Screen capacitor... is it a current source to power stage??

[Mick Bailey]

Try this, stick a diode after the reservoir cap. Now the screen cap CAN'T run current back to the first cap. Report back.

This was a feature of some Japanese Hi-Fi designs a few years ago. The claim was that it improved gain consistency and I can see why the idea emerged. I tried this, but couldn't hear any difference myself. Perhaps I never reproduced the conditions necessary for this to be a factor. Certainly with Hi-Fi amps there's a lot of discussion about stiffening or regulating screen supplies - more so than with guitar amps. I guess its down to a music source v music reproduction.

[Alan0354]

Requirement of hifi amp and guitar amp are so different.

[trobbins]

Maybe its good to think of a high frequency (relative to ripple) tone passing through a PP output stage as the output stage gets ever more cranked. The OT winding voltage is fairly linear for most of the high-frequency cycle, but all the fun starts to occur as grid conduction is approached and changes from being a small portion of the cycle to almost square wave duty cycle. Screen current starts to skyrocket relative to idle or non-cranked average levels. Depending on the tube characteristics, and the screen stopper, and the screen bypass cap, the screen voltage can dip many volts for that portion of the cycle. So many variables - so much influence on the voltage waveform - no doubt its so noticeable and relative to the particular amp.

A year or two ago Loudthud generated an easy to see V-I curve with resistive loading as cranking started for a small portion of the waveform. My view was the cathode current (I) was linearly increasing for most of the cycle but then ramped up to a peak level with very little if any further reduction in anode voltage as the screen current starting to rocket up. A nice example.

The anode current certainly increases as the anode voltage heads towards its minima and falls below screen voltage, but as screen current starts to increase in response, then screen voltage starts falling (depending on stopper), and depending on the load curve etc I reckon the anode current could start bending over to follow a collapsing characteristic V-I curve with the right conditions. But amps don't follow nice load lines

A transient attack waveform would certainly pull B+ cap voltage down before Vscreen cap voltage started dropping in response - that's normal filter response; and similarly there is a rebound - which infers that the averaging conditions across the B+ to screen choke are changing - but those averaging conditions have response times down in the low Hz. Changing bypass cap size changes those response times.

Putting a series diode in with the choke makes it a form of half wave rectifier - so there will be a spike at diode turn-off at the choke-diode node from L dI/dt - I'd be putting some kind of resistance or MOV across the choke, and maybe extra capacitance across the diode.

[Alan0354]

Maybe its good to think of a high frequency (relative to ripple) tone passing through a PP output stage as the output stage gets ever more cranked. The OT winding voltage is fairly linear for most of the high-frequency cycle, but all the fun starts to occur as grid conduction is approached and changes from being a small portion of the cycle to almost square wave duty cycle. Screen current starts to skyrocket relative to idle or non-cranked average levels. Depending on the tube characteristics, and the screen stopper, and the screen bypass cap, the screen voltage can dip many volts for that portion of the cycle. So many variables - so much influence on the voltage waveform - no doubt its so noticeable and relative to the particular amp.

A year or two ago Loudthud generated an easy to see V-I curve with resistive loading as cranking started for a small portion of the waveform. My view was the cathode current (I) was linearly increasing for most of the cycle but then ramped up to a peak level with very little if any further reduction in anode voltage as the screen current starting to rocket up. A nice example.

The anode current certainly increases as the anode voltage heads towards its minima and falls below screen voltage, but as screen current starts to increase in response, then screen voltage starts falling (depending on stopper), and depending on the load curve etc I reckon the anode current could start bending over to follow a collapsing characteristic V-I curve with the right conditions. But amps don't follow nice load lines

A transient attack waveform would certainly pull B+ cap voltage down before Vscreen cap voltage started dropping in response - that's normal filter response; and similarly there is a rebound - which infers that the averaging conditions across the B+ to screen choke are changing - but those averaging conditions have response times down in the low Hz. Changing bypass cap size changes those response times.

Putting a series diode in with the choke makes it a form of half wave rectifier - so there will be a spike at diode turn-off at the choke-diode node from L dI/dt - I'd be putting some kind of resistance or MOV across the choke, and maybe extra capacitance across the diode.

