I was just commenting on the low Vds question posed to ZM. I know in the ZV8 article, we are dealing with a device that is triodish down at low Vds levels. Not sure if these fets could be examined in a similar light. I have seen in the RO85 curves a similarity at low Vds, but not real sure they exist in this form for these fets. Was not poking at design in any way, just grasping at possible explanation until ZM comes along and makes me look like a big dodo.
So far, Nelson hasn't responded for my plea for LU1014s and their P-channel complements? 😀 That would be ideal here if the transconductance is high enough. Remember that the transformer provides no voltage gain and all voltage (and current) gain is from the output FETs. This is truly a low OLG and low feedback amplifier. 😎
Last edited:
sorry - usual ZM's brainfart
of course that I meant exactly opposite of what I wrote ( "that right , not that right ! "
) - rails voltage increase ........ meaning on modest rail voltage increase , to have more space for both , gain and cascoding , parts - to breatheThe plots in posts #4694 http://www.diyaudio.com/forums/pass-labs/216616-f6-amplifier-470.html#post3536985 and #4713 http://www.diyaudio.com/forums/pass-labs/216616-f6-amplifier-472.html#post3539181 clearly show the following with the cascoded circuit:
- H2 increases roughly linearly with frequency (above 1kHz), while H3 is more constant with frequency. I suspect that this is mostly due to capacitances that cannot be cancelled at higher frequencies.
- H3 increases roughly linearly with watts, while H2 is more constant with watts.
Last edited:
H2 can be approximately cancelled at one particular frequency and power level, but not across all frequencies and power levels, primarily due to the unbalanced circuit capacitances in the upper and lower push-pull circuits. In a balanced (bridged) amplifier or a circlotron, with well matched FETs, H2 (near) cancellation is possible at all frequencies.
Last edited:
On F3 brainfart and thinking about F6, I have question about modulation cap on output. Would changing this value affect amp performance. In ZV8, nèlson ßpoke of testing at 1W to determine this vàlue, suggesting that it had an effect.
Are you asking about the cap from the output FET source pins to the secondary winding? In the case of the F6, the value was set very high (1000uF) because the upper secondary winding it tied (AC-wise) to the output and the coupling cap must overcome the interwinding capacitances and the FET Cgs and Cgd capacitances. If the coupling cap is too large, then the 10K x Ccoup time constant gets very long for amplifier start-up and bias adjustments. The N and P halves of the circuit need the same time constant, otherwise to output offset voltage would be very large until everything stabilizes.
My complementary design does not need such large coupling caps, since the windings AC-wise approximately the input voltage rather than the output voltage.
So far, Nelson hasn't responded for my plea for LU1014s and their P-channel complements? 😀 That would be ideal here if the transconductance is high enough. Remember that the transformer provides no voltage gain and all voltage (and current) gain is from the output FETs. This is truly a low OLG and low feedback amplifier. 😎
- Unlike F5, Teaser F6CC does not derive any voltage gain from the front end; which I theoretically advocate in the feedforward approach.
- A quasi complementary N or P Mosfet has 2 gain stages versus the one of the parent. Thus, the transconductance of a quasi complementary compound is expected to be higher in magnitude than its parent. Now, one has the opportunity to make complementary quasi complementary Mosfets from high gain complementary bjts, and complementary N and P Mosfets.
So what effects would result from lowering it? Would the output dignal not fully couple signal to gate of fet due to "competition" from other sources mentioned above.
The main effects will be higher H2 distortion, increasing with frequency. In the "official" F6, you can easily get by with 220uF without much loss. I did this because I was seeing low frequency (< 0.1 Hz) oscillation problems which went away with smaller coupling caps. You will see my many earlier posts about this issue.
So it is reasonable that this cap also influences final disotrion spectra. Is this because cap size influences amount of local feedback? Sorry for not looking through thread. So much OT stuff
Thanks for the explanation.
Thanks for the explanation.
Another attempt to explain the issue: As frequencies get very low, the voltage drops across those coupling caps increase, reducing the AC voltages into the gates of the MOSFETs (a capacitor divider effect). This reduces the openloop gain which in turn increases the distortion.So it is reasonable that this cap also influences final disotrion spectra. Is this because cap size influences amount of local feedback? Sorry for not looking through thread. So much OT stuff
Thanks for the explanation.
I did some different measurements today: Closed-loop and open-loop rolloff and open-loop gain. The rolloff frequencies were estimated from the oscilloscope time base.
Closed-loop rolloff: Gain=6, 1.3A bias, Cfb=10nF compensation cap
-3dB 5.5 div @ 1us/div = 182 kHz (LTSpice predicts 141 kHz)
-6dB 8 div @ .5us/div = 250 kHz (LTSpice predicts 234 kHz)
Open-loop rolloff
-3dB 6 div @ 2us/div = 83 kHz
-6dB 8.6 div @ 1us/div = 116 kHz
Open-loop gain measurement: 1kHz sine wave, 2 volt peak output
Fluke DMM measurement
32mV RMS input for 2V peak output
Oscilloscope measurement
4.5 div P-P @20mV/div = 90mV P-P
0.5*0.090*sqrt(0.5) = 32mV RMS - agrees with Fluke
Open-loop gain = 2.0/0.032 = 62.5 (LTSpice predicts 70)
Closed-loop rolloff: Gain=6, 1.3A bias, Cfb=10nF compensation cap
-3dB 5.5 div @ 1us/div = 182 kHz (LTSpice predicts 141 kHz)
-6dB 8 div @ .5us/div = 250 kHz (LTSpice predicts 234 kHz)
Open-loop rolloff
-3dB 6 div @ 2us/div = 83 kHz
-6dB 8.6 div @ 1us/div = 116 kHz
Open-loop gain measurement: 1kHz sine wave, 2 volt peak output
Fluke DMM measurement
32mV RMS input for 2V peak output
Oscilloscope measurement
4.5 div P-P @20mV/div = 90mV P-P
0.5*0.090*sqrt(0.5) = 32mV RMS - agrees with Fluke
Open-loop gain = 2.0/0.032 = 62.5 (LTSpice predicts 70)
Hi Antoinel,
Look here http://www.diyaudio.com/forums/pass-labs/216616-f6-amplifier-438.html#post3517149
First picture there you can see what lhquam said about k2 changing over the frequency and the difference of behaviour of k2 and k3 over the frequency band.
🙂
Lhquam, those measurements were for the cascode pp? So cascoding a fet increases the gain vs non cascode stage ?
No, with cascoding, the gain is slightly lower. The differences between the cascode and non-cascode plots are due to:
- Reduction/elimination of the FET non-linearity due to variations in Vds (Early effect). This effects all frequencies about equally and probably explains why H3 is much flatter vs. frequency in posts #4694 F6 Amplifier and #4713 F6 Amplifier.
- Reduction of the effects due to the FET Cgd capacitance since Vds is kept nearly constant. The effect of this parasitic capacitance increases with frequency.
Last edited: