I am having too much fun in a simulator and discovered that by measuring collector currents for transistors in the output stage can show if it distorts the signal by deviating from the perfect sine wave. Also, spikes of current apparently indicate of quasi saturation operation.
Throwing together 3EF with two output pairs I have measured currents on all three transistors: pre-drivers, drivers, and outputs.
What I observed was that out of the three - driver transistor was working the hardest with current spikes up to 80mA while driving 4 ohm load with 20khz sine wave signal at full power.
I therefore was wondering how to fix it and got an idea - what if I add another driver transistor in parallel?
I measured currents from pre-driver and driver collectors and attached are screenshots to compare both.
Visually comparing between both it can be seen that spikes were reduced to under 36mA for driver transistors.
Just a though for discussion.
Throwing together 3EF with two output pairs I have measured currents on all three transistors: pre-drivers, drivers, and outputs.
What I observed was that out of the three - driver transistor was working the hardest with current spikes up to 80mA while driving 4 ohm load with 20khz sine wave signal at full power.
I therefore was wondering how to fix it and got an idea - what if I add another driver transistor in parallel?
I measured currents from pre-driver and driver collectors and attached are screenshots to compare both.
Visually comparing between both it can be seen that spikes were reduced to under 36mA for driver transistors.
Just a though for discussion.
Attachments
Last edited:
Use modern outputs (NJW0302/0281) or faster Sanken TO-3P devices as drivers. Works VERY well. I've run 5 Pair @ 400W.
Some of the Sankens are 50Mhz devices (with higher Cob). Just use a larger "suckout" cap.... As a whole , the EF3 will outperform a typical
OEM and have no "droop" at high current.
MJE15032/33 hit the upper limit beyond 3 pairs and <4R loads.
OS
Some of the Sankens are 50Mhz devices (with higher Cob). Just use a larger "suckout" cap.... As a whole , the EF3 will outperform a typical
OEM and have no "droop" at high current.
MJE15032/33 hit the upper limit beyond 3 pairs and <4R loads.
OS
It seems that driver transistors will operate in quasi saturation region regardless if it is EF3 or EF2.
Or if it’s MJW1302/3281 ,
MJE15032/33,
TTC004/A004, or
BD139/140
All those were simulated and showed same peaks in current up to 60-80 mA.
Just an observation.
Those peaks can be reduced in the same way as paralleling output transistors I think.
Or if it’s MJW1302/3281 ,
MJE15032/33,
TTC004/A004, or
BD139/140
All those were simulated and showed same peaks in current up to 60-80 mA.
Just an observation.
Those peaks can be reduced in the same way as paralleling output transistors I think.
Interesting. 72mA is not a problem. The problem I see is the blip on the way up. That is distortion.
I don't use capacitors if they introduce reactance at audio frequencies. I bias the drivers at somewhat higher current.
Of course, my amp has had neither simulation nor full-power testing. 😉
Ed
I don't use capacitors if they introduce reactance at audio frequencies. I bias the drivers at somewhat higher current.
Of course, my amp has had neither simulation nor full-power testing. 😉
Ed
No difference , drivers will still have the same Ic regardless of what (group of semi's) they are driving. Output load is the major factor.
Amp i'm using now will get hot with to-220 drivers at a 2R subwoofer load. TO-3P runs cool to the touch. Output Hfe makes the difference ,
OEM's usually use the highest Hfe Sanken outputs to overcome small drivers. My HK680 has NO heatsink for the drivers , at all. They just get slightly
warm.
Last edited:
I blame @printnik for pointing me to the quasi saturation concept lol
I just learned it myself.
Basically we have three regions of operation:
Saturation - current freely goes through collector->emitter->out for NPN BJT
Quasi saturation - current is somewhat impeded by the base current but still freely goes through - I think it is also referred as beta droop
Cut off - no current goes through collector-emitter junction
Forward active - current at the base is modulating current through the collector-emitter->out in proportion defined by beta.
Now, if I understand correctly, spike in current means transistor approaches saturation region and hence the increase in the current.
Or maybe it is the other way?
I guess another question is - will it matter that much if transistor crosses a little into a saturation region?
