Hi Bas,
Do you aware that Cambridge A500 uses SAP15N/P? This is the manual http://x0546.orbbox.com/files/stora...ublicsession=1ae486d482ef86e98e56c2ba1a6b3a08
Panson
Do you aware that Cambridge A500 uses SAP15N/P? This is the manual http://x0546.orbbox.com/files/stora...ublicsession=1ae486d482ef86e98e56c2ba1a6b3a08
Panson
Hi Bas,
If base stopper resistors are used to avoid high-freq oscillation, did you try to use a ferrite bead or a little ferrite core in series with the base?
Regards,
Panson
oscillation vs loading cap value
The book "Intutitive Analog Circuit Design" derives input impedance of a capacitive loaded emitter follower. When the loading cap value is small, the input impedance exhibits negative resistance. In contrast, large loading cap does not cause the input showing negative resistance. Base resistor is commonly used to offset the negative resistance. I tested the amp (49811+03N/P) with small 270 pF and large 2 uF load cap. The result is consistent with the book. I got oscillation with 270 pF load cap (10 R base resistors installed), but not for 2 uF. Increasing the base resistor to 240 R does not stop the oscillation. I then put 1.5 k which stops the oscillation for 270 pF capacitive load.
The book "Intutitive Analog Circuit Design" derives input impedance of a capacitive loaded emitter follower. When the loading cap value is small, the input impedance exhibits negative resistance. In contrast, large loading cap does not cause the input showing negative resistance. Base resistor is commonly used to offset the negative resistance. I tested the amp (49811+03N/P) with small 270 pF and large 2 uF load cap. The result is consistent with the book. I got oscillation with 270 pF load cap (10 R base resistors installed), but not for 2 uF. Increasing the base resistor to 240 R does not stop the oscillation. I then put 1.5 k which stops the oscillation for 270 pF capacitive load.
and if it wasn't an integrated Darlington, you could try base stoppers on both the driver and output. How about 47r and 2r2 instead of 1k5
Dear Panson,
That is a very kind offer from you. Thank you in advance. Please send it to bas_music@yahoo.com
I think the sanken's really need the cap's from collector to base to stay stable. Without I never got the system really stable. But maybe with some practice you can do it. In my subjective listening sessions, The capacitors from collector to base made the sound better (warmer and less sterile) and the base stoppers make the sound dull and it lost it's "magic" So I would try to stay without the base stoppers but go with the cap's. The cap's between the source and sink really improve the performance. The bigger the electrolytic the better it sounded in my setup.
With kind regards,
Bas
Dear Panson,
I want to add, that if you increase the (miller) compensation of the driver chip a bit the problems would be far less. And don;t forget the 3.3K and 75P to ground.
And yes you are right about the other things. If I loaded my first design with the STD03's with bigger value cap everything was fine, but small capacitance as result from mixed speaker wires caused oscillation.
With kind regards,
Bas
I want to add, that if you increase the (miller) compensation of the driver chip a bit the problems would be far less. And don;t forget the 3.3K and 75P to ground.
And yes you are right about the other things. If I loaded my first design with the STD03's with bigger value cap everything was fine, but small capacitance as result from mixed speaker wires caused oscillation.
With kind regards,
Bas
Thanks Panson, Yes I already had this schematic. In this case they put the 10uF capacitor between the diode string, but I believe it does the same trick to stabilize.
With kind regards,
Bas
Dear Panson,
Sorry for the raining of post's... 😀 If it can be of any help, this is my final circuit that is tested with all kind of loads's and works super stable under all possible circumstances.
Note, the supply rails is relatively low, because we needed more current headroom in lower impedances. In this setup the STD3's can double output power into 4 ohm which was very important for this application. The base stoppers are zero ohm links and not used after all. Also not visible on this part of my schematic, each supply pin from as well the driver as the darlingtons are de-coupled with a 100nF ceramic SMD cap.
With kind regards,
Bas
Sorry for the raining of post's... 😀 If it can be of any help, this is my final circuit that is tested with all kind of loads's and works super stable under all possible circumstances.
Note, the supply rails is relatively low, because we needed more current headroom in lower impedances. In this setup the STD3's can double output power into 4 ohm which was very important for this application. The base stoppers are zero ohm links and not used after all. Also not visible on this part of my schematic, each supply pin from as well the driver as the darlingtons are de-coupled with a 100nF ceramic SMD cap.
