This again. You posted a scope shot showing 30V/us. My point at the time was that you can't make the amp slew any faster than the maximum possible limit dictated by the LTP tail current and the value of Cdom.
And of course increasing the LTP tail current won't increase the slew rate if the VAS current/driver Cin is the limiting factor.
As for the LTP transistor dissipation, 200mW isn't a problem for a to-92, so a 5mA tail current (if you wan't it) isn't a problem unless the rail voltages are very high. But if you are still worried, stick a pair of zeners in series with the LTP collectors.
Hey!
The oscillation was gone with no output stage connected. To avoid further misunderstanding, I'll add images not links to them. Posts will be longer, but let's hope this will not be a problem.
I have heard lot about the bulb connection, but haven't implemented it yet. Thil will be next in my todo list!🙂 For now replaced the fuses with 50 ohm / 5W resistors.
Before going back to connecting the output stage, let us look at the input and VAS stage, do you think this is good enough? Connected the bias multiplier too and increased the tail current to 5 mA. Scope is connected to VAS current source collector.
Current schematics:
Noise with input shorted (20mV/250ns):
Clipping(50V/250us). Little concernered with the instability in negative clipping.
Square wave 10kHz 100Vpp showing rise time ~40V/us:
Any comments are welcome!🙂
Rait
The oscillation was gone with no output stage connected. To avoid further misunderstanding, I'll add images not links to them. Posts will be longer, but let's hope this will not be a problem.
I have heard lot about the bulb connection, but haven't implemented it yet. Thil will be next in my todo list!🙂 For now replaced the fuses with 50 ohm / 5W resistors.
Before going back to connecting the output stage, let us look at the input and VAS stage, do you think this is good enough? Connected the bias multiplier too and increased the tail current to 5 mA. Scope is connected to VAS current source collector.
Current schematics:
An externally hosted image should be here but it was not working when we last tested it.
Noise with input shorted (20mV/250ns):
An externally hosted image should be here but it was not working when we last tested it.
Clipping(50V/250us). Little concernered with the instability in negative clipping.
An externally hosted image should be here but it was not working when we last tested it.
Square wave 10kHz 100Vpp showing rise time ~40V/us:
An externally hosted image should be here but it was not working when we last tested it.
Any comments are welcome!🙂
Rait
Hi,
Is that clipping artifact an oscillation burst or "just" large overshoot? What frequency is it, if oscillation? The VAS transistor goes into really deep saturation during clipping with about ~50mA base current (can the beta enhancer really take this safely?) and ~20mA collector current so it isn't too strange if there is overshoot.
If the VAS is on the verge of local oscillation it may get upset when the output stage is connected, so it's probably a good idea to sort any such issues out first. 🙂
Is that clipping artifact an oscillation burst or "just" large overshoot? What frequency is it, if oscillation? The VAS transistor goes into really deep saturation during clipping with about ~50mA base current (can the beta enhancer really take this safely?) and ~20mA collector current so it isn't too strange if there is overshoot.
If the VAS is on the verge of local oscillation it may get upset when the output stage is connected, so it's probably a good idea to sort any such issues out first. 🙂
Last edited:
the ratio of r18 to r24 are a bit too close, currently 2.74:1
Q13 may start triggering on large VAS currents.
Increasing the ratio to between 3:1 and 4:1 may prevent premature triggering.
Q13 may start triggering on large VAS currents.
Increasing the ratio to between 3:1 and 4:1 may prevent premature triggering.
Try to connect small diode (something like BAV21,200V, fast) in parallel to C14, cathode to colector of Q30...Q30 is in deep saturation at overdrive, overshoot is caused by NFB after recovery from saturation.
Andrew, why does the VAS need to have +23mA / - 69mA output current capability? (4:1 ratio in resistor values, standing current unchanged) Don't you think symmetric output with some margin so that Q13 be kept well turned off during normal operation would be enough? If not, why?
To me, 2.74:1 seems more than enough. The 460mV from 46mA (-23mA output current) isn't near enough to turn on Q13 appreciably. What use would there be to have more driving capability in the negative direction than in the positive? The current setting is about +23mA / -40mA max VAS output current.
BV: Good idea! 🙂
MJL21193: You have a point there, +-20mA from the VAS is enough to drive the output stage to +-40A even if the predrivers are bypassed while output devices and drivers are of the lowest beta... 😛
To me, 2.74:1 seems more than enough. The 460mV from 46mA (-23mA output current) isn't near enough to turn on Q13 appreciably. What use would there be to have more driving capability in the negative direction than in the positive? The current setting is about +23mA / -40mA max VAS output current.
BV: Good idea! 🙂
MJL21193: You have a point there, +-20mA from the VAS is enough to drive the output stage to +-40A even if the predrivers are bypassed while output devices and drivers are of the lowest beta... 😛
Last edited:
I think he could easily cut that down to more than half.
