| lumanauw |
In many writings that can be found on the net, many stated the advantages of having a low value of openloop of a power amp. This when combined with sufficient feedback will result in good reproduction power amp.
First I don't understand this. But after looking at the schematic of "Hi-End" power amp from Motorola, it seems that this low-openloop gain is achieved by putting "just enough" gain on every stage of the power amp.
The gain is roughly can be seen by RC/RE, and in Differential stage and VAS stage, Motorola put combinations of RC/RC that makes the openloop gain is not so big.
The most obvious one is in the VAS, where usually the collector of the VAS transistor dont have RC, but to lower gain, they put RC to ground, so the gain of this VAS is lower than if it hasn't RC.
This low openloop gain theory is also strengthen by Mr.Pass design on Aleph amp. Although it don't have such as RC/RE that can be calculated, but he stated that the mosfet itself has low gain even on open loop.
I have a few question about this theory
1. If I use the output stage of CFP instead of EF, wouldn't that CFP has already gain in it? So it is impossible to have low open loop gain with CFP output?
2. What happens if we have a openloop gain lower than the closed loop gain? (like we design the open loop of 50x, but put the feedback arrangement of 100x), what happens? Will this amp can control the DC offset, inspite of the openloop VS closedloop conditions like above? What will be the gain of the closed loop?
3. What makes the sound tends to be better in low openloop audio power amp? (At least that is what some people said). Is making an openloop as big as possible then limit them by feedback resistors in closed loop is a bad thing? Some designs do this, what is the merits and drawbacks VS the low openloop design? What are the audible difference? Will it influence damping factor or slewrate? |
|
|
| Tube_Dude |
| quote: | Originally posted by lumanauw
In many writings that can be found on the net, many stated the advantages of having a low value of openloop of a power amp. This when combined with sufficient feedback will result in good reproduction power amp.
First I don't understand this. But after looking at the schematic of "Hi-End" power amp from Motorola, it seems that this low-openloop gain is achieved by putting "just enough" gain on every stage of the power amp.
The gain is roughly can be seen by RC/RE, and in Differential stage and VAS stage, Motorola put combinations of RC/RC that makes the openloop gain is not so big.
The most obvious one is in the VAS, where usually the collector of the VAS transistor dont have RC, but to lower gain, they put RC to ground, so the gain of this VAS is lower than if it hasn't RC.
This low openloop gain theory is also strengthen by Mr.Pass design on Aleph amp. Although it don't have such as RC/RE that can be calculated, but he stated that the mosfet itself has low gain even on open loop.
I have a few question about this theory |
OK..let's go!!!:)
| quote: | | 1. If I use the output stage of CFP instead of EF, wouldn't that CFP has already gain in it? So it is impossible to have low open loop gain with CFP output? |
The CFP has current gain in it, as a EF output stage but no voltage gain.
Many people here confuse current gain with voltage gain.
The CFP has the some voltage gain (actualy a litle more) than a EF, but both are less than unity...
The straight answer..yes!! ...it's possible to have a low open loop gain with a CFP output!| quote: | | 2. What happens if we have a openloop gain lower than the closed loop gain? (like we design the open loop of 50x, but put the feedback arrangement of 100x), what happens? Will this amp can control the DC offset, inspite of the openloop VS closed loop conditions like above? What will be the gain of the closed loop |
The amp will hve in this case a litlle less than his 50 X gain...negatif feedback can not
increse voltage gain...only steal some ;)
In a trade of for lower distortion ,noise,output impedance and DC off set control.
And in the case of offset control the answer is yes... if you use a capacitor in the lower arm of the feedback voltage divider..in that case the gain become unity at DC and...Bingo.. your DC offset is controled!
| quote: | | 3. What makes the sound tends to be better in low openloop audio power amp? (At least that is what some people said). Is making an openloop as big as possible then limit them by feedback resistors in closed loop is a bad thing? Some designs do this, what is the merits and drawbacks VS the low openloop design? What are the audible difference? Will it influence damping factor or slewrate? |
This is the eternal question...feedback is a good or a bad thing??
Feedback make the output be similar to the input...(see the null test)...some people don't like acuracy and prefer euphonic colorations... feedback is a powerfull tool in electronic design.
Yes the feedback will influence the damping factor...slew rate is more dependent of the design of the amp... |
|
|
| lumanauw |
Thanks, for the explenation. But I'm not clear about CFP output. Maybe it is a different meaning or different schematic for me. For me CFP is built by 2 transistor, one is smaller transistor, which its emitor is to output, and its collector (Usually about 100 ohm) feds the big transistor. The emitor of this big transistor goes to VCC rail, and the collector goes to output. Since the output of this big transistor is from its collector, so it must have voltage gain, more than unity.
