I chose a pair of FQA28N15 and FQA36P15 FETs that had transconductances of 4.59S and 3.90S, giving a gm ratio of 1.18, which is higher than what I distributed as "matches" to the Holy Grail OS builders.
I made spectral measurements without the RX pot in circuit (column 1 of image 1) and with the RX pot adjusted for equal AC voltages across sense resistors R1 and R2 when for a 1 Watt (column 2 of image 1), 60Hz output into 8 Ohms.
I also chose a very rare pair of FQA28N15 and FQA36P15 FETs that had nearly identical gm to start with and made spectral measurements (column 3 of image 1).
The rows of image 2 are wattage sweeps for 8 Ohms, wattage sweeps for 4 Ohms, and frequency sweeps at 1 Watt into 8 Ohms.
With the RX adjustment, the 2nd harmonic of the unbalanced FQA FETs is somewhat reduced by around 6dB.
I performed the same spectral measurements and RX pot adjustment for the pair of IRFP140 and Harris IRFP9140 FETs, shown in image 2 compared to the well matched FQA FETs in column 3.
With the RX adjustment, the 2nd harmonic of the IRFP FETs is reduced by 20dB at 1 Watt and interestingly, the 3rd harmonic is also significantly reduced at high wattages.
Conclusion: The RX pot adjustment made a significant reduction of all harmonics for the IRFP FETs, but not much for the FQA FETs that were already reasonably well matched.
I made spectral measurements without the RX pot in circuit (column 1 of image 1) and with the RX pot adjusted for equal AC voltages across sense resistors R1 and R2 when for a 1 Watt (column 2 of image 1), 60Hz output into 8 Ohms.
I also chose a very rare pair of FQA28N15 and FQA36P15 FETs that had nearly identical gm to start with and made spectral measurements (column 3 of image 1).
The rows of image 2 are wattage sweeps for 8 Ohms, wattage sweeps for 4 Ohms, and frequency sweeps at 1 Watt into 8 Ohms.
With the RX adjustment, the 2nd harmonic of the unbalanced FQA FETs is somewhat reduced by around 6dB.
I performed the same spectral measurements and RX pot adjustment for the pair of IRFP140 and Harris IRFP9140 FETs, shown in image 2 compared to the well matched FQA FETs in column 3.
With the RX adjustment, the 2nd harmonic of the IRFP FETs is reduced by 20dB at 1 Watt and interestingly, the 3rd harmonic is also significantly reduced at high wattages.
Conclusion: The RX pot adjustment made a significant reduction of all harmonics for the IRFP FETs, but not much for the FQA FETs that were already reasonably well matched.
But you need to know which FET has a higher transconductance first, and place the trimpot there.
And you need a distortion analyser.
Also, in doing so, you increase effective output impedance, because essentially, you are reducing the effective transconductance of the FET with the trimpot to enable better harmonic cancellation.
No free lunch.
Patrick
And you need a distortion analyser.
Also, in doing so, you increase effective output impedance, because essentially, you are reducing the effective transconductance of the FET with the trimpot to enable better harmonic cancellation.
No free lunch.
Patrick
No distortion analyzer is needed. The relative transconductances are measured by the AC voltages across the sense resistors when the amplifier is driven at around 1 Watt , 60Hz into an 8 Ohm load. The FET whose drain sense resistor has the larger AC voltages across it has the larger transconductance.
Yes, obviously decreasing the gm of one of the FETs also decreases the combined gm of the output stage, increasing the effective output impedance.
Yes, obviously decreasing the gm of one of the FETs also decreases the combined gm of the output stage, increasing the effective output impedance.
That would work if the 2nd and higher harmonic terms have the same proportional constant to the transconductance between the two devices,
I.e. you can achieve harmonic cancellation by balancing current sharing between NMOS & PMOS.
Cheers,
Patrick
So in your experiments, the minimum distortion point happens to correspond to equal current sharing between NMOS and PMOS ?
And that applies to all FET pairings, correct ?
Interesting,
Patrick
With my limited tests I would not claim that the minimum distortion point happens to correspond to equal current sharing between NMOS and PMOS.
But there is something unexplained that puzzles me. The 2nd harmonic can be adjusted in many ways. Why is the growth of the 3rd harmonic vs. watts greatly reduced when the gms are balanced?
But there is something unexplained that puzzles me. The 2nd harmonic can be adjusted in many ways. Why is the growth of the 3rd harmonic vs. watts greatly reduced when the gms are balanced?
Good afternoon!
I’ve a question about input-connection:
@lhquam , you are pulling 3 wires from the input-rca‘s, one is in+, one is in-, the third goes to gnd… would I make it less perfect if I went with a „jumper“ from in- to gnd instead of using a separate wire from the connector to gnd? And could you please explain why this is done this way? I believe I have come across this configuration in the BA18 Wayne‘s Line Stage, right?
I’ve a question about input-connection:
@lhquam , you are pulling 3 wires from the input-rca‘s, one is in+, one is in-, the third goes to gnd… would I make it less perfect if I went with a „jumper“ from in- to gnd instead of using a separate wire from the connector to gnd? And could you please explain why this is done this way? I believe I have come across this configuration in the BA18 Wayne‘s Line Stage, right?
The primary reason that I use 3-wires to the RCA connector is that it can be difficult to insert 2 wires, one to the RCA connector and one to In- into the same Gnd (or GndR) position of the small 4-pin connector block.
However, you could solder the jumper wire to the RCAGnd wire so that only one wire needs to be inserted into the connector block.
However, you could solder the jumper wire to the RCAGnd wire so that only one wire needs to be inserted into the connector block.