals that are the same size for all frequencies and exactly mirrored (in
antiphase), so that the power tubes are driven in perfect balance (push-pull).

The amplification is successful (factor 800 x if R8 = 250 Ohm), but the phase shift is defective and
depends on the frequency. I see this in almost all measurements because (in particular) the 2nd
harmonic distortion starts to increase above 1 kHz. The balance principle (of power tubes and output
transformer), which should perfectly suppress the second harmonic, does not work properly. The
counterphase signals produced are not equal in size and are offset (delayed) with respect to each
other. The higher the frequency, the more serious these abnormalities become.

To free the two fets from too much heavy work, I developed an external phase shifter, using two
opamps. See the circuit below.

Figure-14 : Phase shifter-1 for TS2019-PP80 project amplifier.

A short explanation: The first opamp inverts (so as not to suffer from the non-constant input
capacitance of the opamp) and amplifies exactly -1 x. To somewhat limit the frequency range in
the high end (this prevents overshooting of square waves), 22 pF is parallel to the 47 kOhm negative
feedback resistor, which yields an f-3dB of 154 kHz (high enough: f-3dB = 1 / 2*pi*RC)).

The output impedance of the left opamp is very low, which is necessary to properly drive the DN2540 fet
[so as not to suffer from the high (150 pF) internal gate-source capacitance inside the fet].
This signal is sent to the - input (the lower DN2540).

The second opamp inverts again (without frequency limitation, once at the beginning is enough). Its
output is also low impedance and that signal goes to the top DN2540.

The plus and minus 8 V supply voltage are already present on the power supply board, so that is not a concern.

The total circuit then looks like this:

Machine Translated by Google

Figure-15 : TS2019-PP80 with external phase shifter-1

In the working photo below you can see how I placed this phase shifter in Wim's amplifier.

Figure-16 : New phase shifter-1 in Wim's amplifier.

The tests and measurements showed that the circuit rotates the phase much cleaner. The
measured distortion decreases considerably. So on that note.....

However, the 2nd harmonic still ramps up above 1 kHz. A little less, but still. Could I have
overlooked something, not taken into account something, forgotten something? Yes,
unfortunately. But what then? See below.

Later I will describe some experiments in Wim's amplifier with different negative feedbacks, in
which this phase shifter-1 played an important role, even though it has flaws. Later it was
replaced by the better phase shifter-2, of which the print data is available. Phase shifter-2
is recommended, is the best.
Machine Translated by Google

Many things are now much better under control, if only for the better adjustment of minimum
distortion at 1 kHz sine @ 1 Watt into 4 Ohm load.

11. External Phase Inverter-2 :

The 'mistake' I made in phase shifter-1 is that I didn't take into account the delay that takes place
internally in an opamp. The transit time of the signal in the opamp is not infinitely small, so the opamp
internally shifts the output signal relative to the input signal (depending on the GBP of the opamp,
see below).

By building up the phase shifter in a smarter way, you can compensate opamps externally for those
internal time delays and ensure that the + and – outputs get exactly the same time delays (so keep in
step with each other).

Below I show the results of a Mathcad program that I used to calculate this.

150 200

100
Aoj

50 100 ÿni j
Aij

0 ÿi j
Anijÿ1
ÿ 50 0 diffÿ j
Adif j

ÿ 100

ÿ 150 ÿ 100
1 10 100 1 103 ÿ 1 104 ÿ 1 105 ÿ 1 106 ÿ 1 107 ÿ

fj
Figure-17 : Left vertical axis (in dBV/V): the red descending line shows the open-loop gain Aoj of the opamp.
(The actual Ao of the OPA1656 is almost the same up to over 100 kHz).
The bottom dashed purple line Adiffj shows that the two outputs differ less than -130 dB.
Right vertical axis (in degrees): The black dashed line difÿj shows that the phase difference between both output
signals is zero, even though the phase of both signals rotates (see ÿnij and ÿij ).
Horizontal frequency axis: this is in [Hz], so from 1 Hz to 10 MHz.

How do I handle this? Under stringent conditions I send the audio signal simultaneously through two
equal opamps. The upper one is non-inverting (connect to the + input of the opamp) and the lower one
is inverting (connect to the – input of the opamp). Since the two opamps are of the same type, their
internal delay times become nearly equal.