With kind regards,
Bas
Attachments
Last edited:
Hi Bas,
I just e-mail the schematic to you.
The 1.5k base stopper solves the oscillation problem with small load cap. But, the test unit then oscillates with 2 u load cap. EF output inductive reactance is proportional to the stopper. It is larger with a 1.5 k stopper. The test amp may become a oscillator at freq different from that of 220pF load cap. Need to check the freq today. I probably should try e.g. 1 k to see whether it is the optimum value.
I agree with you that C-B cap is an effective oscillation killer. On the other hand, I will use discrete Darlington to learn more the oscillation issue. Is it also happening in discrete form? My goal is to use as less cap as possible.
Cheers,
Panson
Last edited:
Dear Panson,
I see that the Musical Fidelity A1 2008 edition has a cap between the bases as well. Since I've seen this cap in all designs the the SAP and STD devices I think this is a thing you should accept. I have no experience with other darlington's, but with other tripple darlington circuits, and my experience is there is always the danger of oscillation if one not careful compensate and choose a careful layout. If the amplifier was for myself only I would't mind to let it operate on the edge and omit as much as possible capacitors. But if the design is for a commercial/DIY market, you never know what the end user will do with it under what circumstances, and then I think better be safe then sorry at a certain cost of absolute fidelity.
Btw. I forgot to askm which supply you rung the STD03's on? My experience was as well the higher the supply voltage the more risk on oscillation.
I think the high value base stoppers is no go and that should't be the solution. Those resistors highly slow down the system, and even a few ohm already degrade the subjective sound performance in my opinion. If the cap between the bases is the solution I would go for that if I was you. It also improve the switching behavior of the transistor pair.
I wonder why Musical fidelity put the extra diodes in series with the diode string. Must be to reach the Class A level bias? I have no doubt the A1 is a reliable amplifier, but according to the SAP application note, the thermal tracking get less stable with bias idling above 40mA. D If the A is a truly 20 watt Class A amplifier the bias current must be way above 40mA total Re. Do I have your permission to put this schematic online Panson, so others can give their insights as well?
With kind regards,
Bas
I see that the Musical Fidelity A1 2008 edition has a cap between the bases as well. Since I've seen this cap in all designs the the SAP and STD devices I think this is a thing you should accept. I have no experience with other darlington's, but with other tripple darlington circuits, and my experience is there is always the danger of oscillation if one not careful compensate and choose a careful layout. If the amplifier was for myself only I would't mind to let it operate on the edge and omit as much as possible capacitors. But if the design is for a commercial/DIY market, you never know what the end user will do with it under what circumstances, and then I think better be safe then sorry at a certain cost of absolute fidelity.
Btw. I forgot to askm which supply you rung the STD03's on? My experience was as well the higher the supply voltage the more risk on oscillation.
I think the high value base stoppers is no go and that should't be the solution. Those resistors highly slow down the system, and even a few ohm already degrade the subjective sound performance in my opinion. If the cap between the bases is the solution I would go for that if I was you. It also improve the switching behavior of the transistor pair.
I wonder why Musical fidelity put the extra diodes in series with the diode string. Must be to reach the Class A level bias? I have no doubt the A1 is a reliable amplifier, but according to the SAP application note, the thermal tracking get less stable with bias idling above 40mA. D If the A is a truly 20 watt Class A amplifier the bias current must be way above 40mA total Re. Do I have your permission to put this schematic online Panson, so others can give their insights as well?
With kind regards,
Bas
Last edited:
Hi Bas,
Please feel free to put the schematic online.
The writer of the article measured the bias current of the new A1. It is 680 mA +/- 10 mA. The new A1 is indeed a deeply biased class AB. He also mentioned the original A1 bias current equal to 800 mA.
The supply of my test amp is +/- 38 V.
Panson
Please feel free to put the schematic online.
The writer of the article measured the bias current of the new A1. It is 680 mA +/- 10 mA. The new A1 is indeed a deeply biased class AB. He also mentioned the original A1 bias current equal to 800 mA.
The supply of my test amp is +/- 38 V.
Panson
Hi Bas,
I definitely agree with you.