Is it CFP don't have voltage gain? In many of QSC audio power amp, the rail is more than +/-100V, but the differential is using opamp (only+/-15V supply) combined with CFP output. Both voltage and current gain are in this CFP output, while the opamp is only to fix the gain factor. |
|
|
| Christer |
| quote: | Originally posted by lumanauw
Thanks, for the explenation. But I'm not clear about CFP output. Maybe it is a different meaning or different schematic for me. For me CFP is built by 2 transistor, one is smaller transistor, which its emitor is to output, and its collector (Usually about 100 ohm) feds the big transistor. The emitor of this big transistor goes to VCC rail, and the collector goes to output. Since the output of this big transistor is from its collector, so it must have voltage gain, more than unity.
Is it CFP don't have voltage gain? In many of QSC audio power amp, the rail is more than +/-100V, but the differential is using opamp (only+/-15V supply) combined with CFP output. Both voltage and current gain are in this CFP output, while the opamp is only to fix the gain factor. |
The "trick" is that the CFP pair has a built-in local feedback that
takes away all the voltage gain. you still have the current gain.
Since the collector of the "output device" is tied to the emitter
of the input "device", any "tendecy to voltage amplification" in
the output device will give a "tendency to reduce" the Vbe of
the input device, and the net result is that you get no voltage
gain, but current gain and improved linearity etc.
I think a reasonable way to think of the CFP is as an emitter
follower with a current booster.
Note that an ordinary emitter follower also has local feedback,
but the CFP usually has a much higher loop gain.
For your seccond question, it is hard to say anything definite
without a schematic, but as you describe it there must be some
voltage amplification stage between the op amp and the output
CFPs. |
|
|
| Workhorse |
| quote: | Originally posted by lumanauw
Thanks, for the explenation. But I'm not clear about CFP output. Maybe it is a different meaning or different schematic for me. For me CFP is built by 2 transistor, one is smaller transistor, which its emitor is to output, and its collector (Usually about 100 ohm) feds the big transistor. The emitor of this big transistor goes to VCC rail, and the collector goes to output. Since the output of this big transistor is from its collector, so it must have voltage gain, more than unity.
Is it CFP don't have voltage gain? In many of QSC audio power amp, the rail is more than +/-100V, but the differential is using opamp (only+/-15V supply) combined with CFP output. Both voltage and current gain are in this CFP output, while the opamp is only to fix the gain factor. |
Hey Lumanauw
Ampman has illustrated ur question on the diagram.
Check it out. |
|
|
| Workhorse |
| Check this out! |
|
|
| janneman |
One of the questions is: does lower open loop gain sound better?
Lets look at high open loop gain. Generally, that is good, because it allows strong feedback which means low THD, low offset, very high damping factor, immunity from drifting and component aging. I think most people will agree that in itself these are desirable properties.
The problem is that high loop gain comes from almost always with largish phaseshifts, so when you DO apply the strong feedback, the large phase shift at higher frequencies causes instability. To tame them, you have to roll of the freq response open loop to make sure it is less than 1 at the freq where the phase shift reaches 180 degrees. (There are opamps that start open loop roll off at 10Hz). This then again means that the strong feedback is no longer effective at higher frequencies. Also, the means to roll of the gain in itself can have bad effects like limiting slewrate.
So how to avoid this? Use low open loop gain, meaning less phaseshift (generally) and accept less effective feedback, that is however constant across most of the freq range. The trick here is to find a fine compromise to avoid the pitfalls of high feedback and still get the benefits from feedback itself.
Whether the low open loop gain amp sounds better is in the end depending on how clever the designer can balance these conflicting requirements. Important is the ability to construct a low open loop gain amp that is linear in itself, before the application of feedback.
BTW, a CFP pair is a high open loop gain stage. It is the (100%) feedback that reduces the closed loop gain to unity.
Jan Didden |
|
|
| pjacobi |
| quote: | Originally posted by janneman
[...](There are opamps that start open loop roll off at 10Hz). |
Bug or feature?
I'm all fine with this behaviour, if it comes with high enough open loop gain at 20kHz. At least the pase of this constant over the audio range.
Instead of gasping of low ol gains, just state that it is an integrator. Sounds much better.
