We are approaching systems theory - finally growing up and behaving in a civilized manner
Dear adversary,
I've just had a conversation with my old friend Haudegen and this old warrior gave me the following conciliatory advice:
"Take the original L.Stellema circuit, separate the preamp from the power amp and consider both separately. Use PSpice in its currently popular form of the MicroCap12 program package. Proceed slowly and in small, incremental steps. In the first instance, focus all your attention on building a tutorial so that the youngsters can benefit from your work and time, namely learning. Along the way, you'll add specialist knowledge - and by the end, you'll have introduced your grateful students to systems theory." [Haudegen]
@Bernhard,
I'll meet you in our classroom. Please give the first block seminar.
Thank you very much, I'll join you later (I'm not a pensioner yet and still have to work for a living, so I only have limited free capacity).
Yours sincerely,
HBt.
Meetingplace:
classroom
Lecture (today and in the near future) Mr. Bernhard
Dear adversary,
I've just had a conversation with my old friend Haudegen and this old warrior gave me the following conciliatory advice:
"Take the original L.Stellema circuit, separate the preamp from the power amp and consider both separately. Use PSpice in its currently popular form of the MicroCap12 program package. Proceed slowly and in small, incremental steps. In the first instance, focus all your attention on building a tutorial so that the youngsters can benefit from your work and time, namely learning. Along the way, you'll add specialist knowledge - and by the end, you'll have introduced your grateful students to systems theory." [Haudegen]
@Bernhard,
I'll meet you in our classroom. Please give the first block seminar.
Thank you very much, I'll join you later (I'm not a pensioner yet and still have to work for a living, so I only have limited free capacity).
Yours sincerely,
HBt.
Meetingplace:
classroom
Lecture (today and in the near future) Mr. Bernhard
Last edited:
The reference, the TIM principle
Since we should all be a little too distracted by the pseudo-sledgehammer "THD(+N)", I would like to add the forgotten OLG. And as a reminder: the power dissipation of the "Diamond" power amplifier is more than 4*2.5W = 10W in idle mode.
Any Questions?
Since we should all be a little too distracted by the pseudo-sledgehammer "THD(+N)", I would like to add the forgotten OLG. And as a reminder: the power dissipation of the "Diamond" power amplifier is more than 4*2.5W = 10W in idle mode.
Any Questions?
Attachments
On the way to a more than adequate SNR
The noise was quite legitimately used as an argument by the terrorists, except for the omitted fact that the rescue submersible was so far only a proposal from the planning office and was to be regarded as absolutely nothing definitive. The wiring of the inputs has not even been determined yet. The wiring shown is only representative, i.e. an example of one of the countless possibilities.
Anyone who thought that the sketch was already the final schematic to be published was mistaken.
Even today, nothing works without a real test setup!
HBt.
The noise was quite legitimately used as an argument by the terrorists, except for the omitted fact that the rescue submersible was so far only a proposal from the planning office and was to be regarded as absolutely nothing definitive. The wiring of the inputs has not even been determined yet. The wiring shown is only representative, i.e. an example of one of the countless possibilities.
Anyone who thought that the sketch was already the final schematic to be published was mistaken.
Even today, nothing works without a real test setup!
HBt.
Attachments
Ref_1
If the VAS does not die an uncontrolled thermal death, then this version of the planning office is exactly the one that the lab technicians are now allowed to use.
The
THD_1k
(under 25Wrms load) is (including the ninth harmonic) < 0.005%.
That's really not bad
.
Now it's time for the first practical test!
Bye,
HBt.
If the VAS does not die an uncontrolled thermal death, then this version of the planning office is exactly the one that the lab technicians are now allowed to use.
The
THD_1k
(under 25Wrms load) is (including the ninth harmonic) < 0.005%.
That's really not bad
.Now it's time for the first practical test!
Bye,
HBt.
Attachments
Appendix
The question of the meaning of the 5k1 should now be clear even to the most nagging questioner; this placeholder and reminder stands (among other things) for the final clarification of the question of stability. Without considering the Bode plot, one is unfortunately blind in this eye - or without the time domain.