Your latest schematic shows no output coil. Is it also true for your hardware? With the output coil and damping resistor, I don't get oscillation problem. I think in practice an output coil is essential for robustness.
Panson
Dear Panson,
Does this mean all your oscillation problems are solved now with adding the output inductor and resistor?
And I am agree with you, that in any amplifier for general purpose this coil and resistor should be included. However, the last schematic I provided is part of an active 3 way system, where two STD03 amplifiers are used for the tweeter and midrange. Because we know the load, and because there isn't a passive crossover and because of the speaker wires are really short we can omit the output inductor and resistor in this particularly case.
With kind regards,
Bas
Bias
Dear Panson and other readers,
Since this is a solid STD03 tread, I think this stuff is also useful for other readers. I've seen many other STD03 DIY and commercial designs where the pre bias isn't right, or get measured the wrong way. This degrade the stability and thermal tracking. You can change the idle bias more forgiving, but the pre bias must be a truth fixed 2.5mA and as precise as possible.
It is hard to measure exact current, so here my way to measure and to be sure the pre bias current is 2.5mA
Who reads the data-sheet carefully knows that the diodes in the STD03N will have a forward voltage of 705mV by a forward current of 2.5mA
The diodes in the STD03P will have a forward voltage of 1540 by a forward current of 2.5mA.
So to set the pre bias on exactly 2.5mA do as follow:
1: Set the pot-meter (the one between the diode string) for the idle bias all the way to 0 ohm
2: Measure the voltage across the two bases of the STD03N/P
3: The reading should be 2.25VDC
4: Now after warming up adjust the idle bias current to 40-60mA.
It is very advisable to make the pre bias adjustable with a potential meter to avoid variations.
With kind regards,
Bas
Dear Panson and other readers,
Since this is a solid STD03 tread, I think this stuff is also useful for other readers. I've seen many other STD03 DIY and commercial designs where the pre bias isn't right, or get measured the wrong way. This degrade the stability and thermal tracking. You can change the idle bias more forgiving, but the pre bias must be a truth fixed 2.5mA and as precise as possible.
It is hard to measure exact current, so here my way to measure and to be sure the pre bias current is 2.5mA
Who reads the data-sheet carefully knows that the diodes in the STD03N will have a forward voltage of 705mV by a forward current of 2.5mA
The diodes in the STD03P will have a forward voltage of 1540 by a forward current of 2.5mA.
So to set the pre bias on exactly 2.5mA do as follow:
1: Set the pot-meter (the one between the diode string) for the idle bias all the way to 0 ohm
2: Measure the voltage across the two bases of the STD03N/P
3: The reading should be 2.25VDC
4: Now after warming up adjust the idle bias current to 40-60mA.
It is very advisable to make the pre bias adjustable with a potential meter to avoid variations.
With kind regards,
Bas
Bas,
The whole point of the diode stack is to apply a constant voltage between the bases. "Constant" is a relative term here as everybody knows (or should know) that the voltage across a diode depends on temperature. But as the Vbe's also vary with temperature, the "constant" voltage across the diodes set up a constant quiescent current in the output transistors.
However, the diodes aren't perfect. They have some dynamic impedance and other non-idealities that prevent them from being perfect, ideal voltage sources. As a result, the audio signal applied at the base of the output transistors will cause a little bit of AC current to flow in the diodes. Recall, that an emitter follower has a voltage gain of slightly less than unity. Thus, the voltage applied at the base is roughly equivalent to the output voltage. This is pretty big compared to the 2.5-ish volt across the diode stack. So... The current through the diode stack varies slightly (both from the diode non-idealities, and from the fact that an increasing load current on the amp output will cause higher base current to flow --> less current available for diode stack) and, hence, the bias voltage across the diode stack varies slightly as function of the input signal voltage vs time. This means the quiescent point for the output devices changes ever so slightly as the input signal changes. It should be fairly intuitive that this has the potential for causing all sorts of harmonic mixing and, thus, harmonic distortion.
The 10 uF cap from base to base (across the diode stack) makes it so that the bias voltage between the bases doesn't change significantly when audio frequencies are applied to the amp input. This does lower the THD. I measured about an order of magnitude reduction in THD with/without cap. I.e. from 0.0x % THD without cap to 0.00x % THD with the cap in place.
