How can it be done?
I'm just an amateur with osciloscope and signal generator in hand. I don't have any lab-measurement devices. How can I do this?
At first, I think of 2 possibilities to do this.
ONE is not having any caps on the circuit (Wouldn't that caps introduce phase shift that will lead to non-linearity behavior?). This may be contradictive. In handbooks by Self or Slone, it is writen the importance of Miller cap (CC on VAS), to limit openloop bandiwith. But in this term of "linearity on every stage", would it be that having this CC on VAS will produce non-linearity?
TWO is based on Mr.Pass articles. He stated that in Mosfets, linearity is obtained by having as much current as possible (having big standing current). But this is for Mosfets. Aleph amps have about 20mA for the differential with mosfets.
I dont know how this can be implemented with bipolars. Bipolars have Hfe figure, the more current at the collector will need bigger current to fed the base. Maybe this will make the input not so sensitive, because it needs bigger current to fed the base.
Is this "having big standing current" implementable in Bipolars?
What is the guidance for designing power amp, to have "the best possible linearity of every stage "?
This is about QSC design. They use alot of this "Transnova" circuit. One thing interest me, that for a certain output power, QSC uses less output transistor than other brand. Is this "Transnova" circuit makes the whole power amp needs less output transistor than ordinary configuration? Will the output transistors works "more relaxed" than ordinary amp?
I'm just an amateur with osciloscope and signal generator in hand. I don't have any lab-measurement devices. How can I do this?
At first, I think of 2 possibilities to do this.
ONE is not having any caps on the circuit (Wouldn't that caps introduce phase shift that will lead to non-linearity behavior?). This may be contradictive. In handbooks by Self or Slone, it is writen the importance of Miller cap (CC on VAS), to limit openloop bandiwith. But in this term of "linearity on every stage", would it be that having this CC on VAS will produce non-linearity?
TWO is based on Mr.Pass articles. He stated that in Mosfets, linearity is obtained by having as much current as possible (having big standing current). But this is for Mosfets. Aleph amps have about 20mA for the differential with mosfets.
I dont know how this can be implemented with bipolars. Bipolars have Hfe figure, the more current at the collector will need bigger current to fed the base. Maybe this will make the input not so sensitive, because it needs bigger current to fed the base.
Is this "having big standing current" implementable in Bipolars?
What is the guidance for designing power amp, to have "the best possible linearity of every stage "?
This is about QSC design. They use alot of this "Transnova" circuit. One thing interest me, that for a certain output power, QSC uses less output transistor than other brand. Is this "Transnova" circuit makes the whole power amp needs less output transistor than ordinary configuration? Will the output transistors works "more relaxed" than ordinary amp?
Actually it is better to use other ways of stabilising gain and limit phase shift then to use Miller compensation. Using Miller gain compensation the amplifier need to give higher low frequency open loop gain then with other methods in order to achieve the same open loop perfomance, (bandwidth). Other methods involve more complicated stabilising networks and may not be so easy to adapt for the DIYer.
Some Opamp application notes include some info about other ways of stabilising amplifiers otherwise there are a lot of other litterature regarding how to stabilise feedback amplifiers.
By using more optimal ways of stabilistation it is possible to make a more simple amplifier with less gain stages and still achieve the same level of feedback in the audio range, this can possible lead to better performance, I think there are some audio amplifiers on the market where these ideas are implemented.
Regards Hans
Re: How can it be done?
A transistor has a very non-linear voltage-dependant capacitance
between base and collector. By paralleling this by an ordinary and
much bigger capacitor you rather improve linearity, since the non-linear
capacitance will largely be swamped by the external capacitor.
However, the purpose of the compensation capacitor is usually not
mainly to improve linearity but to create a well known low-frequency dominant pole in the transfer function of the amp. If you
just want to minimize the Miller effect (and especially the non-linear
intrinsic capacitance) it may be better to cascode the transistor,
for instance.
