Thank you very much MarsBravo for that analogy, it really helps to see the processes through more common technology, very educational, thank you.
Zjjwwa, thanks for the tip, I will have a look at it, it is good to know it is basic in my case, and I also want to thank you for your previous contribution in post #46, a very good explanation and very well described.
Zjjwwa, thanks for the tip, I will have a look at it, it is good to know it is basic in my case, and I also want to thank you for your previous contribution in post #46, a very good explanation and very well described.
Glad to be of help. Maybe one additional word to that article.
It is stated that the Open Loop Gain is very high. The Closed loop gain is much much lower.
So the question arises, when we close the feedback loop, do we simply "loose" or "waste" all that extra gain?
No, we do not.
One of the effects that was described in the article is stabilizing the predictability of the closed loop gain, irrespective of fluctuations of open loop gain.
Which is good for our purposes.
But there is yet another benefit: Output impedance.
All that "trimmed" open loop gain, when closing the loop, also serves the purpose to lower the output impedance of the amplifier. So basically, a closed feedback loop system has a much smaller output impedance, because all that "extra gain" is working into effectively lowering the output impedance.
As stated in post #77:
"Closed Loop Output Impedence = Open Loop Output Impedance / Feedback Factor
In other words: The Closed Loop Output Impedance is "Feedback Factor" times lower, than the output impedance without the feedback (open loop)
Last edited:
In the above referenced paper, there is no explanation about the output impedance of an amplifier. But one can assume that whenever feedback act upon open loop gain to settle the overall or closed loop gain at a lower value for purpose of better control, this also reduces the open loop output impedence to a lower closed loop impedance by the same factor.
There are some assumptions and calculation errors in this paper also.
Nowhere is the open loop frequency response taken into account, so the given (simplified) examples are true for dc operation only. As soon as frequency, time, delay and frequency-depending internal components are involved, the picture shifts somewhat. High gain operational amplifiers exchange this (high gain) feature for low frequency roll off, with a zero in the range of 10 - 100 Hz. A second zero appears often near the open loop unity gain, say some MHz.
Hence the output impedence of an (operational) amplifier is related to internal and external components in a less optimal fashion then preferred.
A mistake of perception is made in the section:
However with the addition of negative feedback the systems gain has only fallen from 34dB to 33.5dB, a reduction of less than 1.5%, which proves that negative feedback gives added stability to a systems gain.
The relative difference (dB's) might differ 1.5% (the ratio of two ratio's), but the absolute and true value is 4.3% (the ratio of the values 49.75 and 47.6).
edit: crossing posts!
Very true. That is the next level of finesse of all of it.
But the "DC approximation" seemed to be a good, easy reading "overall starter". That was my initial intent.
Some quick questions after having looked trough linked explanation:
What is the difference between local feedback and global feedback, and which one of these are being shown in the linked explanation.
Of these two, which one is more directly dependent of the load characteristics?
I may have missed something here, if so I'm sorry, it is not your job to educate me, that is my task, and I'm grateful for the assistance.
What is the difference between local feedback and global feedback, and which one of these are being shown in the linked explanation.
Of these two, which one is more directly dependent of the load characteristics?
I may have missed something here, if so I'm sorry, it is not your job to educate me, that is my task, and I'm grateful for the assistance.
Consider a regular main amplifier with three stages: input, voltage and current stage.
Local feedback is feedback per stage, overall feedback is feedback of the total system consisting of three stages.
One might consider a miller capacitor, connected between collector (output) and base (input) of a bjt (bipolar junction transistor) as a local feedback of this basic stage, whatever is in front or connected as load (!). This miller cap however, is highly voltage-dependent (Vcb) and is capable of 'modulating' the amplification and causing unwanted distortion.
So, feedback is not always a magical trick to get rid of distortion. It can produce distortion also. This miller cap is a very nasty fellow...
And as all practical systems are not exactly operation according to theory, there are numerous other things to take into account, such as frequency response of all used components, to name one. Signals do not propagate at infinite speed through amplifiers, so overall feedback is susceptible for time-lagging. Combined with our nasty fellow and you'll understand that if one desires to build an amplifier, an oscillater can been created instead.
Local feedback is feedback per stage, overall feedback is feedback of the total system consisting of three stages.
One might consider a miller capacitor, connected between collector (output) and base (input) of a bjt (bipolar junction transistor) as a local feedback of this basic stage, whatever is in front or connected as load (!). This miller cap however, is highly voltage-dependent (Vcb) and is capable of 'modulating' the amplification and causing unwanted distortion.