As far as I know the cap doesn't do anything for the overall amp stability. It may help a little with bias thermal stability, but that's really a guess. I chose to add the cap because of the 10-fold reduction in THD.
~Tom
The whole point of the diode stack is to apply a constant voltage between the bases. "Constant" is a relative term here as everybody knows (or should know) that the voltage across a diode depends on temperature. But as the Vbe's also vary with temperature, the "constant" voltage across the diodes set up a constant quiescent current in the output transistors.
However, the diodes aren't perfect. They have some dynamic impedance and other non-idealities that prevent them from being perfect, ideal voltage sources. As a result, the audio signal applied at the base of the output transistors will cause a little bit of AC current to flow in the diodes. Recall, that an emitter follower has a voltage gain of slightly less than unity. Thus, the voltage applied at the base is roughly equivalent to the output voltage. This is pretty big compared to the 2.5-ish volt across the diode stack. So... The current through the diode stack varies slightly (both from the diode non-idealities, and from the fact that an increasing load current on the amp output will cause higher base current to flow --> less current available for diode stack) and, hence, the bias voltage across the diode stack varies slightly as function of the input signal voltage vs time. This means the quiescent point for the output devices changes ever so slightly as the input signal changes. It should be fairly intuitive that this has the potential for causing all sorts of harmonic mixing and, thus, harmonic distortion.
The 10 uF cap from base to base (across the diode stack) makes it so that the bias voltage between the bases doesn't change significantly when audio frequencies are applied to the amp input. This does lower the THD. I measured about an order of magnitude reduction in THD with/without cap. I.e. from 0.0x % THD without cap to 0.00x % THD with the cap in place.
As far as I know the cap doesn't do anything for the overall amp stability. It may help a little with bias thermal stability, but that's really a guess. I chose to add the cap because of the 10-fold reduction in THD.
~Tom
Thank you Tom for this good explanation. It makes me understand the function of the cap. Did you also experiment with different values?
With kind regards,
Bas
Yes, it does. I can remove all caps (cap across B pins, cap across C and B) now.
We should look at the voltage across the diode stack or differential voltage across base pins instead of the base to ground voltage. The differential voltage across base pins (or across the diode stack) is nearly constant. It can be verified by a DVM in AC mode measuring the base to base AC voltage. The amp input can be 1 kHz generated by a PC sound card. You won't see the level comparable to the output signal. You may see mV over there.
Dynamical resistance of a forward bias diode is less than one ohm. For instance, a 1N4148 dynamic resistance at 2.5 mA bias point is about 50 mR. The 03N/P diode dynamic resistance will be more or less the same. This tiny resistance will not cause any bias point variation during operation. Even it does. It will be negligible compared to that caused by the trim pot, thermal tracking error.
A 10 uF cap impedance is 16 Ohm at 1 kHz and 0.796 Ohm at 20 kHz. It is higher than the dynamic resistance of the diode stack.
In my test unit, I don't see any difference between using and not using a cap (470 uF || 0.1 uF) across the base pins (diode stack and trim pot included).
Attachments
Last edited:
I did not experiment with different values, I just slapped 10 uF on there and called it good.
Bottom line is that the cap needs to make the voltage across the diode stack constant at audio frequencies. So figure it should set a pole no higher than 1~2 Hz.
The input impedance of the bipolar transistor should be on the order of (beta)*Re. With two pairs of transistors in parallel, this becomes:
8000*0.22*0.5 = 880 ohm
2 Hz cutoff: 1/(2*pi*R*C) < 2 Hz --> C(min) = 1/(2*pi*880*2) = 0.90 uF.
Note that I'm using the high-beta versions of the devices. The minimum beta is about 8000. The factor of 0.5 is because the input impedance of the two pairs in parallel is 0.5x that of just one transistor.
~Tom
Odd... I included the cap because I wanted the voltage across the diode stack to remain constant. I've also seen commercial designs (Parasound A23) that used the cap. Theory says the voltage should be constant. Good commercial guys put it there... So I put it in there. You don't like it? Take it out... Easy... 🙂 It's not mission critical.
Thanks for sharing your data, though. Data (usually) doesn't lie.
~Tom
- Status
- Not open for further replies.