The VAS in my big sub amp is running at ~9mA feeding a triple with 8 pairs of outputs - no shortage of current.
The VAS in my big sub amp is running at ~9mA feeding a triple with 8 pairs of outputs - no shortage of current.
I didn't say that 27r4 (R18) was right or wrong.
I said the ratio of the resistors was too low.
10 to 15mA through the VAS may be sufficient, if the designer wants this level of VAS bias current.
Now let's look at the CCS biasing the VAS.
The 27r4 sets the bias current when the Vbe ~600mV. schematic states 22mA.
If the VAS at maximum output passes 22mA+-22mA then it can change from 0ma to 44mA over the whole cycle.
The protection transistor will see 440mV across the 10r and will have started to turn on, shorting out the LTP collector resistor.
This is very likely to affect sound quality.
I advocate that any protection should allow a valid audio signal to pass to a valid load.
I believe that to meet this requirement the VAS protection transistor should never have >=400mVbe. This requires the ratio of R18 to R24 to be >3:1
I said the ratio of the resistors was too low.
10 to 15mA through the VAS may be sufficient, if the designer wants this level of VAS bias current.
Now let's look at the CCS biasing the VAS.
The 27r4 sets the bias current when the Vbe ~600mV. schematic states 22mA.
If the VAS at maximum output passes 22mA+-22mA then it can change from 0ma to 44mA over the whole cycle.
The protection transistor will see 440mV across the 10r and will have started to turn on, shorting out the LTP collector resistor.
This is very likely to affect sound quality.
I advocate that any protection should allow a valid audio signal to pass to a valid load.
I believe that to meet this requirement the VAS protection transistor should never have >=400mVbe. This requires the ratio of R18 to R24 to be >3:1
Hello!
megajocke
The artifact is an overshoot:
AndrewT
Increased the vas current set resistor to 33R2 to give more headroom and also this allows VAS to run cooler.
BV
Added a BAV99 diode across the cap and the artifact dissapeared:
Is this diode safe for normal operation - doesn't it disturb the signal? And if the artifact is an overshoot, is it neccesary?
Thanks!
I'll try adding the output stage also, first with drivers only.
Until then,
Rait
megajocke
The artifact is an overshoot:
An externally hosted image should be here but it was not working when we last tested it.
AndrewT
Increased the vas current set resistor to 33R2 to give more headroom and also this allows VAS to run cooler.
BV
Added a BAV99 diode across the cap and the artifact dissapeared:
An externally hosted image should be here but it was not working when we last tested it.
Is this diode safe for normal operation - doesn't it disturb the signal? And if the artifact is an overshoot, is it neccesary?
Thanks!
I'll try adding the output stage also, first with drivers only.
Until then,
Rait
Andrew, what you wrote was:
What was it that made you think the VAS needed to deliver +-22mA? The triple darlingtion OPS used needs no more than about +-0.2mA of drive for +-40A of output current and Cdom needs no more than +-5mA from the VAS output. Thus "Q13 may start triggering on large VAS currents." is not true. Q13 Vbe won't even exceed 300mV.
What was it that made you think the VAS needed to deliver +-22mA? The triple darlingtion OPS used needs no more than about +-0.2mA of drive for +-40A of output current and Cdom needs no more than +-5mA from the VAS output. Thus "Q13 may start triggering on large VAS currents." is not true. Q13 Vbe won't even exceed 300mV.
Last edited:
Those have only 75V voltage rating. Did you use both diodes in series? It's still a bit low as the diode needs to be able to stand off the full rail-to-rail voltage. Also, two diodes in series reduces the clamping action greatly.
An alternative you could use instead is a resistor in the collector of the beta enhancer selected so that the amount of overdrive for the VAS transistor is limited to a sensible value. That would probably be enough.
The biggest problem without any of these measures is that a LOT of current flows through the beta enhancer when in negative clipping, degrading or even destroying it.
Last edited:
You must use diode with 200V reverse voltage, on yor scope screenshot is clearly visible, that 75V diode is acting as "zener" (rounded edges on positive clipping, lower positive output voltage). Diode as BAV21 is OK, without any negative influence on signal. Any saturation (long recovery time) in signal path means amplifier is without control, open NFB, full open loop gain, possible instabilities at recovery.... To avoid this, yes, diode is the simplest way and is necessary
you are doing it again.
I did not say the VAS needed 22mA of bias current.
I said the ratio of resistors was too low and when prompted showed how I arrived at a ratio that usually works to avoid early onset of loading the LTP output, i.e. triggering the protection transistor.
Oscilations are still present? If yes, try to remove Q13 and replace it with 3pcs 1N4148 conected in series (it is simplest way), first anode to base Q7, last cathode to former emitor Q13. Gain in local feedback (Q13) acting against overall NFB can cause oscilations. And lover current in VAS, increase resistors R18, R24 about 3 times. Faster transistors in VAS would be much better (replace MJE340/350)..