The OPA1656 I'm using here has a GBP (Gain Bandwidth Product) of about 50 MHz (extremely
fast). With a small trim capacitor of a few pF, the minimal remaining difference in delay times can
be smoothed out.

A real balance input stage has been added to the test print that Erwin is now making. We will soon
test and if everything works out, publication will follow.

Behold: the solution is on the horizon!

Machine Translated by Google

I remember telling you from the beginning of project TS2019-PP80 that the phase shifter was going to
be the bottleneck. Practice now also shows this. I then told that an input transformer of my hand
could solve the problem. However, such a transformer, with associated development
costs and long development time (= project delay) and small sales market, is relatively expensive.
That's why I'm still striving for a solution with opamps first.
[But there may be a transformer customer on the way who wants such a transformer and then the
development costs will be with him/her. That makes things different, depending on rights, obligations and agreements].

Why did I know from the start that the phase shifter would cause trouble? Because of my
experience that nothing is perfect and also because I asked Frank to perform a Spice simulation of
the TS2019-PP80 design. I then received the following response: “Menno, what a great amplifier; how
do you just get -150 dB distortion while designing in your head?” To which I replied, “Wait until it is
built; practice and Spice are not each other's friends”. Frank had programmed an ideal phase
shifter and when I asked him to program an ordinary phase shifter (the kind you build from separate
parts), that -150 dB distortion was not much left. Well…

I had already announced in my previous post (November 23/24, 2020) that phase shifter-2 is
coming. The circuit is now complete and will be discussed here.

The complete schematic of phase shifter-2 is below. It is a bit more complex than the previous version.
That's mainly because I added a balanced input amplifier.

Figure-18: The complete schematic of phase shifter-2

The balanced input is the circuit around U1-a. The ALPS volume control pus U1-b together provide
a clean volume control. The actual phase shifter consists of the opamps U2-ab. First I discuss the
ALPS volume control.

11-1: Section-1: The optimized ALPS volume controller.

The diagram below only shows the volume control. The audio input is connected directly to the top
of the ALPS pot.
Machine Translated by Google

The volume control has a value of 10 to 100 kOhm, depending on the type you purchased.
From 'audio-in' to the plus input of opamp U1-b, the volume control can be represented
by a resistor divider whose effective resistance depends on the position of the volume
control.

I drew an extra two green capacitors: Cin-a and Cin-b. These together form the input
capacitance of the op amp. In a regular Jfet opamp, Cin-ab are NOT constant, but vary in value
with the instantaneous magnitude (AMPLITUDE) of the audio signal.

Together with 'the resistance' of the potentiometer, these capacitors form a low-pass filter,
whose -3dB high-frequency (f-3H) is not constant, but changes with the magnitude of the
audio signal. f-3H is around 300 kHz, depending on the ALPS resistance.

So you get a low-pass filter whose high f-3H moves with the size of the audio signal.
This causes a non-constant phase shift in the audio range, which you experience as
unrest in the sound image. John van der Sluis discovered this many years ago and described
it in his magazine Audio en Techniek.

As far as I know there is currently only one opamp (the OPA164x series) that has constant
Cin capacitances. With all other opamps Cin is audio voltage dependent. This is the explicit
reason why I apply the OPA1642 here. The ALPS potentiometer is now loaded with a high
impedance (so that it accurately follows its logarithmic control curve) and there is no longer
any variable filter influence, so that the phase unrest disappears from the sound image.

Can you hear that? Of course. I have never heard such a clean and pure and quiet signal
with an ALPS potentiometer.

'The potentiometer rises a class in quality'. I put this in parentheses because nothing has
changed in the ALPS. It is now loaded very correctly and constantly, so that the
disturbing phase distortion no longer occurs. At the output of U1-b, the signal is low-
impedance and therefore insensitive to whatever you connect to it. The ALPS no longer notices this.

Below you can see the graph with which the manufacturer shows the constancy of Cin-ab.

This constant Cin is a great gift, making potentiometers work optimally.

You can also minimize the phase effect mentioned above by using 1kOhm potentiometers.
This is very beneficial for low noise, but it requires a lot of current capacity from the driving source.
Machine Translated by Google

11-2: Section-2: The phase shifter.

In the diagram below I have redrawn U1-b, because the output impedance of U1-b is an essential part of the proper
functioning of this circuit.

