Class D & E amplifiers

In a class D amplifier, a pair of transistors switch on and off, out of phase across an O/P transformer.
Efficiency of 75% can be expected but class D amplifiers are relativley complex.
Class E amplifiers are about 90% efficient.
Because of the low loss no cooling fan is needed, no TR switch is nedded, the received signal is piped through the amplifier itself.
The Class D & E amplifiers were invented and patented by Nathan Sokal and Alan Sokal (Father/Son) in 1975
They minimize heat loss by having as little overlap as possible between voltage and current.
In class D and E amplifiers,the devices act as switches, half time completely on, half time completely off.
But transistors are not perfect switches; Mosfets have a resistance of 1 ohm (on) and several hundred Pf when off.
Losses are greatly reduced in switching amplifiers but the penalty that the O/P power no longer depends on the drive power, but rather on the supply voltage.
This means that switching amplifiers are not linear amplifiers and therefore not suitable for SSB without aditional limiting and modulating circuits.
However they are fine for CW, FSK and FM.

From designs cited in May 1997 QST

POWER 300W 500W
Device (I.Rectifier) IRFP440 IRFP450
Vds 500v 500v
Id 8.8amp 14 amp

Class 'E' transmitter for 1.8MHz (2 design's) - "click to enlarge"


The following is reproduced from "evolution of the damper diode" in television systems
an original work by Dr Hugo Holden (see

IMHO it surpasses any other textbook explanation of how this type of circuitry really works.

The first person to suggest the use of a "damper" diode in magnetic deflection scanning, in 1936,
was Alan Dower Blumlein, the "inventor" of stereo audio.
He patented "binaural" audio recording in 1931.
Alan Blumlein was killed in a plane crash in 1942 while testing radar for the war effort.
This loss was described by Winston Churchill as a national tragedy.
Following this, damper diode function was well examined by RCA laboratories presumably during wartime and in the immediate two years thereafter.
RCA produced a series of review articles and in "Magnetic Deflection Circuits for Cathode-Ray Tubes" Otto. H. Schade Volume V 1947-1948 Pg 105, RCA Labs, reference is made to Blumlein's original 1936 patent, and damper diode technology is thoroughly explained.
The circuits have been reduced to their basic forms without linearity or width controls so as to show their basic configuration.

Coupling the yoke to the line output tube by a transformer is shown in Fig 1.
At flyback the tube is cut-off and the magnetic field in the transformer and yoke collapses and resonates due to the self-inductance and distributed capacitance of these structures.
There are oscillatory voltages and currents representing relatively undamped oscillations.
These oscillations, which are visible in the scanning raster, decay away, and become damped out when the line output tube is again driven into conduction by the drive voltage.
These oscillations must be eliminated for a satisfactory scanning raster.

Fig 2 shows resistive damping.
In this case the damping occurs across the entire duration of the sawtooth current scanning waveform on both the positive and negative excursions of current, so it can be called bidirectional damping.
This is wasteful of energy, lengthens the flyback period, and reduces the opportunity to utilize the positive going high voltage spikes generated at the line output tube’s anode, or via an OVERWIND coil to generate EHT.

Fig 3 shows an improvement to resistive damping.
This technique is used in the HMV Marconi 904 (1939).
The RC network is frequency selective, damping the parts of the waveform with the highest rates of change.
This reduces the oscillations of currents (shown in red) however, because the flyback period contains high frequency (Fourier) components, this is also damped.
Again this wastes energy and lengthens the flyback period.

Fig 4 Shows what might appear to be the introduction of an efficiency diode in the RCA TRK9 (and TRK 12) but is in fact, not.
This circuit has the damper conducting only over flyback time alone, and is really a spike suppressor.
A true efficiency diode conducts during the active scan time on the left hand side of the scanning raster and recovers energy from the magnetic field of the yoke and line output transformer.
The recovered energy is stored in the magnetic field at the end of scan time at the right side of the raster.
The circuit of fig 4 damps the flyback voltage oscillations and absorbs energy when the output tube is cut off.
This arrangement can’t be used in a system to generate EHT from the flyback voltage spike. In 1938 the Baird/Bush TV and radio company in the United Kingdom were using the circuit shown in Fig 5:
(Provided by Mr Victor Barker (VK2BTV AUSTRALIA)

This is probably one of the first examples of energy recovery scanning.
When the magnetic field in the line output transformer collapses, the diode conducts on the first negative half cycle of voltage on the diode’s cathode, to produce a more linear rate of change of current.
This damps the oscillations and also returns energy to the power supply.
As can be seen this was the precursor of the typical transistorised line output stage that appeared in early transistor televisions in the early 1960’s.