So, feedback is not always a magical trick to get rid of distortion. It can produce distortion also. This miller cap is a very nasty fellow...
And as all practical systems are not exactly operation according to theory, there are numerous other things to take into account, such as frequency response of all used components, to name one. Signals do not propagate at infinite speed through amplifiers, so overall feedback is susceptible for time-lagging. Combined with our nasty fellow and you'll understand that if one desires to build an amplifier, an oscillater can been created instead.
The reason for this mismatch is the fact transistors are extremely complex to represent algebraically accurately. Therefore, to reduce algebraic complexity, transistor representation is approximated. For instance, in low frequency circuit analysis, the base-collector capacitance is usually ignored, although, it still affects circuit operation.MarsBravo said:
An amplifier will behave exactly per theory. The trouble is making a complete model for it. That includes at least a planar EM solution fior the board layout, models for the passives that are good up to at least the first resonance, and accurate models for the transistors. Models are typically worth exactly what you pay for them. Built-in spice models are rough approximations only - if actual transistor behavior deviates from the mathematical formulations by enough, no set of model parameters or coefficients will fit it. In many cases you can get “good enough”. Sometimes you can’t. A proper model for the power supply,, reasonable transistor models, and reasonable approximations for the PCB trace inductances and capacitances will model an amp over the audio band. May or may not be good enough for proper stability analysis, or to analyze a very high speed amp with bandwidth into the 100’s of kHz. Won’t model distortion down to the single digit ppm’s, but most amps aren’t that good. It takes a sophisticated model to design one that accurately. I have to laugh, roll my eyes, and shake my head every time I see another design posted that simulates .000003% distortion without doing a full board level simulation and using transistor models found for free on the internet.
the audio signal is too complex and speakers and room too problematic to generalize anything.
There are people who like classD , kevlar drivers, or poly, and there are those who like paper drivers with SET.
I think that the speaker designer goal should be taken into consideration....
A high sensitivity alnico with large 15 inches woofers alnico will benefit from a lower power softer sounding amplifier such as classA SET...
But a very modern big magnet poly cone with a very good XO will benefit from hundred of watts of available power. Especially with heavy magnets , underhung coils, and ventilation and lot of Xmass
So, you either hear the amplifier soft clipping or you hear the speakers scratching your ears. The best system should make everything come together and sound in harmony.
There are people who like classD , kevlar drivers, or poly, and there are those who like paper drivers with SET.
I think that the speaker designer goal should be taken into consideration....
A high sensitivity alnico with large 15 inches woofers alnico will benefit from a lower power softer sounding amplifier such as classA SET...
But a very modern big magnet poly cone with a very good XO will benefit from hundred of watts of available power. Especially with heavy magnets , underhung coils, and ventilation and lot of Xmass
So, you either hear the amplifier soft clipping or you hear the speakers scratching your ears. The best system should make everything come together and sound in harmony.
I have read through this whole thread and while there has been some great discussion I believe there a possibility that the root cause of the problem has been completely missed. You have not articulated in any detail what exactly you don't like about the sound of this system so I'm reading between the lines here, but I think it is entirely possible it just needs some corrective processing to account for non linearities in the design and how the box interacts with the listening environment, and perhaps the system needs time alignment if the sub output lags or sounds sloppy.
I don't see how it would be possible to achieve anything remotely close to a flat response in room even with a relatively simple reflex or sealed sub unless you just get crazy lucky, and I know this is not a standard BP design but those are regularly regarded as sounding slow, so I think it's a big mistake not to take advantage of the processing power at your disposal. You will need something other than a test mic plugged into the DRPA though, you need to measure impulse response as well as frequency response so something like REW will be needed.
As for the question of amplifier sound with regard to low frequencies I think there are differences but they generally only begin to appear at the upper limits, at the 1/10th of rated power level everything is operating in it's linear range and is well under control. As an experiment you could bridge the amp into the sub instead of just using 1 channel, I know that's not the same as bringing in a totally different amp but if that makes a positive difference then maybe you need more power than you think, if it's noticably worse then maybe there is something to your suspicions... or maybe your copy isn't as healthy as it should be because really this should not make any difference one way or the other.
You also couldn't ask for a more well behaved driver.. the klipple test results I have seen on these neo B&C's are excellent and it has one of the most powerful motors available these days, you would be hard pressed to find something better.
Yes, it is. Several members proved it that measurement result can very close to simulation result. Of course, the implementation is matter, example: good pcb layout, choosing right component type, good wiring, etc.
Only who do not understand theory, said opposite.