In my earlier post I told something about the calculation of this circuit. In short, it comes down to this: in U2-b
and in U2-a the signal is slightly delayed. For correct phase rotation, the delay times in U2-a and in U2-b must be
exactly equal. You can set this precisely (trimming) with the adjustable capacity C10. I'll tell you how to do that
in practice later.

U2-b amplifies +1 times and U2-a amplifies -1 times, provided that R8,9 are exactly equal and the output impedance
of U1-b is negligible with respect to R8 = R9. That's why I redrawn U1-b and I repeat what was said earlier with the
volume control: the output impedance of U1-b is very low impedance (do the math, you might be shocked
by what's happening there). For our application, all conditions are met and we only need to consider R8.9.

You could now say, “the OPA1656 has no constant Cin; is it bothering you now?"
No, because U2-b is driven directly by the low output impedance of U1-b, putting the varying filter
frequency f-3H in the neighborhood of 7 GHz. You will no longer notice this in the audio range up to 20 kHz.

Similarly for U2-a: inverting; so virtual earth; thus variation of Cin does not play a role. Is that really true? No, it
is not true if you measure the distortion at 20 kHz. I've hinted at this before, due to an unexpected increase in H2.
However, for our application such micro effects play a negligible role because the tube output stage itself already
distorts more H2 at 20 kHz.
This new super simple circuit is able to keep the phase shift constant at 180 degrees up to 1 MHz. You will never
achieve that with transformers.

A final note: I very consciously chose the OPA1656 because it has a GBP (= gain bandwidth product) of 55 MHz.
This opamp is extremely fast, so you can make the circuit work perfectly up to 1 MHz with a very small
correction (with C10). In addition, this opamp has an extremely low natural distortion. It really is a wonderful thing.
Machine Translated by Google

11-3: Section-3: The Balanced Input Amplifier.

• This circuit is a standard (text book) circuit. He reinforces the tension between the
in+ and in- inputs 1 x.

• The input impedance is approximately 2 x 10 kOhm (R1,2). This value is one

balancing between low noise (1 kOhm) and large enough input impedance (100
kOhm) to minimize strain on tube signal sources. In the end I opted for a modern
intermediate value of 10 kOhm.

• The frequency range is limited at the top by C3.4 and the corresponding f-3H is located
constant at 194 kHz.

• The input is decoupled for DC voltages with C1,2 and the associated f-3L is at 0.6 Hz. It
is important to use non-polar (also called bi-polar) types for C1,2 (because they have
minimal audible influence when combined with f-3L<< 20 Hz).

• In this circuit, the constant of Cin is an essential condition (the common mode voltage is
not equal to zero), which is why I also deliberately chose the OPA1642 here.

• (But let's be honest: U1-b was supposed to be the OPA1642. The OPA1642 contains 2
opamps, so I had one left. To please everyone, I used the remaining opamp U1-a to
make a balanced input stage. to make).

• Is this circuit the best balanced input amplifier? No, it could be better with
precisely equal-trimmed resistors R1,2,3,4 or with a professional differential
amplifier (requiring a total of 3 opamps). However, for our TS2019-PP80
application, this amplifier is more than good enough.

• (I will indicate later on how you can bypass this circuit or use it smartly for an asymmetrical
RCA-Cinch-Tulip input).
Machine Translated by Google

11-4: The circuit diagram and the PCB and the connections and the SOIC8 ICs

I asked Erwin to design the print and after cycling back and forth a beautiful print was created.
Erwin then starts with a different kind of scheme: a logistics scheme in which the most logical
connections are chosen (from the point of view of the print layout). See below:

Figure 11-4: Phase shifter-2 logistic scheme

With this scheme, Erwin started swimming laps and doing puzzles. I had set the additional
requirement that the OPA1645 and OPA1656 ICs must be soldered directly to the PCB (SOIC8
sockets). See the beautiful result below.

Figure 11-5: PCB of phase shifter-2

If you look closely at this you will see that Erwin has added extra 0.1uF/25V capacitors from pins 4
and 8 of each IC to ground (C6,7,8,9; necessary to prevent IC oscillations). https://www.digikey.nl/
products/nl?keywords=81-RDER71E104K0K1H3B

The trim capacity C10 is now also clearly visible; see link below for details of C10. https://
www.digikey.nl/product-detail/nl/sprague-goodman/GYA15000/SG3002-ND/65241

All resistors are ¼ Watt metal film. If you don't use the balanced input, don't put R1,2,3,4,5,6 and
C1,2,3,4. All these components are located to the left of the ICs. Not mounting it causes no problem.
Machine Translated by Google

11-5: Adjusting C10.

