Don't worry, most people here (Catalonia) don't like to be called spanish (me included). By the way i'm trying a more radical design. Instead of using EC, that is a kind of NFB despite Bob Cordell saying it's feedforward, i will do a unity-gain output stage (maybe gain of 2 if it's too harsh) and use another opamp to give the desired voltage gain. Hopefully 140 dB nfb will make make all output-stage distortion disappear.
ionomolo said:
This is possible I agree, the difficult part is to retain phase coherency in other words not to shift the phase in time which creates an even bigger problem since you have non related spurious to contend with.
Some years back, my software guru friend and I tried error correction using fast microprocessors to do this, But sadly, microprocessors work with a watch, one that needs winding up and what happened in real time now calculating with the processor a result was available only when you tart listening a new CD some days later.
You have to compensate in real time and do it with damn fast components.
My biggest gripe in audio remains phase shift and I claim that phase shift, not distortion is the mechanism that colours the characteristics of real instruments and voices.
I have been occupied now for some weeks with a very simple amplifier that will give zero phase shift from DC to 30kHz, It sounds simple but it is not as you have to bias components hard to operate in an absolute linear region.
Once they are linear you find at some odd instance you have 360 deg phase shift causing a strong radio transmitter at say 20MHz,
perfect frequencies 19 MHz, 20 MHz, etc,
If I want to make something to generate at 20 MHz it is a lot of trouble, Maybe I will build John's uncompensated amplifier and save me many weeks work to make a stable oscillator of >200 watts
You're right. I've choosen mosfets for two reasons: ( 1 ) A single device makes the same than 2 bjt's and ( 2 ) They are really fast.
I guess i'll get more phase correctness if i set a lower value for the gate resistors, i will try to go to 235 ( 470 || 470 ) to see if it's stable.
Douglas Self states in his book that a well-designed Szikai output can have 0.014% open-loop THD at 40W. This book says really weird things, like that a quasi-complementary mosfet output stage measures slightly better than the complementary one. Is this right? Would not be interesting to try other output stages?
I guess i'll get more phase correctness if i set a lower value for the gate resistors, i will try to go to 235 ( 470 || 470 ) to see if it's stable.
Douglas Self states in his book that a well-designed Szikai output can have 0.014% open-loop THD at 40W. This book says really weird things, like that a quasi-complementary mosfet output stage measures slightly better than the complementary one. Is this right? Would not be interesting to try other output stages?
Another interesting point, gate capacitances. I get the worst because the lower must be increased to prevent asymetrical behaviour.
2SK1058 600 pF 2SJ162 900 pF (900 pF) [RENESAS]
2SK1530 900 pF 2SJ201 1500 pF (1500 pF) [TOSHIBA]
IRFP240 1300pF IRFP9240 1200pF (1300 pF) [INTERNATIONAL RECTIFIER]
It seems that the best option regarding low capacitance (and thus phase correctness) is the 2SK1058/2SJ162 pair. May be being manufactured by toshiba better than being by renesas?
Linearity is also really complicated to deduce from datasheets, since they show gm vs current and drain current vs gate-source voltages in log axis.
I must also say that i have no preferences between mosfet and bjt. Only that zero dc gate current is a plus when designing. Keep in mind that bjt transistors also have capacitive behaviour that is not negligible, but rather masked by high base current that adds more design challenges.
2SK1058 600 pF 2SJ162 900 pF (900 pF) [RENESAS]
2SK1530 900 pF 2SJ201 1500 pF (1500 pF) [TOSHIBA]
IRFP240 1300pF IRFP9240 1200pF (1300 pF) [INTERNATIONAL RECTIFIER]
It seems that the best option regarding low capacitance (and thus phase correctness) is the 2SK1058/2SJ162 pair. May be being manufactured by toshiba better than being by renesas?
Linearity is also really complicated to deduce from datasheets, since they show gm vs current and drain current vs gate-source voltages in log axis.
I must also say that i have no preferences between mosfet and bjt. Only that zero dc gate current is a plus when designing. Keep in mind that bjt transistors also have capacitive behaviour that is not negligible, but rather masked by high base current that adds more design challenges.
My best advice is put it together with space wires and mount MOSFETs on heat sink, then listen. Whether you hear the difference in gate capacitance is debatable. If you have one of each pair , then see which you like best and use those in the final amplifier and in future designs, but I don't think that you will hear a markable difference, I think nothing at all.ionomolo said:
It's not about hearing capacitance but rather about hearing what capacitance does to the poor driver opamp. I don't expect it to like driving a 3 nF cap, despite phono preamps in typical applications have these in the feedback loop and the thing does work and sound good. Your advise is the best, of course, but it needs money and time 
.
.
Someone knows how much contributes each stage in a power amplifier to overall phase shift?
The LM4562 claims near a 90º phase shift at 10 KHz. I really feel this to be very high.
I guess that phase shift in the common cathode arrangement should be lower (well, 180º but less dependant with frequency). Maybe i'll fix "everything" by removing the LM4562 and replacing it with a low-voltage tube arrangement, but here gate capacitance would be a serious issue.
The LM4562 claims near a 90º phase shift at 10 KHz. I really feel this to be very high.
I guess that phase shift in the common cathode arrangement should be lower (well, 180º but less dependant with frequency). Maybe i'll fix "everything" by removing the LM4562 and replacing it with a low-voltage tube arrangement, but here gate capacitance would be a serious issue.
New design
This simulates way better (maybe 20 times less open-loop THD than the original schematic). Havn't already built it.
New design
The caps paralleled with the diodes might make the circuit unstable, remove them if this happens. If still unstable use the common trick (bandwith limiting through 50 to 220 pF between inverting and noninverting inputs of the opamp).
This simulates way better (maybe 20 times less open-loop THD than the original schematic). Havn't already built it.
New design
The caps paralleled with the diodes might make the circuit unstable, remove them if this happens. If still unstable use the common trick (bandwith limiting through 50 to 220 pF between inverting and noninverting inputs of the opamp).
Link to freeware program for opening of .MDI files posted by ionomolo:
http://www.hot.ee/mdiviewer/MDIviewer.exe
http://www.hot.ee/mdiviewer/MDIviewer.exe
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.