5k1 -> transformed or morphed into [100Ohm+j(1/ Omega * 10nF)] - clear from the beginning. DC offset (in the simulator) is less than +/- 0.5 mV. To full scale is 2Vrms as Inputvoltage requiered - 25W at 8 Ohm!
#
The practical test setup follows, unfortunately my schedule is a little delayed.
The question of the meaning of the 5k1 should now be clear even to the most nagging questioner; this placeholder and reminder stands (among other things) for the final clarification of the question of stability. Without considering the Bode plot, one is unfortunately blind in this eye - or without the time domain.
5k1 -> transformed or morphed into [100Ohm+j(1/ Omega * 10nF)] - clear from the beginning. DC offset (in the simulator) is less than +/- 0.5 mV. To full scale is 2Vrms as Inputvoltage requiered - 25W at 8 Ohm!
#
The practical test setup follows, unfortunately my schedule is a little delayed.
Typo!
Correction: 1/+j(omega * 10nF) => -j/(omega * 10nF)
Of course, in our case the complex resistor [100V/A - j( 1/ omega * 10nF)] can also be connected directly to ground; this is the same in terms of ac voltage view.
Take a look at the open loop case, the Bode plot or the root locus curves - and you'll know what's going on. Does MC12 have a wizard who can do this thinking for us?
Correction: 1/+j(omega * 10nF) => -j/(omega * 10nF)
Of course, in our case the complex resistor [100V/A - j( 1/ omega * 10nF)] can also be connected directly to ground; this is the same in terms of ac voltage view.
Take a look at the open loop case, the Bode plot or the root locus curves - and you'll know what's going on. Does MC12 have a wizard who can do this thinking for us?
Attachments
As much as I propagate and praise the intuitive approach from time to time, it is nothing more than the well-known method of trial and error and packages like MC12 tempt us to work (only in) this way.
In future threads I will abandon the primitive-intuitive view in favor of the formal way and always refer to the complex variable s. etc.
#
In the appendix I have intuitively scribbled one last time. Please never again forget the case of the open loop, no matter how many sliders MC12 provides ... Finally, I will create a new and thus concluding thread, that shows everything possible about the experimental setup. And by the way, of course it is also possible for us to acquire the original IP-VA stage of the L. Stellema proposal stable /stability. Only one thing is not possible: to conjure up THDs <<< 0.01%; not with this concept.
Bye,
HBt.
here ends now the story.
In future threads I will abandon the primitive-intuitive view in favor of the formal way and always refer to the complex variable s. etc.
#
In the appendix I have intuitively scribbled one last time. Please never again forget the case of the open loop, no matter how many sliders MC12 provides ... Finally, I will create a new and thus concluding thread, that shows everything possible about the experimental setup. And by the way, of course it is also possible for us to acquire the original IP-VA stage of the L. Stellema proposal stable /stability. Only one thing is not possible: to conjure up THDs <<< 0.01%; not with this concept.
Bye,
HBt.
here ends now the story.
Attachments
The diabolical thing about infectious terrorism is ...
Why a CLG of 26dB and a bandwidth limitation of <48kHz? I would prefer a maximum of 20dB and fh = 10 * 20kHz. But this is not immediately possible due to the stability criterion of negative feedback systems (in this case a 2-pole), in this case (actually) not at all.
A glance at the Bode plot of the open loop is sufficient. But even in iterative, intuitive mode, the virtual trial & error method (simulator used as a tool) leads us to a benchmark value -> indirectly via the yellow Post-It 5k1. I would have expected the opposition to be capable of an immediate transfer -> namely the realization that a CLG of less than (105.1/5.1) 26.3dB will lead to dynamic instabilities. Even without the correct system view, the triple B gang should have recognized immediately what it was all about. It is no coincidence that this quotient corresponds to that of the original. Instead, the battlefield shifted to the issue of noise.
Which in this case is directly dependent on the magnitude of the input impedance (the sums of the noise currents in the direction of the virtual zero) when the non-inverting input is grounded.
But before we look at this, we need to know the two poles and look at the left half-plane and the plot.
Rollover f2/f1 -> approx. 11MHz/15kHz = 733.33 -> 57.31dB. Now everything is again very simple and almost intuitive, we are looking for the frequency at which the OLG has dropped from left to right to approx. 57dB. This is at >100kHz; this intersection point should later represent our -6dB point in the final negative feedback case, i.e. we already know that fh is set by us to <100kHz by means of the small ceramic capacitor in parallel with Rfeedback.