Just for comparison two screenshots of 20khz sine wave fed at 2.82Vp and 0.5Vp into amp with gain of 11 and powered by 37V rails.
Current spikes reach 60mA and 10mA respectively.
I suppose in practice it probably does not matter though as most driver transistors are way over 1A rating.
I just learned it myself.
Basically we have three regions of operation:
Saturation - current freely goes through collector->emitter->out for NPN BJT
Quasi saturation - current is somewhat impeded by the base current but still freely goes through - I think it is also referred as beta droop
Cut off - no current goes through collector-emitter junction
Forward active - current at the base is modulating current through the collector-emitter->out in proportion defined by beta.
Now, if I understand correctly, spike in current means transistor approaches saturation region and hence the increase in the current.
Or maybe it is the other way?
I guess another question is - will it matter that much if transistor crosses a little into a saturation region?
Just for comparison two screenshots of 20khz sine wave fed at 2.82Vp and 0.5Vp into amp with gain of 11 and powered by 37V rails.
Current spikes reach 60mA and 10mA respectively.
I suppose in practice it probably does not matter though as most driver transistors are way over 1A rating.
Attachments
Last edited:
Cap across gain setting resistor also helped a little in my simulations.
Also combining and trying different transistor combinations helped me to find those that don't have that blip when working together.
Attachments
Last edited:
What available Sanken has high gain and Ft around 50 MHz ?
I see gain around 40 to 150 which of course is respectable
compared to vintage transistor. But not impressive compared to On Semi
Maybe need to look further but they dont provide models? or did I miss that
as well.
I was going to mention Gain of power transistor is only likely way of course to reduce
driver current.
To keep gain higher as well, increase the number of power devices.
Basically looking at data sheets, best performance seems to usually fall around
no more than 1 amp of current to be expected.
King of transistor seemed to be on semi MJL4281 and MJL4302
Gain is 80 to 240 and typical Ft is 35 MHz
Looking closely at datasheet you will see if current is kept to 1 amp per device
Gain will remain around 200 to 240 and Ft is closer to 60 MHz
Looking closer at the Sanken devices around the same as well
about 1 amp for devices has highest gain and Ft gets closer to 80 MHz
either way it seems more logical to have driver for every transistor
and use more linear devices to drive them. Since 3503 and 1381
are very linear around 10 to 15 ma
And briefly looking at current to drive power resistor at 1 amp is
only around 8 to 12 ma Since modern power devices have such incredible gain
if current is kept to around 1 amp
Maybe people over bias drivers? 3503 and 1381 dont seem to need more
than 5 ma of DC current and AC current will only be 8 to 12 ma to drive
power transistor @ 1amp
In numerous simulations once you increase 3503 and 1381 DC current past 6ma
distortion will increase. 5ma seems ideal
I see gain around 40 to 150 which of course is respectable
compared to vintage transistor. But not impressive compared to On Semi
Maybe need to look further but they dont provide models? or did I miss that
as well.
I was going to mention Gain of power transistor is only likely way of course to reduce
driver current.
To keep gain higher as well, increase the number of power devices.
Basically looking at data sheets, best performance seems to usually fall around
no more than 1 amp of current to be expected.
King of transistor seemed to be on semi MJL4281 and MJL4302
Gain is 80 to 240 and typical Ft is 35 MHz
Looking closely at datasheet you will see if current is kept to 1 amp per device
Gain will remain around 200 to 240 and Ft is closer to 60 MHz
Looking closer at the Sanken devices around the same as well
about 1 amp for devices has highest gain and Ft gets closer to 80 MHz
either way it seems more logical to have driver for every transistor
and use more linear devices to drive them. Since 3503 and 1381
are very linear around 10 to 15 ma
And briefly looking at current to drive power resistor at 1 amp is
only around 8 to 12 ma Since modern power devices have such incredible gain
if current is kept to around 1 amp
Maybe people over bias drivers? 3503 and 1381 dont seem to need more
than 5 ma of DC current and AC current will only be 8 to 12 ma to drive
power transistor @ 1amp
In numerous simulations once you increase 3503 and 1381 DC current past 6ma
distortion will increase. 5ma seems ideal
That is a good study. I would go back to EF3, remove C7, and lower R21 so that the drivers aren't so close to cutoff.
