Rosemary Ainslie Quantum Magazine Circuit COP > 17 Claims

[MarkE]

That's looking very solid.

[TinselKoala]

I think so too. All the raw data is preserved in the still images and the short videos I make from them, and the calibrations are justifiable and repeatable. I've even done ice-and-boiling point calibrations of the several thermometers I use. Frequencies and duty cycle numbers come from the Philips counter and the Link DSO.


Now here's an issue I'd like to kick over a bit. Please let me know what your thoughts are.

When the 5-mosfet circuit is in "Q2 oscillation mode", that is, with a suitable _negative voltage_ and current supplied to the Source pin of the Q2 mosfet and the rest of the circuit essentially inactive.... where is the appropriate place to take the input voltage reading? Both the main battery and the bias source contribute power to the load in this case, and the main supply and bias supply are in series. So I am thinking that the correct place to measure the input voltage will be between the battery positive and the Q2 mosfet's Source pin.... but the circuit winds up having the Current Sense Resistor connected from the Main battery negative to the mosfet Gate. This means the CSR isn't measuring the true load current, and the ordinary way of measuring voltage across the main battery only, isn't measuring the true input voltage.

Right? See the stripped-down diagram below.

The "normal" arrangement of measurements is to measure the input Voltage from TP A to "NERD Reference", and the Current as the Vdrop from Tp B to "Nerd Reference". But as we know this does not capture the full power in the circuit while Q2 is oscillating. The true supply voltage, it should be clear now, is actually the series voltage of the Main Battery + the Bias source voltage, and so should be measured from TP A to TP C at the mosfet source pin. The Existing CVR, which shows the big oscillations in current, isn't even involved in the main power loop during Q2 oscillations. The only single place to measure the true load current, which is the total supplied by the main and bias sources, would appear to be where I have placed the DMM A in the diagram.

So, this is how I intend to measure the input power when testing the Common Gate Oscillator portion of the circuit separately, using continuous Q2 oscillations. VSupply will be from TP A to TP C, and Current will be from the DMM A or an isolated non-inductive CVR in that location. I'll also compare readings from this DMM with similar readings taken between TP C and the negative terminal of the bias source.

(The circuit shown in the diagram is what happens during the Q2 oscillation parts of the cycle. The Q1 can be removed from the circuit entirely without affecting the Q2 oscillation portion of the cycle at all, as I've shown in several video demonstrations, and this results in the circuit in the present diagram.)

[MarkE]

I think so too. All the raw data is preserved in the still images and the short videos I make from them, and the calibrations are justifiable and repeatable. I've even done ice-and-boiling point calibrations of the several thermometers I use. Frequencies and duty cycle numbers come from the Philips counter and the Link DSO.


Now here's an issue I'd like to kick over a bit. Please let me know what your thoughts are.

When the 5-mosfet circuit is in "Q2 oscillation mode", that is, with a suitable _negative voltage_ and current supplied to the Source pin of the Q2 mosfet and the rest of the circuit essentially inactive.... where is the appropriate place to take the input voltage reading? Both the main battery and the bias source contribute power to the load in this case, and the main supply and bias supply are in series. So I am thinking that the correct place to measure the input voltage will be between the battery positive and the Q2 mosfet's Source pin.... but the circuit winds up having the Current Sense Resistor connected from the Main battery negative to the mosfet Gate. This means the CSR isn't measuring the true load current, and the ordinary way of measuring voltage across the main battery only, isn't measuring the true input voltage.

Right? See the stripped-down diagram below.

The "normal" arrangement of measurements is to measure the input Voltage from TP A to "NERD Reference", and the Current as the Vdrop from Tp B to "Nerd Reference". But as we know this does not capture the full power in the circuit while Q2 is oscillating. The true supply voltage, it should be clear now, is actually the series voltage of the Main Battery + the Bias source voltage, and so should be measured from TP A to TP C at the mosfet source pin. The Existing CVR, which shows the big oscillations in current, isn't even involved in the main power loop during Q2 oscillations. The only single place to measure the true load current, which is the total supplied by the main and bias sources, would appear to be where I have placed the DMM A in the diagram.

So, this is how I intend to measure the input power when testing the Common Gate Oscillator portion of the circuit separately, using continuous Q2 oscillations. VSupply will be from TP A to TP C, and Current will be from the DMM A or an isolated non-inductive CVR in that location. I'll also compare readings from this DMM with similar readings taken between TP C and the negative terminal of the bias source.

(The circuit shown in the diagram is what happens during the Q2 oscillation parts of the cycle. The Q1 can be removed from the circuit entirely without affecting the Q2 oscillation portion of the cycle at all, as I've shown in several video demonstrations, and this results in the circuit in the present diagram.)

