[Amps] Plate modulation from power supply ripple?

[Manfred Mornhinweg]

Hi all,

let me add my way to see this ripple modulation matter:

First, let's think tubes. In a tube, cathode current flow depends on the 
electric field around the cathode, which is created by the combination 
of the voltages of all other elements, relative to the cathode, and 
considering their relative distances and gridding densities.

So, for a triode, cathode current is determined both by grid and plate 
voltage, with grid voltage having the larger influence. With a tetrode 
or pentode, instead, the plate voltage will only have a small effect, 
while the cathode current is mostly determined by grid and screen voltages.

Most of the cathode current will flow into the anode. Caution has to be 
applied to tetrodes and pentodes in which under some conditions a 
significant current can flow into the screen instead.

Any radio-frequent variation of the plate current will translate into RF 
output from the amplifier. Given that the plate matching isn't altered, 
RF power output depends on the square of the RF component in the plate 
current.

Plate RF voltage depends by plate RF current multiplied by the loading 
resistance offered to the plate by the matching network and the load 
(antenna).

When the above means that the RF voltage on the plate should become 
larger than it can be, according to the existing DC plate supply 
voltage, the tube will go into saturation. In deep saturation, the 
output power will be dependent almost solely on plate voltage.

As a result of all the above, a tetrode or pentode amplifier can be 
highly linear, AND INDEPENDENT OF PLATE VOLTAGE RIPPLE, as long as it 
never saturates. That means driving it low enough so that even at the 
deepest valleys of the ripple waveform on the supply, the plate supply 
voltage does not yet significantly influence the output power.

But if this amplifier is overdriven, its saturated output power will be 
directly modulated by that ripple.

In simplified, practical words: A tetrode or pentode amplifier will not 
produce any significant hum modulation, as long as it is driven low 
enough so that it stays out of clipping even in the supply voltage 
ripple lows. If it's overdriven, hum modulation will appear on the 
peaks. Since voice peaks are usually of higher frequency than the 100 or 
120Hz ripple on the supply, the resulting hum on the output will not be 
very obvious. It's peceived as an unclean, somewhat raspy sound, but not 
really as hum. But in continuous power modes, such as CW or RTTY, hum on 
the output will be readily noticeable when the amplifier is driven into 
  saturation. Instead if operated out of saturation, the output will be 
essentially hum free.

The more ripple there is on the supply voltage, the lower the amplifier 
needs to be driven, to avoid hum modulation. In the extreme case, that 
of having no filter capacitor, the maximum output possible without hum 
decreases to zero.

The above assumes that the other voltages, those for the screen and the 
grid bias, are free from ripple. Any ripple on the screen will directly 
modulate the gain of the tube, and thus introduce hum on the output, 
even while not saturating the amplifier. Hum on the grid, instead, is 
less critical, because it doesn't cause much change to the tube's gain. 
Hum on the grid will just cause a hum component on the plate current, 
but the RF signal will remain superimposed on this hum, without being 
significantly modulated in its amplitude. So, as long as the total plate 
current, including the hum component, doesn't drive anything into 
saturation, ripple on the grid bias should not have a big effect. It's 
like varying the idle current setting at 100 or 120Hz rate. Depending on 
the specific tube, this has only a very small effect on gain.

Now the case of triodes: Since teh plate current of a triode depends 
heavily on plate voltage, in addition to grid voltage, a triode used 
without strong feedback will be nonlinear (distorting). In such an 
amplifier, any ripple on the plate supply will modulate the RF output.

But due to the resulting nonlinearity, triodes are normally never used 
in low feedback circuits, when linear amplification is desired. Instead 
they are used in some setup that provides a high negative feedback, such 
as the grounded grid configuration. A typical triode will have around 
10% to 25% of the output voltage fed back to the input, in opposed 
phase. This tends to establish the output voltage of the tube as a 
certain proportion of input voltage, thus linearizing it to some degree. 
This reduces its sensitivity to ripple on the supply, and this is what 
allows us to run triodes in grounded grid from simple power supplies 
that do have significant ripple.

The exact amount of hum modulation suffered by a specific triode, fed 
from a supply with a certain amount of ripple, could be calculated from 
the tube's transfer functions and the circuit it is used in. The more 
negative feedback it has (the lower its gain), the less sensitive it 
will be to ripple in the plate supply.

Now let's think transistors:

Both bipolar transistors and MOSFETs behave rather close to a pentode, 
in terms of transfer function. That is, their gain does not depend a lot 
on collector or drain voltage, because their collector or drain current 
doesn't vary much with voltage, but instead is mostly controlled by base 
current or gate voltage. For this reason, amplitude modulation of the 
output by ripple on the supply is much as it is for pentodes and 
tetrodes, that is, it becomes significant only when the transistor is 
driven into saturation.

But there is another problematic effect, which is not a big factor in 
tubes: The internal capacitances of transistors change a lot with 
applied voltage. Specially in MOSFETs, in which these capacitances are 
rather high, this leads not only to some amplitude modulation through 
gain changes, but also to phase modulation, causing an FM hum on the 
signals, when the supply voltage has ripple on it.

But this effect is small, and depends a lot on the specific devices 
used. Some MOSFETs show a significant change of capacitance with 
voltage, even at the supply voltage level. Others show this big change 
of capacitance only at low voltage levels, say, one fifth of the supply 
voltage or so. With such a MOSFET, a small signal will be confined to 
voltage swings well out of the area where capacitances change a lot, and 
this will be rather unaffected by ripple on the supply. As we go further 
toward large signals, close to clipping, a MOSFET will suffer from these 
capacitance changes, causing both nonlinear gain reduction and phase 
shifting, and at this level it will also show increased sensitivity to 
supply ripple modulating the signal both in amplitude and phase.

So, this is how I see it. It's certainly not a complete picture, but I 
hope I didn't forget any of the most important effects.

In practice, some ripple on the supply (say, 10%) seems to be just a 
minor issue, in most ham amplifiers. On the other hand, with today's 
technology of switching regulators it's really easy and inexpensive to 
make power supplies that are essentially ripple free. A switching supply 
with just a single stage output filter typically has less than 0.1% 
ripple at full load, less at partial load, and this small ripple is not 
at line frequency, but at its switching frequency of several tens of 
kHz. An additional stage of filtering, made with a small inductor and a 
modest capacitor, can easily reduce the ripple to 0.01% or even less. 
With a switching supply feeding an amplifier it becomes very important 
to have very low ripple, because the sidebands introduced by the higher 
frequency ripple fall into rather distant channels, where they are far 
more noticeable. But since it's so easy to produce very low ripple 
output from such a power supply, this isn't a problem at all. It's just 
a matter to be aware of during the design phase.

Manfred

========================
Visit my hobby homepage!
http://ludens.cl
========================