Returning to this later, let's look at this Bush circuit in the following three equivalents: Rather than returning the anode to ground (zero volts), it can be returned to B+ provided B+ is cancelled to zero volts (or close) by another “-B+” supply as seen in Fig 5A. This added supply can then be replaced with an RC network, as seen in Fig 5B, which charges to a value Y, say close to the value of B+ but in practice is a little less as the line opt tube anode voltage doesn’t go completely to zero during active scan time.

Then simply this diode and RC network is placed on a secondary winding, not the primary, and the position of the diode and RC network reversed as they’re in a series circuit then you end up with the following seen in Fig 5C:

This circuit, although looking a little similar to that for the TRK12, is in fact quite different.
Observe the transformer polarity.
As will be seen below this is in fact the basic circuit used in the RCA 621 TS, except that the voltage generated across the capacitor is added in series with the B+ voltage to create what we now know as B+ boost voltage.

This same basic circuit, with the diode and RC network is also shown in Fink’s Principles of Television Engineering 1940 page 152, fig 87, placed in the primary circuit

When the line output tube is cut-off at flyback, the first half cycle of voltage oscillation takes the damper anode negative (cutting it off during flyback).
The damper anode has the opposite polarity to the anode of the line output tube. Then on the first positive half cycle of the voltage oscillations at the damper anode, the damper conducts.
This damps the oscillations and results in a near linear scanning current, at the left side of the raster, as the magnetic field in the yoke and transformer now collapse in a controlled (damped) linear way toward zero.
Before the current reaches zero however, the line output tube is driven into conduction and the process repeats.

The voltage you see across the transformer or yoke's terminals represents the voltage across the capacitive component, which lags behind the circuit current by 90 degrees.
When the output tube is cut-off, the circuit current, during the flyback period, is associated with a negative peak voltage on the damper anode and a positive peak on the line output tube's anode.
These peaks occur in time within about the middle of the 10.16usec fly-back interval (American system).
At the time of this peak, the yoke's current value is zero (but has its greatest rate of change) and the rate of change of voltage on the diode's anode, although at its peak, is zero at this time.
After that the secondary voltage returns to zero, at the end of flyback, and the current is at a negative maximum, now with the beam at the extreme left of the raster.
Then as the voltage at the damper anode attempts to oscillate in a positive direction, at the damper's anode with respect to its cathode, the damper diode conducts, damping the oscillations and resulting in a more linear current at the beginning of active scan time on the left side of the raster.
Moving on to the post war period we find Fig 6 below which really represents the Baird/Bush concept 1938,or the basic function outlined by Fink 1940, in the format seen in Fig 5C.

Damped current charges capacitor Cb and provides energy to a load R. Cb charges up and lifts the cathode potential of the damper diode.
This means that the plate potential has to rise to a higher value to establish conduction.
This helps ensure that the diode is not conducting until the start of active scan time, so there is negligible damping during the fly-back period.
This system is “recovering energy” from the magnetic field of the yoke and transformer which was stored at the end of active horizontal scan time, and delivering it to a load.
The load resistor can now be replaced with the primary circuit. This is shown in Fig 7 below

This basic circuit was used by RCA in the 621 TS, and this, or a modified version of it became the "Modern Standard" for line output stage deflection, using tubes, ever since.
Cb's negative can either be returned to ground, or to B+ as shown, which is at ground from the AC perspective.
The recovered potential energy generated by the magnetic field of the yoke and transformer, which was in fact provided by the primary circuit at the end of the scan (right side of the raster) is used to generate a boost voltage to help supply the primary circuit.
This gives a larger primary supply potential, the B+ Boost voltage, which helps attain the required picture width from a smaller B+.
It should be pointed out that, as is always the case, no additional energy is created that was not already supplied by the power supply in the first place.
The circuit is simply more efficient because overall, the damped current is not primarily wasted as heat, which it is in all cases of resistive damping.
One definition of a resistance is an energy wasting or heat dissipative device.
Moving on the Fig 8, we can see what happens if we simply re-draw the above circuit with Cb connected to ground.