I realize that the linked document regarding feedback is a generic and basic run-trough of how it works, I assume it can be applied both locally and globally, or perhaps both, and even though this is interesting and helps the understanding of how it works I cannot really see the load control part of it beyond a gain stabilizing function.
However, as far as I can understand (and mind you that is not a lot at this stage, so this may be wrong) it seems that a large amount of negative feedback, globally, might act destructive if the load is highly reactive, meaning a big delta from the intended signal at the summing point (as per the schematic in the link), making the feedback circuit work a lot harder than intended or beyond the purpose for which it was designed, if designed for a nice stable dummy load (resistor) for example.
I currently think there are several different parts/processes ina an amplifier that makes up the load control, here is how far I've come, and again, I just starting out here so I may be wrong about this.
1. / To ensure accurate initial signal or transient reproduction we need to have a sufficient current delivery capacity and slew rate (V/microsecond)(although the slew rate may be well above need in most cases), the current need I have been told can be up to five times higher than the basic theory would have you believe.
2. / To ensure accurate signal or transient end or stop we have the output impedance acting as a damper on the reactive signal where a lower output impedance seems favorable. (I'm trying to get to grips with that).
The amount and type of feedback (local/global) or mix there of is the most tricky part for me to understand, since at least on the surface it does not seem wise to "bother" the amp with corrective functions even if it means an increase in efficiency (gain), it should come at a cost of load control, or signal integrity (in the case of extensive global negative feedback) if the amplifier is not designed for it (beyond a dummy load).
This is how far I have come, but I'm sure this is grossly oversimplified and even perhaps plain wrong, and I'm very sure there are a lot more things to consider.
As for the processing and speakers, I have applied sufficient processing in terms of delay and crossovers to comply with what the design requires to operate as intended, both alone and together with the top system.
I'm very pleased with the result, both the integration between the two speakers, and each of their stand alone performances in their respective ranges.
I'm not all that concerned about a dead flat response curve, and I am aware that big variations in the response may be perceived as a lack of load control, sounding heavy, slow or muddy etc. but it is a good point and good to bring it up.
I find that you may kill the joy of the system if apply EQ to aggressively on order to flatten it out, especially in a room, make sure to take care of the most evident problems and leave the rest alone, all speakers have a character, just make sure you agree with it.
I will try to bridge the IP450 and see what happens, I'm not expecting miracles to be honest, and as for the drivers I am aware of their design features, performance and T/S parameter balancing, I chose them specifically based on that.
But it could be good to know that they are a bit on the extreme side, the 200g of moving mass is somewhat compensated for by a 30 in BL (motor strength), the inductance is fairly low for a 3,4kW program capacity driver with a well ventilated 4,5" voice coil (1.8mH) indicating a high efficiency in transferring the current to a force acting on the cone, and more important perhaps wise versa.
It is housed in a highly efficient 6'th order series tuned quarter wave design, meaning that it may generate a lot of back-emf for the amp to handle since there are a lot of secondary forces acting on the efficient drivers diaphragm, most likely way beyond what you may encounter in more traditional designs such as bass reflex (Helmholtz) or a closed box, and so well hence my question regarding load control.
The Labgruppen IP450, although being a nice enough amp, is by my guess not intended to handle a load such as this, and so I found the FP6400 which most likely is designed for this, and so I wondered if there might be an improvement in load control (sound quality) if I sprung for the more powerful amp, that's how it started at least, but I'm learning a lot.
However, as far as I can understand (and mind you that is not a lot at this stage, so this may be wrong) it seems that a large amount of negative feedback, globally, might act destructive if the load is highly reactive, meaning a big delta from the intended signal at the summing point (as per the schematic in the link), making the feedback circuit work a lot harder than intended or beyond the purpose for which it was designed, if designed for a nice stable dummy load (resistor) for example.
I currently think there are several different parts/processes ina an amplifier that makes up the load control, here is how far I've come, and again, I just starting out here so I may be wrong about this.
1. / To ensure accurate initial signal or transient reproduction we need to have a sufficient current delivery capacity and slew rate (V/microsecond)(although the slew rate may be well above need in most cases), the current need I have been told can be up to five times higher than the basic theory would have you believe.
2. / To ensure accurate signal or transient end or stop we have the output impedance acting as a damper on the reactive signal where a lower output impedance seems favorable. (I'm trying to get to grips with that).
The amount and type of feedback (local/global) or mix there of is the most tricky part for me to understand, since at least on the surface it does not seem wise to "bother" the amp with corrective functions even if it means an increase in efficiency (gain), it should come at a cost of load control, or signal integrity (in the case of extensive global negative feedback) if the amplifier is not designed for it (beyond a dummy load).