I don't think with a diode, there is any big spike at the choke/diode junction. I've gone through the analysis in post #15. I don't see any need of TVS/MOV across the choke.

We all agree the time constant of the sag is in low Hz or even sub Hz. So what is the difference in the sound characteristics with different size caps when the amp is cranked and with screen conducting?

[trobbins]

Maybe you need to put a 12V battery in series with a 12V AC secondary and half-wave choke input rectify that power source in to a capacitor bypassed load. There is always a significant dI/dt when the choke current falls through zero and is commutated off by the diode - that spike voltage level is dependant on all the parasitics surrounding the choke and the diode/choke node.

[Alan0354]

Maybe you need to put a 12V battery in series with a 12V AC secondary and half-wave choke input rectify that power source in to a capacitor bypassed load. There is always a significant dI/dt when the choke current falls through zero and is commutated off by the diode - that spike voltage level is dependant on all the parasitics surrounding the choke and the diode/choke node.

I don't follow what you are saying, please explain.

[trobbins]

I was just indicating that there really will be a level of voltage transient at the diode-choke node if a diode was inserted between the B+ filter cap and the choke feeding the screen bypass/filter cap. And apart from simulating that voltage spike (eg. using PSUD) one could easily confirm that a significant spike is generated on the bench by quickly making a simple test circuit (as described).

How much of a spike occurs in an actual screen supply circuit would depend on many variables - and as such, it would be prudent to add some over-voltage protection or spike suppression across the choke when experimenting with the added diode.

[Alan0354]

I was just indicating that there really will be a level of voltage transient at the diode-choke node if a diode was inserted between the B+ filter cap and the choke feeding the screen bypass/filter cap. And apart from simulating that voltage spike (eg. using PSUD) one could easily confirm that a significant spike is generated on the bench by quickly making a simple test circuit (as described).

How much of a spike occurs in an actual screen supply circuit would depend on many variables - and as such, it would be prudent to add some over-voltage protection or spike suppression across the choke when experimenting with the added diode.

I don't think I can agree. But let's take a look.

Inductor build up a magnetic field when the current going through. If you cut off the current, the inductor do whatever it takes to keep the current going.....by raising or lowering at the end. This is a general theory of inductors. First look at Fig.2 which is a typical relay driving circuit.

From Fig.2 attached, the top drawing is a typical relay circuit where the coil is excited by one end tied to +V and the other end pull to ground by a switch. When the switch is closed, relay at point A is SOURCING a current I to the ground through the switch:

(1) First case is without the protection diode, the moment the switch opens, point A will swing up to high voltage to try to keep SOURCING the current I. This is shown in waveform (1). This is dangerous as it can fry other circuits and shock people.

(2) Standard way of prevent spike is with the diode connected as shown in the top part of Fig.2. When the switch opens, point A starts to swing up trying to keep SOURCING the current I. But when point A swing to 0.7V above +V, the diode turn on and current from point A flow back to the top of the relay coil to discharge the magnetic field inside the coil. This is shown in waveform (2).


Let's use this idea and extend to putting the switch between the top of the relay coil, and then ground the bottom of the coil as shown in the bottom drawing of Fig. 2. In this case, point B is SINKING current I.

(1) First case is without the protection diode. The moment the switch open, the relay coil wants to keep SINKING the current I. Point B will swing to negative high voltage trying to keep SINKING the current. This is shown in wave form (3).

(2) With the protection diode, as soon as point B swing 0.7V below ground, the diode turns on and keep the current flowing until the coil is demagnetized. Then point B goes back to ground.


Fig.1 is the screen grid with diode we are talking about, it is very similar to the case where the switch is on the +V side and the relay on the ground side in the lower drawing of Fig.2. Lets look at Fig.1 The top drawing is in normal operating condition where point A is higher than point C. The diode is on and point B is 0.7V below point A. The lower drawing shows when a large guitar signal causing point A to sag to 430V which is 5V lower than point C at the screen. Right at that moment, Point B is still at 439V, the diode tries to turn off as it is reverse biased. But the choke is charged up with magnetic field and want to keep the current going. The choke is SINKING current at point B. So the choke will try to swing point B NEGATIVE. It will swing down until 429.3( 0.7V below 430V), then the diode turns on again and keep the current flowing. This will keep going until the equilibrium again. The waveform is shown with point A, B and C.