Regards,
Peter Jacobi |
|
|
| Tube_Dude |
| quote: | Originally posted by lumanauw
Thanks, for the explenation. But I'm not clear about CFP output. Maybe it is a different meaning or different schematic for me. For me CFP is built by 2 transistor, one is smaller transistor, which its emitor is to output, and its collector (Usually about 100 ohm) feds the big transistor. The emitor of this big transistor goes to VCC rail, and the collector goes to output. Since the output of this big transistor is from its collector, so it must have voltage gain more than unity |
Yes !! Your descrition of two transistors ,one smaller feeding one big output is a CFP...and yes it don't have voltage gain!
| quote: | | Is it CFP don't have voltage gain? In many of QSC audio power amp, the rail is more than +/-100V, but the differential is using opamp (only+/-15V supply) combined with CFP output. Both voltage and current gain are in this CFP output, while the opamp is only to fix the gain factor |
Thats is another case...if that case the emiters of the "smaller" resistor don't connect to the output (and colector of the big transistor). but are connected to a voltage divider from output and ground....
Usualy in this last case the gain of the output stage is 2 or 3 and the op amp can swing the full 100 volts at the output..
The voltage divider in that case do all the diference!!;)
Here you will find a exemple!
http://www.audiofanatic.it/Schemi/T...IRF9540_540.jpg |
|
|
| janneman |
| quote: | Originally posted by pjacobi
Bug or feature?
I'm all fine with this behaviour, if it comes with high enough open loop gain at 20kHz. At least the pase of this constant over the audio range.
Instead of gasping of low ol gains, just state that it is an integrator. Sounds much better.
Regards,
Peter Jacobi |
Feature, of course!;)
And yes, it's an integrator. In fact, most audio amps basically are integrators.
Jan Didden |
|
|
| janneman |
| quote: | Originally posted by Tube_Dude
Yes !! Your descrition of two transistors ,one smaller feeding one big output is a CFP...and yes it don't have voltage gain!
Thats is another case...if that case the emiters of the "smaller" resistor don't connect to the output (and colector of the big transistor). but are connected to a voltage divider from output and ground....
Usualy in this last case the gain of the output stage is 2 or 3 and the op amp can swing the full 100 volts at the output..
The voltage divider in that case do all the diference!!;)[snip] |
Basically, a CFP is a two stage amp with open loop voltage gain. The closed loop gain is set with the two resistors as you say. The special case where one of the resistors is zero gives 100% neg feedback and reduces the voltage gain to close to one. But the basic circuit is always the same.
Jan Didden |
|
|
| djk |
"Is it CFP don't have voltage gain? In many of QSC audio power amp, the rail is more than +/-100V, but the differential is using opamp (only+/-15V supply) combined with CFP output. "
http://www.diyvideo.com/forums/atta...p?postid=182299
The schematic shows the Hafler Transnova, the QSC uses the same grounded output stage.
Another way to build it would be to swap ground and the center tap on the transformer (so it looks like a normal amplifier) and run the ground reference for the opamp ±15V on the amplifier positive output (this is called suspended supply operation). |
|
|
| Tube_Dude |
| quote: | Originally posted by djk
The schematic shows the Hafler Transnova, the QSC uses the same grounded output stage. |
I think there are an ATC amp that use the same technology.
If memory serves...;) |
|
|
| lumanauw |
With the method of QSC amp works (output from the CT of the transformer) it is possible for us to make all the stages before the transformer (from the differential, VAS, output stage) not to have ANY voltage gain at all. Both the current and voltage gain are in the transformer itself.
Could this be considered a low openloop gain, since all the electronics dont have any gain is possible? Or this is exactly not low openloop gain, since the gain in the transformer system is indeed very big?
(we only have to drive + and -0.6VDC on final transistor to get full max swing of +-VCC. If the VCC is 1000V, we can obtain this voltage only by driving the final transistor 0.6V, because it is inverted by the CT transformer itself.
Is there any audible excellence in this kind of configuration, or is it tends to be worse than ordinary amp? Or is it producing the same output quality?
One more question about this configuration. In ordinary amp, we use bank of caps (10.000uF per rail). How important this cap-bank in this kind of power amp, since we modulate the transformer? Will the cap only 100uF sufficient in this configuration, or is it still need the same 10.000uf? |
|
|
| janneman |
If you connect the load between the point E and the sources of the FETS, you have a traditional amp except that the ground connection is moved to the other side of the load.
If you have the feedback polarities correct, I don't see any reason why the quality of the sound should be different, in principle.
Jan Didden |
|
|
| fscarpa58 |
I agree Jan
Apart from the different bias scheme, the following
are perfectly equivalent.
Federico |
|
|
| pjacobi |
| Isn't the idea of the less usual lower circuit, to mount the power BJTs unisolated and have the heatsink at GND? |
|
|
| janneman |
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.
There actually are several patents to this and similar topologies, all claiming great benefits. Transnova is one IIRC. It is a way to discern yourself from all the other "standard" amps on the market and as such may be a valuable marketing tool. Most of the times these schemes are picked apart in the trade press.
Jan Didden |
|
|
| Nelson Pass |
| quote: | Originally posted by amp_man_1
Ampman has illustrated ur question on the diagram.
Check it out. |
Actually Ampman, the circuit on the left is a follower,
as it has unity gain via local feedback, so it does not have
voltage gain as seen by the outside world. |
|
|
| PMA |
| quote: | Originally posted by lumanauw
In many writings that can be found on the net, many stated the advantages of having a low value of openloop of a power amp. This when combined with sufficient feedback will result in good reproduction power amp.
|
IMHO this is not an issue. Try to obtain the best possible linearity of every stage (before the NFB is applied). Then after applying negative feedback (global) you will get perfect sonic result. One of the biggest mistakes is to try to compensate non-linearities (cross-over, turn-on-offs etc.) only by global NFB. |
|
|
| Workhorse |
| quote: | Originally posted by Nelson Pass
Actually Ampman, the circuit on the left is a follower,
as it has unity gain via local feedback, so it does not have
voltage gain as seen by the outside world. |
Thanks to Sir Nelson Pass for correcting me.
Regards
AmPmaN |
|
|
| lumanauw |
| quote: | | IMHO this is not an issue. Try to obtain the best possible linearity of every stage (before the NFB is applied). Then after applying negative feedback (global) you will get perfect sonic result. One of the biggest mistakes is to try to compensate non-linearities (cross-over, turn-on-offs etc.) only by global NFB. |
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? |
|
|
| tubetvr |
| quote: | | 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. |
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 |
|
|
| fscarpa58 |
Hi,Lumanauw, All
| quote: | | What is the guidance for designing power amp, to have "the best possible linearity of every stage "? |
There are stage topologies which are better then other (e.g. CFP Vs. Darlington). Usually this is due to the presence of more local feedback.
So,
1) use better devices
2) use more local feedback and less global.
Yes, it is a bit too simplistic but it helps.
Federico |
|
|
| Workhorse |
| quote: | Originally posted by fscarpa58
Hi,Lumanauw, All
There are stage topologies which are better then other (e.g. CFP Vs. Darlington). Usually this is due to the presence of more local feedback.
So,
1) use better devices
2) use more local feedback and less global.
Yes, it is a bit too simplistic but it helps.
Federico |
Yeah I completely agree with Federico that one must use more local feedback and less global feedback.
Regards
AmpmAN |
|
|
| Christer |
| quote: | Originally posted by lumanauw
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?
|
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 |
| quote: | | A BJT has en exponential IC vs. Vbe |
A BJT behaves most linearily when it is used as what it actually represents: a current-controlled current-source.
That means that distortion can be reduced by carefully chosing driving impedances !
Regards
Charles |
|
|
| Christer |
| quote: | Originally posted by phase_accurate
A BJT behaves most linearily when it is used as what it actually represents: a current-controlled current-source.
That means that distortion can be reduced by carefully chosing driving impedances !
Regards
Charles |
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. |
|
|
| djk |
"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. |
|
|
| janneman |
| quote: | Originally posted by djk
[B[snip]Crown went to the grounded ouput stage about 30 years ago.
Their stated reason was for ease in implementing their new ODEP protection circuit.
[snip] [/B] |
Yes. In some applications, adequate protection made be important enough to sacrifice the ultimate in sound quality. It's all a matter of priorities.
Jan Didden |
|
|
| lumanauw |
I can understand Local feedback in final stage as CFP output stage. It is really this CFP is better than triple darlington?
But I do not understand how to make local feedback in Differential and VAS. Could anyone give schematic on how to do this? |
|
|
| lumanauw |
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)? |
|
|
| Christer |
| quote: | Originally posted by lumanauw
But I do not understand how to make local feedback in Differential and VAS. Could anyone give schematic on how to do this? |
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.
| quote: | Originally posted by lumanauw
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?
|
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. |
|
|
| lumanauw |
Is it using RC and RE automaticly gives feedback to the transistor?
What about having local feedback in VAS, is it if I put RE it is automaticly local feedback? Most power amp design have this, so local feedback is already there in all schematic? I just dont realize it. |
|
|
|