This is how far I have come, but I'm sure this is grossly oversimplified and even perhaps plain wrong, and I'm very sure there are a lot more things to consider.
As for the processing and speakers, I have applied sufficient processing in terms of delay and crossovers to comply with what the design requires to operate as intended, both alone and together with the top system.
I'm very pleased with the result, both the integration between the two speakers, and each of their stand alone performances in their respective ranges.
I'm not all that concerned about a dead flat response curve, and I am aware that big variations in the response may be perceived as a lack of load control, sounding heavy, slow or muddy etc. but it is a good point and good to bring it up.
I find that you may kill the joy of the system if apply EQ to aggressively on order to flatten it out, especially in a room, make sure to take care of the most evident problems and leave the rest alone, all speakers have a character, just make sure you agree with it.
I will try to bridge the IP450 and see what happens, I'm not expecting miracles to be honest, and as for the drivers I am aware of their design features, performance and T/S parameter balancing, I chose them specifically based on that.
But it could be good to know that they are a bit on the extreme side, the 200g of moving mass is somewhat compensated for by a 30 in BL (motor strength), the inductance is fairly low for a 3,4kW program capacity driver with a well ventilated 4,5" voice coil (1.8mH) indicating a high efficiency in transferring the current to a force acting on the cone, and more important perhaps wise versa.
It is housed in a highly efficient 6'th order series tuned quarter wave design, meaning that it may generate a lot of back-emf for the amp to handle since there are a lot of secondary forces acting on the efficient drivers diaphragm, most likely way beyond what you may encounter in more traditional designs such as bass reflex (Helmholtz) or a closed box, and so well hence my question regarding load control.
The Labgruppen IP450, although being a nice enough amp, is by my guess not intended to handle a load such as this, and so I found the FP6400 which most likely is designed for this, and so I wondered if there might be an improvement in load control (sound quality) if I sprung for the more powerful amp, that's how it started at least, but I'm learning a lot.
Last edited:
Try a mental exercise with a calculator in hand.
Imagine that you re-do all the calculations of the voltages in the feedback loop.
But this time, Your load will be constituted of an 8 Ohm load AND a 1V battery in series with the load.
The battery will serve the purpose to try to "upset" or kick off-balance the output of the amplifier (or op-amp, or whatever).
Try to re-calculated what is now "happening" at each of the nodes.
You will most probably come up with interesting results, and notice that it is not at all easy to upset the output voltage of the amplifier.
So, basically, in other words: If you apply an EXTERNAL 1V battery to the output of the amplifier, in series with your 8 ohms load, but the output voltage of the amplifier remains almost unchanged, from this data, you can draw conclusions as to the output impedance of the amplifier.
No. Highly reactive load means big delta between the output signal, and the output current. The amplifier with feedback, will provide the output signal depended by the gain, which is set by the feedback. The question if it can provide the current, or the protection will kick in, or maybe the output devices are fail because of the SOA violence.
Sajti
Typical amplifiers actually decrease (not increase) the gain. But it is a bit like trading horses.
For example:
Open Loop Gain (meaning: an amplifier without global feedback, or: the wire going "backwards" as feedback is cut / discontinous) could be 60 dB.
Or in other words:
60 dB = 20 log (V_out / V_in)
3 = log(V_out / V_in)
V_out = V_in * 10^3
V_out = V_in * 1000.
An amplifier with its Open Loop Gain multiplies the input voltage signal by a factor of 1000.
Closed Loop Gain (meaning: an amplifier with global feedback, or: the wire going "backards as feedback is "re-connected" so as to activate the feedback loop), could be 40 dB.
Or in other words:
20 dB = 20 log (V_out / V_in)
1 = log(V_out / V_in)
V_out = V_in * 10^1 = V_in * 10.
An amplifier with its Closed Loop Gain multiplies the input voltage signal by a factor of 10.
Difference between Open Loop Gain and Closed Loop Gain is sometimes called "The Loop Gain":
60 dB - 20 dB = 40dB.
Which basically means: 10^2, or a factor of x100
The output impedance of an amplifier, when you close the loop, will fall by a factor of this "Loop Gain". In our example, the output impedance will become 100 times smaller, when you close the loop.
So basically, you are "trading horses". You "give up" a factor of 100 in terms of gain, but you get a 100 times smaller output impedance instead.
{ I hope I got it right here; A second pair of eyes - please verify }.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.