This circuit, as deployed in the 621 TS, with small modifications is the basis for modern tube line scanning.
At turn on, DC current flows via the secondary winding and the damper diode to charge Cb to B+ potential and to initially supply the B+ to the primary circuit. During operation the voltage across Cb charges to B+ boost.
Therefore Cb needs to be rated to handle this higher voltage.
The circuit is however “inconvenient” in that the transformer cannot be configured as an auto-transformer.
It is a small modification to introduce B+ directly at the anode of the damper diode and then the circuit comprising the secondary, damper diode and Cb can be rotated around to create the circuit of Fig 9 below.
The circuit of Fig 9 has the advantage that the Cb only needs to be rated to handle the Boost component of the total “B+Boost” voltage, rather than the total amount.
Also the primary and secondary can be one tapped winding, with the yoke coupled across any part of it, in an efficient autotransformer configuration. ADMIRAL used this basic configuration in the early 1950’s, for example in their series 23 chassis.
In general, by the time efficient energy recovery line output stages had arrived, it had become the custom, as it is in the 621 TS, to derive the EHT from an OVER-WIND linked to the plate circuit of the line output tube shown in red in Fig 9, and the heater supply for this EHT diode derived from a small number of well insulated turns on the output transformer.

Other variations of damper diode circuits which have occurred in the post war period include a triode pair used as a controlled damper diode, which gives additional control over the linearity of the saw-tooth scanning current. See fig 10 below:


In general, in the line output stage, every effort is made to keep the resistances of the yoke and output transformer windings as low as possible.
In the case of transistor output stages (as shown in fig 5), where the working impedances are lower (lower ratio if dynamic voltages to dynamic currents), the nature of the line output transformer is such that the inductance dominates and the resistances of the coils are very low. This enables the output stage to operate as a saturated switch.
The transistor is driven with a step function, or rectangular wave, to cause it to switch on for 2/3 or more of the active scan time leading to the right side of the raster.
The rise in current in the yoke when the transistor is switched on is linear because the rate of change of current dI/dt is close to a constant for a period after the voltage is initially applied. The fundamental differential equation for an inductor is V = -L.dI/dt, or voltage equals a constant (the inductance L) multiplied by the rate of change of current with time. So if a rectangular voltage is applied to an inductor, the initial rise in current is a saw-tooth, ideal for scanning.
The negative sign in the equation indicates that the EMF of inductance, is reactive to, or opposes the applied EMF.
Of course the current cannot rise linearly indefinitely in an RL circuit, either it will become non-linear as the transformer core starts to saturate, or exponentially taper off to a value given by the applied voltage divided by the resistance.
The proportions of inductance, applied voltage and winding resistances are chosen so over the time interval of the scan when, the transistor is turned on, the rise in current is substantially linear.
At the end of scan, the transistor switched off and the magnetic field collapses for a half cycle of operation, this then forces the semiconductor damper diode into conduction, where it effectively acts like a switch and results in a substantially linear scan on the left hand side of the raster.

The transistor line output circuit, however, is therefore such that the transistor, acting as a switch, is very efficient, but the transistor, and its drive waveform can have little effect over the linearity of the scan it generates (unlike a tube line output stage as will be explained).
To gain linearity control, in the transistor line output stage, usually a capacitor is inserted in series with the yoke (sometimes called an "S" correction capacitor), or an inductance in apposition to a permanent magnet is placed in series with the yoke.

In contrast the impedances in a tube line output stage (ratios of dynamic voltages to dynamic currents) are higher than in the transistor case.
These lower currents (and higher voltages) in the primary circuits require that there are more turns in the output transformer at least, and usually the yoke too.
Horizontal yoke winding resistances are in the order of 10 to 60 ohms for tube work, but very low, less than 1 ohm sometimes in portable transistor TV’s with 12-volt supply rails.
Overall though, in a tube set, the line yoke coils are transformed to the anode of the line output tube as a substantially inductive load, and the anode voltage wave, with the saw-tooth grid drive is substantially rectangular in character.
However, in the case of the tube line output stage, the nature of the grid drive (horizontal drive), unlike the transistor case, is able to influence the linearity, especially on the middle right hand side of the raster.
(The line yoke can also be driven directly from the tube anode; in this case the yoke has a higher inductance and higher resistance than in the transformer coupled case)

In general linearity controls in tube line output stages are introduced as an inductor in series with the B+ boost supply to the primary winding, with varying amounts of capacitive filtering around this for the B+ boost voltage.
The ripple voltage generated alters the output tube’s working load and variations in linearity can be obtained that way, which tends to vary the linearity near the picture centre.

In contrast, due to the much lower working frequency in the frame circuits, transformers with larger inductances (and also resistances) are required and the load presented to the vertical output tube is a combination resistive and inductive.
This requires an overall drive voltage which is a combination of a rectangular wave and saw-tooth (trapezoid) to generate a saw-tooth current is a circuit with series elements of L and R.
The design of tube or transistor vertical output stages very much resembles their audio output stage counterparts, and the exact shape of the trapezoidal waveform and the operating conditions of the output device, has a substantial effect on the vertical linearity.
This is why one common form of linearity control in a vertical output stage consists of a variable resistor in the cathode of the vertical output tube.


Fighting (side effects of) Macrovision

This is an account of some modifications I did to an ancient (circa 1980) Sanyo TV of ours to fix an annoying side effect of the so-called "Macrovision" protection of rental VHS video tapes.

NOTE: If you came to this page as a result of looking for Macrovision removal methods (eg: to enable your DVD player to run to your TV set via your VCR without having the picture go dark every 30 seconds or so), what you really need is a video timebase stabiliser.

This page doesn't provide that. You can buy kits for such items (for the Oz and UK PAL TV standard) from your nearest Jaycar Electronics shop and make your very own in one evening (if you aren't frightened of soldering). Just type "stabiliser" into Jaycar's search box and then look for "Dr video Mk II". They're about $100.

Okay - back to the story of our old Sanyo!

This old warrior has been kept in service over the past 22 years for the simple reason that I fitted a nice big new 63cm (25") tube back in 1985 after the green cathode of the original had gone low in emmision.
(Well, re-gunned actually ... you would give Thomas Tubes here in Melbourne your old CRT, and they'll supply you a re-gunned one of the same type for around $150.)

Anyway, in June of 2003, this old Sanyo had gone "bang" and stopped working completely.
I learned that when the set was later turned on once again, there was now a strong, acrid burning smell in the room and no picture or sound at all.

An old original electrolytic capacitor (4.7uF 350 volt) had shorted out, plus incinerating a 1 ohm series resistor which obviously hadn't enjoyed having 120 volts across it as a result (see Fig.2).
These were part of the overall 120 volt feed to the horizontal output stage.

An 8uF 450 volt electro and a 1 ohm 5 watt resistor soldered in, and sound and picture returned as good as new.

Okay - so far, pretty straightfoward.

At this point, I wondered to myself
"Should I just take the set back to the living room now, or try to do something about it's Macrovision susceptibility?"
(see Fig.4) Because, whenever we'd play rented tapes over the years, every 30 seconds or so, we'd get all these bright white dashed lines coming up over the top half of the picture for 9 or 10 seconds. A most annoying effect.

Finally, I decided to have a shot at getting rid of the problem.
Basically, it seems Macrovision sneaks in a series of high amplitude white-going pulses during the vertical retrace interval. Aha, I thought,
"okay then, let's have another look at that Sanyo circuit
(see Fig.8).
What could be wrong with its vertical retrace blanking?"

One guess.

Sure enough - there was no evidence of any CRT-retrace blanking circuitry at all! One could only assume that more recent (or better designed) receivers must include such blanking, because our old chassis was one of the few on which I'd ever noticed this effect. So - how to get rid of the problem?

It seemed obvious that what was needed here was some additional circuitry that would clamp the video signal down to black (or below) during the vertical (field) retrace interval.
I quickly sketched up a circuit for a single NPN transisor, some resistors and a pair of coupling capacitors, grabbed the components out of the drawer and soldered myself up a little rats-nest for the job. I then connected the transistor input circuit up to the vertical output sawtooth (via a 100k resistor and a 470pF cap) and experimentally tried hooking its output up to various spots in the video-signal chain.

I eventually managed to get this single transistor switch system working to a greater or lesser extent. I could clean out a certain amount of the effect - but it was still disappointingly short of the mark.
For one thing, it wasn't staying on for sufficiently long. For another, because I was taking my driving signal off the 120 volt analogue sawtooth vertical output and differentiating via an RC network into the transistor base, the transistor was only acting as a switch as it turned ON. The turn-off process was really slow and sloppy.

At this point, the idea of some digital logic went through my mind.

Over the following weekend, I put the CRO on the set once more and checked out the waveform and gave it some more thought. "Why not a bistable flip-flop? That should allow some control over the pulse length, if nothing else, and it'll be absolutely "on" or "off", ie: a nice simple, predictable switch. Why not give that a try?"

So once again it was out with the pen and paper, now drawing up a bistable. I then wired it up and tried it, and this time things looked much more promising. Initially it would turn off too early, but this was soon traced to my poor choice of sampling point (on the vertical output circuit) and a coupling capacitor which at 470pf was too low. So I fixed that, moved the input to the actual vertical oscillator circuit (which is an 18 volt square wave instead the 120 volt sawtooth wave I'd been sampling) - and now the timing was much closer to what I wanted. Spot on, in fact.

My only other problem now was that there seemed to be no convenient spot in the circuit to which I connect the output of this little rat's nest circuit to completely blank the video. I could reduce it, sure - but not completely kill it. It was frustrating in the extreme. All I wanted was a spot where I could ground the video signal or pull it up to 18 volts for black, and yet there seemed to be no such point! I could hardly believe it - every obvious output connection point produced problems of one sort or another. So once again I seemed to be going around in circles.

It occurred to me that I should just try running the flip-flop output into a high voltage NPN transistor and use that to ground the CRT control grid.
I'd already checked that circuit and the control grids (all wired together) were normally sitting around +100 volts, with the cathodes on about +250 volts (see Fig 6 and/or Fig 8). In other words, the control grids were sitting around -150 volts with respect to the cathodes during normal running, but grounding them would increase this negative bias by a further 100 volts.

I checked the effect of grounding the grid circuit, and going that extra 100 volts negative certainly blacked things out. So I grabbed an MJE340, wired it up to the CRT board, and connected the output of the flip-flop to it. And hey - bingo - success at long last!

I finally had reason to feel pleased with myself.

Here's a few pictures.

The old 65cm Sanyo on the workbench with the back removed.

The original fault was caused by this little electrolytic capacitor on the left. The poor resistor then died as a result.

Having fixed the above fault, I decided to have a look at our Macrovision problem. Looking at the video signal coming out of the video detector with my old Telequipment D66 CRO, I could clearly see the Macrovision interference signal

And here's the delightful effect on the screen. Believe it or not, not only had the family been tolerating this for 10 years with the more recent rented tapes, they told me not to worry about it! (I know that it severly annoyed our friends)

This is what the signal should look like, as seen between Macrovision interference periods

This is the final version of the add-on circuit to provide vertical retrace blanking - a bistable flip-flop (using discrete transistors, no less).
When the vert oscillator waveform swings positive (ie: at far left of lower trace on CRO screen above), the flip-flop switches the MJE340 on and pulls the CRT grid down to 0 volts. When the vert osc swings negative (about half way across the screen), the bistable flips back and the MJE340 turns off again.

I couldn't be bothered making up a nice circuit-board for something as small as this, so I simply wired it onto the back of the Sanyo's main video processing board.