Ed
That is not too bad comparing to what pre-driver has to go through at 20khz max power out.
Attached are plotted collector currents for VAS, pre-driver, driver and outputs.
No switch-off speedup capacitor was used.
Note that currents were all set pretty low but not as low that transistors go into cutoff region. Seems to work with 4ohm pure resistive load. Bias for output transistors was set at 100mA.
Lastly attached current plot for VAS zoomed in. Note a notch at the bottom.
Attachments
Last edited:
And switching driver transistors to 15032/33 worsens the life of the pre-drivers even more...
Note after adding two 2.2u caps to pre-driver and driver current setting resistors the situation was improved somewhat, but far from perfect.
Adding caps introduces an extra non linear component to the audio path though... It's a catch 22 🤷🏻♂️
Note after adding two 2.2u caps to pre-driver and driver current setting resistors the situation was improved somewhat, but far from perfect.
Adding caps introduces an extra non linear component to the audio path though... It's a catch 22 🤷🏻♂️
Attachments
Last edited:
15032/33 should have hfe well over 100 up to 1 amp current
Closer to 140 hfe driving 2 power transistors.
Since it shouldn't be more than 13 to 30 ma AC to drive
2 power transistors.
So it shouldn't be stressing the pre drivers very much at all.
Since 15032/33 gain would be very high and easy to drive.
since the load they are driving is nowhere close to needing
1 amp
looking at emitter value of 200 ohms for 15032/33
Doesn't Seem DC current be much higher than 6ma
When they behave little closer to a power transistor.
So DC current should be closer to 15 to 20 ma
Closer to 140 hfe driving 2 power transistors.
Since it shouldn't be more than 13 to 30 ma AC to drive
2 power transistors.
So it shouldn't be stressing the pre drivers very much at all.
Since 15032/33 gain would be very high and easy to drive.
since the load they are driving is nowhere close to needing
1 amp
looking at emitter value of 200 ohms for 15032/33
Doesn't Seem DC current be much higher than 6ma
When they behave little closer to a power transistor.
So DC current should be closer to 15 to 20 ma
Your stages seem to be much more "glitchy" than mine (even at 2R loads). My "fast" CFA seems to have the least correction signal evident. You do see
a bigger correction at the VAS itself than any of the EF3 stages. This sim is the same .5 PPM "Wolverine" EF3 setup ... I don't think I want to reinvent the wheel...
The reason I made the statement about the Sanken's is that my HK680 uses them as outputs and uses un-heatsinked (toshiba) to-220 drivers - always cool !
My designs that use ON's as outputs even make the "beefier" MJE15032/33 drivers warm (with a heatsink). Same bias , same EF3.
OS
Attachments
The amp is missing supply decoupling capacitors on the pre-driver and driver supply. The driver is causing a 0.6V drop which couples into the signal through C25.
In addition to avoiding capacitors with reactance at audio frequencies, another rule is to avoid capacitors that couple supply noise into the signal. I see the latter mistake in many designs.
Ed
In addition to avoiding capacitors with reactance at audio frequencies, another rule is to avoid capacitors that couple supply noise into the signal. I see the latter mistake in many designs.
Ed
Added decoupling capacitors but it did not improve simulation.
What really helped to reduce non linear behavior of the pre-drivers collector currents was increasing the current for drivers MJE15032/33 to 20mA.
Running VAS at 4mA, pre drivers at 5mA, drivers at 20mA, and outputs at 100mA. Much much better behaved output stage.
Attached is collector current plot for pre drivers.
For MJW3281/1302 drivers current needs to be increased to 40 mA to allow for minimum non linearities of the pre driver stage. Results in the second attachment.
What really helped to reduce non linear behavior of the pre-drivers collector currents was increasing the current for drivers MJE15032/33 to 20mA.
Running VAS at 4mA, pre drivers at 5mA, drivers at 20mA, and outputs at 100mA. Much much better behaved output stage.
Attached is collector current plot for pre drivers.
For MJW3281/1302 drivers current needs to be increased to 40 mA to allow for minimum non linearities of the pre driver stage. Results in the second attachment.
Attachments
Last edited: