I am NOT insisting on "current mirror" solution. However I must write that explanations to you understand why did I replace resistor with current mirror.
I must warn you again: I am not a solid state professor. I came here from chipamp forum. Moreover I was not even a chipamp guru as you know.
I must warn you again: I am not a solid state professor. I came here from chipamp forum. Moreover I was not even a chipamp guru as you know.
Perhaps when using the emitter resistors it would be best to balance the collector resistors also.
For general information on the input stage refer to: http://www.dself.dsl.pipex.com/ampins/balanced/balanced.htm#7
In Fig 9 on that page Mr. Self recommends the resistors together with the current mirror.
For general information on the input stage refer to: http://www.dself.dsl.pipex.com/ampins/balanced/balanced.htm#7
In Fig 9 on that page Mr. Self recommends the resistors together with the current mirror.
A few observations:
If you use current mirror loading of the input LTP, in real life your current mirror transistors will be different from each other (unlike in a simulator). Because of this they need emitter degeneration or the LTP will again be unbalanced (and may be quite seriously so). Chose a resistor value that results in at least 25mV and preferably 100mV voltage drop with half of the LTP tail current.
Consider using a current source for the tail current of the LTP. This is the more necessary the more the input voltages expected approach an order of magnitude less than the power rail (this sort of thing generates common mode distortion). Power supply modulation will also produce this, but with 100Hz ripple added. If you want to continue using a simple resistor for the tail of the LTP, consider using two resistors in series, with the one from the rail itself being a smaller portion of the total. Then from the joining point, you can put a large cap to ground, or a cap in parallel with a suitable zener diode. This will filter and/or regulate the voltage for the tail supply of the LTP.
Some people swear on bootstrap 'surrent sources' for the VAS, and your amp being class A avoids one of the main pitfalls of bootstrapping - you might want to try it. Keep in mind the bootstrap cap needs to be premium quality and at least 10x larger than what you get using F(lower) to calculate it. In this application bootstrapping may have a certain advantage, which is the ability to swing the VAS output above the positive supply rail. Since you are losing about 4V off your maximum swing due to MOSFET Vgs, this can be an advantage, alowing the output to go near the positive power rail. Remember to put a zener diode across G and S of the MOSFET to limit the maximum Vgs on clipping. The zener is best placed on the MOSFET legs, after the gate damper resistor.
Your darlington current source may work well in the simulation, but it will require a well chosen darlington, or you might end up outside the SOA for it. I've had good results with a MOSFET 'ring of two' current source in this sort of application, especially considering it cah swing all the way down to about 0.7V off the negative rail. Ring-of-two would be the connection you use for your VAS current source.
If you continue using a darlington current source as youa re doing now, replace the 3.3V zener with two red LEDs in series or one blue LED. Low voltage zeners are quite noisy, whereas LEDs, being forward biassed, are considerably less so. If you use bootstrapping for the VAS, in order to maximize swing, you may want to lower the reference voltage for this current source as much as possible, but do not go below some 0.25V (0.5V would be better) on it's emitter resistor. As it stands now, you may even want to increase the voltage drop on the emitter resistor to get symetrical clipping.
If you use current mirror loading of the input LTP, in real life your current mirror transistors will be different from each other (unlike in a simulator). Because of this they need emitter degeneration or the LTP will again be unbalanced (and may be quite seriously so). Chose a resistor value that results in at least 25mV and preferably 100mV voltage drop with half of the LTP tail current.
Consider using a current source for the tail current of the LTP. This is the more necessary the more the input voltages expected approach an order of magnitude less than the power rail (this sort of thing generates common mode distortion). Power supply modulation will also produce this, but with 100Hz ripple added. If you want to continue using a simple resistor for the tail of the LTP, consider using two resistors in series, with the one from the rail itself being a smaller portion of the total. Then from the joining point, you can put a large cap to ground, or a cap in parallel with a suitable zener diode. This will filter and/or regulate the voltage for the tail supply of the LTP.
Some people swear on bootstrap 'surrent sources' for the VAS, and your amp being class A avoids one of the main pitfalls of bootstrapping - you might want to try it. Keep in mind the bootstrap cap needs to be premium quality and at least 10x larger than what you get using F(lower) to calculate it. In this application bootstrapping may have a certain advantage, which is the ability to swing the VAS output above the positive supply rail. Since you are losing about 4V off your maximum swing due to MOSFET Vgs, this can be an advantage, alowing the output to go near the positive power rail. Remember to put a zener diode across G and S of the MOSFET to limit the maximum Vgs on clipping. The zener is best placed on the MOSFET legs, after the gate damper resistor.
Your darlington current source may work well in the simulation, but it will require a well chosen darlington, or you might end up outside the SOA for it. I've had good results with a MOSFET 'ring of two' current source in this sort of application, especially considering it cah swing all the way down to about 0.7V off the negative rail. Ring-of-two would be the connection you use for your VAS current source.
If you continue using a darlington current source as youa re doing now, replace the 3.3V zener with two red LEDs in series or one blue LED. Low voltage zeners are quite noisy, whereas LEDs, being forward biassed, are considerably less so. If you use bootstrapping for the VAS, in order to maximize swing, you may want to lower the reference voltage for this current source as much as possible, but do not go below some 0.25V (0.5V would be better) on it's emitter resistor. As it stands now, you may even want to increase the voltage drop on the emitter resistor to get symetrical clipping.
Gordy said:
In the above message I quoted the wrong page. It should be http://www.dself.dsl.pipex.com/ampins/dipa/dipa.htm#5
Dxvideo said:
A current mirror load is a great improvement over resistive loading. It forces the long tail pair (LTP) collector currents into balance, improves the open loop linearity and gain and provides a huge improvement in power supply rejection with respect to modulation of the LTP tail current. This is important if you don’t want to go with the extra complexity of a constant current source to provide the LTP tail current.