(The protection diode on the RHS is uneccessary, of course - I only included it because I was feeding signal to that side in one of my earlier experiments)

Vertical Blocking Oscillator Tfmr Replacement B & W TV

In my experiences with restoring valve television sets, particularly ones from the 1950's, one of the most common faults is the vertical blocking oscillator transformer failure.

Usually, the transformer shows signs of impending failure before it actually stops oscillation.
Such signs include poor field linearity which cannot be adjusted out, strange looking interlace, and poor synchronisation.
The worst transformers seem to be the ones which are used in the plate circuit of the vertical oscillator (often a 6BM8 triode).
This type of oscillator design is common in the European type of design, used by such brands as Philips and Kriesler. The other type of oscillator circuit employing a blocking oscillator transformer has the transformer in the cathode circuit of the oscillator triode.
This type is common in sets of US type design, such as AWA.
It could be that having the transformer in the cathode circuit means less turns on the windings (lower impedance) thus resulting in better reliability. In fact, I don't think I've ever had a failure of this type!
So, what to do? At first I would replace the transformer with another one, but eventually it would fail again, and I didn't always have the right transformer.
Besides, the vertical locking never seemed to be much good anyway. So, I decided to design an electronic solution and ELIMINATE the transformer off altogether. 
Why this type of circuit found favour I can't understand. Plenty of sets used a multivibrator instead, eliminating the costly transformer.
While a few sets used a separate twin triode for the vertical oscillator (invariably a 12AU7), most actually used the output triode or pentode as the other half.
My design uses the separate oscillator approach.
I did this to allow flexibility of design; i.e.. using it with other sets than the ones I'm describing here.
Also, using the output valve in the oscillator circuit is more critical in that linearity and height settings can change oscillator performance.

This first circuit was implemented in my 1958 Philips 21 CT 335.
A 6C4 triode was added to form the multivibrator in conjunction with the existing 6BM8 triode.
You can of course use other valves; a 12AU7 triode for example. I simply mounted the extra socket on a piece of aluminium and screwed it to the wooden side panel of the chassis. The improvement in performance was amazing. Lock was so strong that the hold control had to be turned right to the extremes to cause loss of synchronisation.
Note that the output side of the circuit is no longer original.
That has nothing to do with the oscillator modification and was actually done with the original oscillator circuit in place.
I had to rewind the output transformer as the primary eventually went o/c.
It's 3000 turns of 39 gauge wire if you're interested. As it happened, after rewinding there wasn't enough room for the feedback winding to go back on, so I ignored it and modified the circuit to a more conventional design.
Note also the blanking level has been increased.
This is due to the teletext problem.

This next circuit was designed for an 1957 HMV F1, but is obviously applicable to the other F series chassis.
These HMV F series sets are notorious for transformer failure. Despite that, they are one of the best performing sets in Australia.
If you wanted a set designed to text book principles with no cost cutting, this is it. As far as the vertical section goes, linearity is so good that a user control is not provided.
As you can see, current feedback is used to optimise the waveform. The yoke current is sensed and fed back to the 6BM8 output pentode grid via a small transformer.

Again, the substitute circuit uses a cathode coupled multivibrator. The transformer was removed and the 6BX6 mounted on an aluminium bracket in its place above the chassis.
In view of how the sync pulses are fed in to the oscillator, I decided to use a pentode with the sync pulses fed into the screen grid.
This worked exceptionally well. Incidentally, an interesting feature of the HMV is that the vertical oscillator plate supply is taken from the normal B+ rather than the B+Boost.
No doubt the feedback network in the output stage compensates for the less than linear waveform.

The third circuit was for a  KRIESLER 79-1. Not suprisingly, the design is similar to the PHILIPS.
Initially, out of curiosity I did try using the 6BM8 pentode as one half of the oscillator.
While it worked, it was unreliable and critical, so went back to the tried and trusted method.
I have a number of 12AU7's with one triode faulty, so I used one of those. Note that R138 is reduced to 1.2M in order to get sufficient height
It should be clear that the circuit can be adapted to other sets but obviously some experimenting will be required to optimise performance.
These are the circuits I've designed and implemented so far and I'm not in a position to design for other sets...that will happen as the need arises.