As you can see, there is no spike of any sort in the operation. Please feel free to point out anything wrong with my theory.

[trobbins]

May I suggest that the basic issue relates to AC choke operation where dI/dt cycles between positive and negative levels and choke current effectively lags choke voltage - the relay example only illustrates unipolar choke current operation and so is a little deficient as an example.

Point A is seen as initially being larger than point C, and cycles down below point C voltage (amount of swing V depends on ripple voltage present on A, assuming steady stage conditions). The current in inductor was in a 'positive' direction in a form of pseudo half cycle, but the current cycles down through zero (with a changed sign of dI/dt) and then wants to head negative in value - but then gets stoped abruptly at the 0A level, where dI/dt is highest at the 0A crossing point, and the induced choke voltage at point B flies positive (not negative) which the series diode doesn't capture. Note that at the time current is going negative through 0A, the choke voltage at point B has already been pushed down below point C and needs to keep below C in order to sustain the current cycling down in to the negative part of the current cycle.

[Gingertube]

I think the original poster was wanting to know why additional capacitance at the screen supply node changes the sound so much.

We need look no further than the fact that the tube current is far more dependent upon screen voltage than anode voltage, The screen also acts as a grid with significant gain. Any ripple on the screen supply will affect both tubes, sure the currents due to that ripple signal at the screen will cancel in the output tranny but intermodulation products (at least the odd order terms) will not. This will make the amps sound "muddy".

This is why the HiFi gurus go to the extent of regulating the screen supply.
The next best thing is to use the old Mullard trick of using a common screen supply resistor. As one push pull side draws more screen current the other side draws less so the total current is reasonably constant and the screen supply voltage is therefore more constant.

If you are using large individual screen resistors to protect the output tubes and/or impart some compression then it is essential to tie those resistors back to a very clean power supply node.

Cheers,
Ian

[loudthud]

I did a short simulation using arbitrary numbers. With somewhat normal numbers the choke current stays positive when currents suddenly increase, although it takes a dip. If the choke resistance is low there can be ringing on the screen node. If the screen node capacitance is increased, the choke current can swing negative.

Normal:
C reservoir 100uF
L 5H 200 ohm, low resistance 50 ohm
C screen 50uF
I plate 80mA idle 250mA max
I screen 5mA idle 50mA max

Choke swings negative:
C screen 100uF

[Alan0354]

May I suggest that the basic issue relates to AC choke operation where dI/dt cycles between positive and negative levels and choke current effectively lags choke voltage - the relay example only illustrates unipolar choke current operation and so is a little deficient as an example.

Point A is seen as initially being larger than point C, and cycles down below point C voltage (amount of swing V depends on ripple voltage present on A, assuming steady stage conditions). The current in inductor was in a 'positive' direction in a form of pseudo half cycle, but the current cycles down through zero (with a changed sign of dI/dt) and then wants to head negative in value - but then gets stoped abruptly at the 0A level, where dI/dt is highest at the 0A crossing point, and the induced choke voltage at point B flies positive (not negative) which the series diode doesn't capture. Note that at the time current is going negative through 0A, the choke voltage at point B has already been pushed down below point C and needs to keep below C in order to sustain the current cycling down in to the negative part of the current cycle.

The relay model is not exact, but it is very close. With a diode, you cannot have reverse current. From my step by step explanation, the choke at point B will swing negative and cause the diode to turn on and conduct current, the current will slowly decrease if point does not come up and eventually decrease to zero. I don't see any spike in this case. The only time I can see spike in simulation is the speed of the diode. The swing at point B can be very high speed and the diode might not turn on until nano seconds later, maybe that's what you see in the simulation. Just like you might see a big spike even in the relay circuit when you simulate the switch open.