A current mirror load, used in conjunction with miller compensation will increase the localised negative feedback around the voltage amplifier stage (VAS) transistor. This helps to linearise the VAS and lowers its output impedance at high frequencies. This would be of benefit in your design, as the input capacitance of the output MOSFET source follower is driven directly from the VAS.
In a design like this, I'd implement a hard current limit on the driven output devices at 2*Iq. This will help with symetrical clipping when the load impedance dips.
Cheers,
Glen
In order to reduce thermal stress and distortion in Q5, may I suggest something more robust, like BF869. If implemented, you maybe want to increase its current to at least 20mA. I would also insert an emitter resistor, mainly to improve linearity.
As proposed earlier, you should get rid of that Darlington device and zener diode, go for FET and a friendlier voltage reference.
The input stage is still poorly designed.
As proposed earlier, you should get rid of that Darlington device and zener diode, go for FET and a friendlier voltage reference.
The input stage is still poorly designed.
Ozgur,
In its present form, it will work fine. Built the !@#$ thing, listen to it, then make the changes suggested here, one by one, listening as you go.
If you are guided by advice from ten people, you will naturally become confused, particularly if you are starting out. There is simply no short cut.
Building and listening is the only way I know to understand how the schematic relates to the sound. I can tell you many of these connections, but it's only words, and you need to hear for yourself.
If you do this, and keep notes, you will be discovering things outside the normal text book knowledge held by most EEs, and well into the sorts of techniques used by real world designers like NP and John Curl. I advise you to become very familiar with a well recorded test track/CD, and use it over and over when testing. It's boring, but it's real music, and if you choose lots of different genres, you will be doing exactly what the consumer does.
In my view, this is what DIY is really about; making that very difficult connection between the schematic and the sound.
Cheers,
Hugh
In its present form, it will work fine. Built the !@#$ thing, listen to it, then make the changes suggested here, one by one, listening as you go.
If you are guided by advice from ten people, you will naturally become confused, particularly if you are starting out. There is simply no short cut.
Building and listening is the only way I know to understand how the schematic relates to the sound. I can tell you many of these connections, but it's only words, and you need to hear for yourself.
If you do this, and keep notes, you will be discovering things outside the normal text book knowledge held by most EEs, and well into the sorts of techniques used by real world designers like NP and John Curl. I advise you to become very familiar with a well recorded test track/CD, and use it over and over when testing. It's boring, but it's real music, and if you choose lots of different genres, you will be doing exactly what the consumer does.
In my view, this is what DIY is really about; making that very difficult connection between the schematic and the sound.
Cheers,
Hugh
Dxvideo said:
To ease the confusion a bit, and to expand upon your theoretical knowledge, I suggest that you sit down and read through the free wealth of information found at the link provided to you in this post:
http://www.diyaudio.com/forums/showthread.php?postid=1258508#post1258508
It addresses most of the design points raised in the replies you have received thus far; and it is written by a very experienced “real world” professional audio designer.
Cheers,
Glen
Ozgur,
This is the problem. No one agrees. There are two camps; the subjectivists, and the objectivists, with both fighting bitterly and many harsh words traded. The best position is a foot in both camps; this is balance, and it should be won with experience.
Furthermore, the technical aspects can be so mathematical and require such extensive training that the whole mess can easily become rather academic, without much reference to what sounds good. It is possible to have two medical Professors arguing about what is wrong with the patient, generally only one will be right!!
That is why if you build it you will assume theory and practice in an appropriate balance without the confusion of two or three different teachers. Then, once you've heard it, start tinkering, and making the various changes suggested.
To assume there is only one way is not appropriate; there are many ways to skin a cat, and each has advantages and disadvantages, like front wheel drive or rear wheel drive on an automobile.
It is a simple circuit, and it sounds very good. Breadboard it, and have a listen!
Cheers,
Hugh
This is the problem. No one agrees. There are two camps; the subjectivists, and the objectivists, with both fighting bitterly and many harsh words traded. The best position is a foot in both camps; this is balance, and it should be won with experience.
Furthermore, the technical aspects can be so mathematical and require such extensive training that the whole mess can easily become rather academic, without much reference to what sounds good. It is possible to have two medical Professors arguing about what is wrong with the patient, generally only one will be right!!
That is why if you build it you will assume theory and practice in an appropriate balance without the confusion of two or three different teachers. Then, once you've heard it, start tinkering, and making the various changes suggested.
To assume there is only one way is not appropriate; there are many ways to skin a cat, and each has advantages and disadvantages, like front wheel drive or rear wheel drive on an automobile.
It is a simple circuit, and it sounds very good. Breadboard it, and have a listen!
Cheers,
Hugh
With 2 amps bias current through the output devices they will of course be dissipating 48 watts each. Assuming you use a TO-220 package for each, they only have a small footprint to transfer that much heat. You will probably find that if you use an insulating washer between the device and heatsink the heat transfer will not be good enough. Better to attach the devices directly to their own =separate= heatsinks and isolate them from ground. Be aware though that because the lower device collector is connected to the heatsink, the entire heatsink will radiate at audio frequencies and may cause stability and feedback problems if it is too close to the input circuitry.
G.Kleinschmidt said:
I wish I would have enoughly time to learn. However I am too old to learn something.
Anyway, There must be a exact way to have a simple and robust input stage. I have just one simple question;
Does my input stage need a single resistor load or a current mirror!
I am not asking the meaning of life!!!!
Thx.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.