The problem with obtaining the total voltage across the circuit is that the actual voltage of the bias source is not accessible.  Whatever the internal impedance of the device acting as a voltage source, be it a battery or function generator or something else, the Q2 transconductance multiplied by that internal impedance develops a voltage drop that works to drive the difference between the voltage source value and the gate voltage to the gate threshold value.   I would use an external 50 Ohm resistor as the source degeneration resistor for Q2.  Then if you have a Vbias source that is half an ohm or less up to ~10MHz then 99% or more of the voltage drop caused by the Q2 gain will appear across the resistor, allowing you to accurately measure the Vbias contribution.  If you want to turn the oscillations on and off you can use the function generator with an additional N channel MOSFET as shown in this drawing.  If you just want to let the oscillations run, then you can replace the additional MOSFET with a short, or use the MOSFET and connect its gate to the Q1 Source node.

To get less than 0.5 Ohm to 10MHz you need to keep the decoupling capacitor network inductance below 8nH.  You will want to parallel with a capacitor big enough to handle the 150kHz pulses you are using.  If you have more than one of those boards Steve designed you can use the capacitor network on that second board.

[TinselKoala]

The problem with obtaining the total voltage across the circuit is that the actual voltage of the bias source is not accessible.  Whatever the internal impedance of the device acting as a voltage source, be it a battery or function generator or something else, the Q2 transconductance multiplied by that internal impedance develops a voltage drop that works to drive the difference between the voltage source value and the gate voltage to the gate threshold value. 

That's right, and that's why the observed Gate drive voltage always bottoms out at around  - 4.2 V, with fuzz, no matter the negative offset or p-p voltage setting of the FG or other bias source. I'll have to do some measurements to see if that is actually a problem as far as the overall power measurement goes, though. Since the current in the system will increase as the applied, or opencircuit, bias voltage increases, the instantaneous current monitor reading x the inst. total series voltage should still give the correct power, I think, maybe.

I would use an external 50 Ohm resistor as the source degeneration resistor for Q2.  Then if you have a Vbias source that is half an ohm or less up to ~10MHz then 99% or more of the voltage drop caused by the Q2 gain will appear across the resistor, allowing you to accurately measure the Vbias contribution.  If you want to turn the oscillations on and off you can use the function generator with an additional N channel MOSFET as shown in this drawing.  If you just want to let the oscillations run, then you can replace the additional MOSFET with a short, or use the MOSFET and connect its gate to the Q1 Source node.

To get less than 0.5 Ohm to 10MHz you need to keep the decoupling capacitor network inductance below 8nH.  You will want to parallel with a capacitor big enough to handle the 150kHz pulses you are using.  If you have more than one of those boards Steve designed you can use the capacitor network on that second board.

All good advice of course, but I don't want to stray too far from the original circuit. Let me do some test comparisons between the "standard" method that has been used by the Ainslie gang and the various suggested methods. I don't have another S board, but it can easily incorporate a Q2 arrangement and then be driven in continuous oscillation mode by either an external power supply proper, or battery, or the FG set to DC output (I don't think Ainslie's FG has a DC output setting) of the proper polarity.

[MarkE]

That's right, and that's why the observed Gate drive voltage always bottoms out at around  - 4.2 V, with fuzz, no matter the negative offset or p-p voltage setting of the FG or other bias source. I'll have to do some measurements to see if that is actually a problem as far as the overall power measurement goes, though. Since the current in the system will increase as the applied, or opencircuit, bias voltage increases, the instantaneous current monitor reading x the inst. total series voltage should still give the correct power, I think, maybe.
All good advice of course, but I don't want to stray too far from the original circuit. Let me do some test comparisons between the "standard" method that has been used by the Ainslie gang and the various suggested methods. I don't have another S board, but it can easily incorporate a Q2 arrangement and then be driven in continuous oscillation mode by either an external power supply proper, or battery, or the FG set to DC output (I don't think Ainslie's FG has a DC output setting) of the proper polarity.

Because of the MOSFET gain the additional net voltage across the heater resistor tends to reduce to approximately:  (V_BIAS - V_THS)*R_HEATER/R_BIAS and the additional power reduces to approximately:2*V_BATTERY*(V_BIAS - V_THS)/R_BIAS + (V_BIAS - V_THS)²*R_HEATER/R_BIAS².  If you can suppress the oscillations, then you should be able to verify this.   

The first term is larger than the second by the ratio: 2*V_BATTERY/(V_BIAS - V_THS) * R_BIAS/R_HEATER.  In rough numbers with a 24V battery voltage the second term is less than 5% of the first.  In the August demonstration, it was around 1.5%.