AndrewT:
Neither did I say that you said that. As usual you seem to be doing all that is in your power to deliberately misinterpret and misquote my posts. What I did say was that your calculation is based on that +-22mA of current delivery is needed, not that the CCS bias needs to be 22mA.
Those are entirely different things, except for this relationship:
CCS bias must be set equal to or greater than the maximum expected output current.
But you assumed that the relationship is like this:
CCS bias must be set equal to the maximum expected output current.
In the calculation you showed you made this assumption: the VAS needs to deliver +-22mA but this does not follow just because CCS current was set to 22mA. It is, of course, entirely possible to set VAS bias current to a greater value than what is strictly needed and it has a couple of small advantages actually. If the VAS only needs to deliver say +-15mA with its 22mA CCS current your 400mV limit won't be reached.
To be specific, what you did was that you said that the resistor values needed changing because:
"Q13 may start triggering on large VAS currents."
Thus you imply that the VAS, in this particular amplifier, has not got enough output current capability. It is an entirely different matter that a 1:3 ratio is a good design choice because of the things you explained. 🙂
In this case, 400mV over the protection resistor is 40mA sunk. 22mA was sourced by the CCS. That means 18mA could be sunk by the VAS without exceeding 400mV Vbe for the protection transistor. By saying "Q13 may start triggering on large VAS currents." you implied that this figure of 18mA was not enough.
Neither did I say that you said that. As usual you seem to be doing all that is in your power to deliberately misinterpret and misquote my posts. What I did say was that your calculation is based on that +-22mA of current delivery is needed, not that the CCS bias needs to be 22mA.
Those are entirely different things, except for this relationship:
CCS bias must be set equal to or greater than the maximum expected output current.
But you assumed that the relationship is like this:
CCS bias must be set equal to the maximum expected output current.
In the calculation you showed you made this assumption: the VAS needs to deliver +-22mA but this does not follow just because CCS current was set to 22mA. It is, of course, entirely possible to set VAS bias current to a greater value than what is strictly needed and it has a couple of small advantages actually. If the VAS only needs to deliver say +-15mA with its 22mA CCS current your 400mV limit won't be reached.
To be specific, what you did was that you said that the resistor values needed changing because:
"Q13 may start triggering on large VAS currents."
Thus you imply that the VAS, in this particular amplifier, has not got enough output current capability. It is an entirely different matter that a 1:3 ratio is a good design choice because of the things you explained. 🙂
In this case, 400mV over the protection resistor is 40mA sunk. 22mA was sourced by the CCS. That means 18mA could be sunk by the VAS without exceeding 400mV Vbe for the protection transistor. By saying "Q13 may start triggering on large VAS currents." you implied that this figure of 18mA was not enough.
Last edited:
show Raitraak how you would determine an optimal value for VAS & CCS bias current. Then all of us can learn.
If you keep the bias current low you can do without putting tehy bias transistor on the heatsink with the output transistors.
For higher bias currents you need the bias transistor on teh heatsink otehrwise thermal runaway occurrs.
Hey!
BV
Currently dealing with the input and VAS stage, no oscillation there.
Found some BAV21s, so this problem is solved. Couldn't imagine that it needs full rail to rail voltage spec.
Others
Let's leave the VAS current for now, if it's no problem for amp stablitity. Both VAS current source transistor and VAS transistor have proper heatsinking. Of course if someone could describe the current selection process in detail, it would be nice!🙂
Added output drivers and noticed some instability.
Current schematics show the actual state of the board: Output trannies are removed, also protection network is disconnected. The midpoint of driver stage current set resistor is connected to the output rail to close feedback loop. This time in pdf as it is too large.
SCHEMATICS
Used 10kHz test signal, output is 100Vpp.
Rising Edge
Falling edge
Does that mean there's too little compensation? Increase the 100pF comp. cap?
Regards,
Rait
BV
Currently dealing with the input and VAS stage, no oscillation there.
Found some BAV21s, so this problem is solved. Couldn't imagine that it needs full rail to rail voltage spec.
Others
Let's leave the VAS current for now, if it's no problem for amp stablitity. Both VAS current source transistor and VAS transistor have proper heatsinking. Of course if someone could describe the current selection process in detail, it would be nice!🙂
Added output drivers and noticed some instability.
Current schematics show the actual state of the board: Output trannies are removed, also protection network is disconnected. The midpoint of driver stage current set resistor is connected to the output rail to close feedback loop. This time in pdf as it is too large.
SCHEMATICS
Used 10kHz test signal, output is 100Vpp.
An externally hosted image should be here but it was not working when we last tested it.
Rising Edge
An externally hosted image should be here but it was not working when we last tested it.
Falling edge
An externally hosted image should be here but it was not working when we last tested it.
Does that mean there's too little compensation? Increase the 100pF comp. cap?
Regards,
Rait
- Status
- Not open for further replies.