The major source of non-linearities in amplifiers is the inherent
non-linear nature of active components. While they may perform
better of worse on secondary effects, the primary effects are
inherent and unavoidable. A BJT has en exponential IC vs. Vbe
transfer fucntion, a JFET a square function etc. This is in their nature
and cannot be avoided. Various topologies can minimze the effects
of this, however. For instance, an emitter follower has local NFB
loop, which makes it more linear but also takes away all voltage
gain. A differential pair is very linear within a small range of input
voltages, but is very non-linear when going outside this range.
lumanauw said:
A transistor has a very non-linear voltage-dependant capacitance
between base and collector. By paralleling this by an ordinary and
much bigger capacitor you rather improve linearity, since the non-linear
capacitance will largely be swamped by the external capacitor.
However, the purpose of the compensation capacitor is usually not
mainly to improve linearity but to create a well known low-frequency dominant pole in the transfer function of the amp. If you
just want to minimize the Miller effect (and especially the non-linear
intrinsic capacitance) it may be better to cascode the transistor,
for instance.
The major source of non-linearities in amplifiers is the inherent
non-linear nature of active components. While they may perform
better of worse on secondary effects, the primary effects are
inherent and unavoidable. A BJT has en exponential IC vs. Vbe
transfer fucntion, a JFET a square function etc. This is in their nature
and cannot be avoided. Various topologies can minimze the effects
of this, however. For instance, an emitter follower has local NFB
loop, which makes it more linear but also takes away all voltage
gain. A differential pair is very linear within a small range of input
voltages, but is very non-linear when going outside this range.
phase_accurate said:
Yes, that is true, and I agree it was somwhat sloppy and possibly
misleading of me to only mention the Vbe vs. Ic characteristic.
On the other hand, it is not always possible to have sufficient
control over the source impedance so it can be considered a current source. Besides, the source will typically be a voltage, at least for
the input stage, so the non-linear variations in Vbe voltage will
affect the base current caused by the source.
"It is an advantage, but minor in my view. The downside is that the power supplies float up and down with the signal, which may give all sorts of problems with parasitic capacitances."
Crown went to the grounded ouput stage about 30 years ago.
Their stated reason was for ease in implementing their new ODEP protection circuit.
Stray capacitance from large power supply components to the chassis should be minimized.
The transformer should have split bobbin construction or a screen.
Do not mount the rectifiers or filter caps straight to the chassis.
QSC follows these same ideas.
Crown went to the grounded ouput stage about 30 years ago.
Their stated reason was for ease in implementing their new ODEP protection circuit.
Stray capacitance from large power supply components to the chassis should be minimized.
The transformer should have split bobbin construction or a screen.
Do not mount the rectifiers or filter caps straight to the chassis.
QSC follows these same ideas.
Since cascode is in the middle of the topic, I have a question about this. Reading the article of Mr.Pass about cascode, Mr Borbely, it shows that cascoding can reduce input capacitance.
How can this happen? What is the explenation?
Although Mr.Pass has his own article about the merit of cascoding, why is it he seldom use this configuration? Some of his design using cascoding is intended to split the heat dissipation.
Is cascode not good enough so it is better not to use it if we are not force to (by thermal /voltage considerations)?
How can this happen? What is the explenation?
Although Mr.Pass has his own article about the merit of cascoding, why is it he seldom use this configuration? Some of his design using cascoding is intended to split the heat dissipation.
Is cascode not good enough so it is better not to use it if we are not force to (by thermal /voltage considerations)?
lumanauw said:
Emitter resistors do this, or a CCS. If you change the base
voltage, this causes a change in the emitter current which causes
a voltage change over the emitter resistor that counteracts the
voltage change at the base. As a net result you get only a small
change in Vbe.
lumanauw said:
It doesn't reduce the capacitance, but the effect of it. Normally
if you have a collector load you get a voltage swing on the
collector. Typically you get a similar voltage swing between
collector and base. The capacitance between collector and base
then acts as a local AC feedback path, where the feedback
increases with frequency, which reduces the frequency response
considerably. To make things worse, this capacitance it non-linear.
If you cascode, you fix the collector voltage at a DC level. You get
only very small AC voltages at the base and the collector,
so there will be almost no AC voltage over the capacitor to
generate a feedback current.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.