• It is best to do this before you put the PCB in your amplifier. • Use
two separate 9 V batteries for the power supply. • Set the
sine wave generator to 1 MHz with an output voltage of approximately 1 V. •
Use a two-channel oscilloscope in XY mode. For example, the Picoscope-2000 series
fits very well here. Thanks to Frank, it is now connected to my measurement computer
as standard and I can easily paste pictures.
• I have put the adjustment diagram below in a drawing.

Figure 11-6: adjustment setup of phase shifter-2

Adjustment: Rotate C10 until the XY scope image is a slanted straight line (not an oval).
The setting is one-off; after that you don't have to change it anymore.

What if you don't have this measuring equipment? Consult with your fellow students via TS-
mail and make arrangements to help each other. If you come to the TS clinic, I can do the
adjustment for you in a curse and sigh.

11-6: Connecting an unbalanced RCA RCA RCA input.

See the connection diagram below. The section U1-a is not used now.

Figure 11-7: unbalanced input connection

Machine Translated by Google

11-7: Connecting a balanced input that you can also switch unbalanced.

Figure 11-8: Switching between balanced and unbalanced

11-8: Connecting to the main board and preferred setting of your PP80

Figure 11-9: Phase shifter-2 wiring

• Remember that R2 on the control board must now be 100

kOhm. • Remember that R8 on the control board must now be 1 kOhm (trim pot Trim M1
= M2) • And what else I want to say ……………………………
• Subjective with regard to phase shifter-1: cleaner, more space and details, sss and ttt and more beautiful

12. Study on different forms of feedback.

In addition to cathode negative feedback, can you add additional negative feedback?

The above is about the amplifier that each of you is now building or has ready. So I don't have to do
any extra tests in that area, but I do want to say something about it.

In addition to the cathode negative feedback, you should NOT apply additional global negative
feedback. If you have cathode negative feedback (as we do now), then you don't have to
make an extra negative feedback on top of that. The sound image quickly turns dead. The
sound becomes flat and totally uninteresting. So don't do it.
Machine Translated by Google

How do I know: from my 1987 Radio Bulletin draft of the VDV100 power amplifier. There I did
anode negative feedback and global negative feedback. The two fought each other (although
I didn't realize that at the time). Until I chose -or- anode negative feedback -or- global
negative feedback. Then the soundscape opened up more. Later tests with cathode AND global
negative feedback showed the same picture. You do -or- one or the other, but not at the same time.

That is why I advise everyone to choose R8 = 1000 Ohm, as long as you only apply cathode
feedback. If you still want some negative feedback, choose R8 = 500 Ohm. Then you can
apply a little extra global feedback from output to input, but please not too much. Just play
with it and then set what YOU like sounding.

R8 = 250 Ohm is only for freaks like Menno, who still want to play with global negative
feedback, although it is known in advance that this is not an easy route. Despite that, why not
try again what massive global backlash could yield?
See some of the results below.

In the basic design of the TS2019-PP80 amplifier, I very deliberately chose cathode feedback,
so that the damping factor (at the output) can be reasonably large without great effort, so that
you do not have to listen to 'blubber' bass sound. The distortion at cathode negative
feedback is also fairly low, because the power tubes are quite linearized by this local negative
feedback.

NOTE: In the diagrams that follow, the cathodes are disconnected from the secondary winding.

So the cathode negative feedback has been removed. The cathodes of the power tubes now
go directly to ground via 10 or 1 ohm resistors.

12-1: Local feedback from the primary to the source of the DN2540 :

The circuit below is driven by the phase shifter-2. From the anode of the power tubes
there is negative feedback (per balance half) to the source of the driving DN2540 fet. Note that
the output transformer is NOT inside the negative feedback loop. We are heading towards
Trans here, but now with extra amplification in front of the power tube (LinearAudio: issues 5 and 8; MTA).
Machine Translated by Google

Figure-12-1: Local negative feedback BEFORE the output

transformer, from the anodes of the power tubes to the sources of the driving DN2540 Fets.

The technique I apply here has been known for a long time. In older designs, this is then coupled to
the cathodes of the driving steerer tube (ECC81 for example). Remarkably, this negative
feedback works well with coupling capacitors (Trans does not), provided they have the correct
value. See again also that R8 = 250 Ohm.

The negative feedback is 125 Ohm (half of R8) compared to the red 100 kOhm negative feedback
resistors. The high voltage of the anodes must be withheld, and that is done half by two 10 uF/450V
electrolytic capacitors connected in series. Parallel to the electrolytic capacitors, resistors of 1
MOhm are mounted to divide the high voltage equally between the two electrolytic capacitors.
See photo below. Make sure that the + side of the electrolytics points towards the anode.

Figure 12-2: Series of electrolytic capacitors in local negative feedback branch.

I repeat one more time: the cathode negative feedback is gone. There is negative feedback per
balance half from the anode to the source of the driving fet. The OPT is not inside the negative feedback loop.
This is very similar to Trans. What is the end result? See the distortion measurement below:
Machine Translated by Google

Figure 12-3: TS2019-PP80 with local feedback to Fet sources; at 1 Watt into 4 Ohm

First a few notes:

• At D5 and D6+ you see a steep ramp on the right. This is caused by the fact that I measure
(sample) with 192 kHz, and therefore effectively measure up to 90 kHz bandwidth. The input
amplifiers of my E-mu 0202 sound card then no longer process the H5 and H6 properly. To ignore!
• On the left you see food junk. Do not pay attention. This is about experiments.

Now look at the distortion numbers (bottom left of the graph) at 1 kHz: do you see how low
the distortions are already? This all seems great. Yet …..

The practice is that it measures reasonably well, but that this circuit is rejected by me in
advance because:

• feedback is fed back to the source, while the audio signal is applied to the
gate. The fet ultimately amplifies the signal between gate and source; so the input
signal minus the negative feedback signal. This is exactly according to the theory. But
this subtraction is by definition NOT linear, because the fet itself has its own distortion
and adds it to this subtraction. This distortion is strongly addressed because the fet AND
on the gate AND on the source receive fairly large signals, of which the fet must then
determine the difference.
• Conclusion: it has to change. This seems like a good route, it measures very nice, but
there is a serious error in advance.

After trying it, I'm rejecting this approach and going to do better with the next circuit.
Machine Translated by Google

11-2: Local negative feedback from the primary to the input of the DN2540 :

I again fed back from the anodes, but this time to the inputs of the fets. In order to make the
phase correct, I have applied the negative feedback crosswise (it is about negative feedback
and not positive feedback). The Source now remains at rest and the Gate is only presented
with the small difference signal.

Again, NO cathode feedback is applied and R8 = 250 ohms.

Between the driving phase shifter-2 and the inputs I added extra 1 kOhm resistors.
The negative feedback is therefore the negative feedback resistance of 470 kOhm relative
to these 1 kOhm resistors.

The necessary subtraction of input signal and feedback signal now only takes place at the
gate of the fet (and not between gate and source), so that the gate-source a-linearity of
the DN2540 is hardly used. This is serious cake and I will go into this in more detail.

Figure 12-4 : Local negative feedback BEFORE the output transformer, from the anodes of
the power tubes to the inputs (gates) of the driving DN2540 bikes. See further details in the text.
Note: In Jacques van Eijk's report this circuit is indicated by fig.8

Figure-12-5: TS2019-PP80 with local negative feedback to fet gates; at 1 Watt into 4 Ohm
Machine Translated by Google

Now compare the results of Figure 12-3 and Figure 12-5. Then you see that I have
rightly rejected the negative link to the sources. The same applies to feedback to cathodes in
tube circuits. The results of Figure 12-5 are much better.

So: do not couple between source and gate. Counter-couple in such a way that the input fet on
its gate only sees the difference between Vin and beta*Vout (which produces a small gate signal,
resulting in significantly less fet distortion).

12-3: What do I do with these excellent results?

Nothing at the moment. My schedule doesn't allow me to work this out in full, get all the
bugs out, guarantee high-frequency stability, and so on. And what about the sound effect of
the coupling capacitors? All things that require careful study. To work something like this out
properly I need a lot of time (if I can schedule it at all). I did listen to one channel for a while,
and that made me very happy. A promising route !!!!

For the latest results, see Jacques van Eijk's report, who successfully applied this circuit and
saved me an enormous amount of work.

13. Closing Remarks

In this article you find no pictures of the many PP80 amplifiers that were built by my students.
Please see the reports of my students in the section TubeSociety  Projects  TS2019