We recognize almost immediately where the limits of variability or adjustability of the circuit topology basically lie.
tbc
- the most important connections are concealed
- side scenes are set up
- the insistence on illusions or fantasies
- digressing from the topic
- pretending to be in possession of the only truth
- basically badmouthing
Why a CLG of 26dB and a bandwidth limitation of <48kHz? I would prefer a maximum of 20dB and fh = 10 * 20kHz. But this is not immediately possible due to the stability criterion of negative feedback systems (in this case a 2-pole), in this case (actually) not at all.
A glance at the Bode plot of the open loop is sufficient. But even in iterative, intuitive mode, the virtual trial & error method (simulator used as a tool) leads us to a benchmark value -> indirectly via the yellow Post-It 5k1. I would have expected the opposition to be capable of an immediate transfer -> namely the realization that a CLG of less than (105.1/5.1) 26.3dB will lead to dynamic instabilities. Even without the correct system view, the triple B gang should have recognized immediately what it was all about. It is no coincidence that this quotient corresponds to that of the original. Instead, the battlefield shifted to the issue of noise.
Which in this case is directly dependent on the magnitude of the input impedance (the sums of the noise currents in the direction of the virtual zero) when the non-inverting input is grounded.
But before we look at this, we need to know the two poles and look at the left half-plane and the plot.
Rollover f2/f1 -> approx. 11MHz/15kHz = 733.33 -> 57.31dB. Now everything is again very simple and almost intuitive, we are looking for the frequency at which the OLG has dropped from left to right to approx. 57dB. This is at >100kHz; this intersection point should later represent our -6dB point in the final negative feedback case, i.e. we already know that fh is set by us to <100kHz by means of the small ceramic capacitor in parallel with Rfeedback.
We recognize almost immediately where the limits of variability or adjustability of the circuit topology basically lie.
tbc
Alternatively, you could simply proceed from right to left and shift the 2nd pole; in the intuitive approach, we shift the curve from top to bottom until f2 cuts the 0dB mark.
But I don't want to have to explain all the processes in detail now either. Especially because the opposition refuses to provide detailed knowledge.
Nevertheless, a look at the angle, at the rotation of the phase, could prove to be extremely helpful. But let's take a different route and use a trick: (Rf+Rn_) / Rn_ = 30 --> Rn_ = Rf/29 = 68kOhm / 29 = |2k345Ohm|, we keep this value in the back of our minds as a mark.
But I don't want to have to explain all the processes in detail now either. Especially because the opposition refuses to provide detailed knowledge per se. Nevertheless, a look at the angle, at the rotation of the phase, could prove to be extremely helpful. But let's take a different route and use a trick: (Rf+Rn_) / Rn_ = 30 --> Rn_ = Rf/29 = 68kOhm / 29 = |2k345Ohm|, we keep this value in the back of our minds as a mark. Our aim is to fool the system into thinking that a 1st order low-pass filter with an OLG of perhaps 56.66dB is following. For this, Rn_ must be a complex resistance to ground, [R-jXc] ---> R is already known, namely conveniently 100Ohm (this is often or always the case!) ---> -j2k343Ohm, but at what frequency? At least two ways come to mind immediately, based on the graphical representation of the system in the OLG, we look at the phase and move from -225° to the left and immediately plumb to the X-axis (somewhere to the right of sqrt(fcorner^2/10) we find a suitable value and calculate and choose 10nF.
We are now on the home stretch and the somewhat cleverer, but still primitive and intuitive procedure has been outlined. With this trick we have bypassed the 26dB mark, which is inherent in the present concept as a system.
The yellow Post-It was the 5k1 representative that the simulator threw out for us while we were playing with the tool.
#
I hope I haven't gotten too entangled in my narrative structure, otherwise the mathematically correct spelling will have to do.
Please excuse any linguistic inconsistencies, they are due to the complexity of the matter and the translation process.
Best regards,
HBt.
Psst ...
after the static view, everything revolves around dynamic stability - and how do I achieve this? The distortion factor or the inversion and the noise were not really the topic of this thread, nor was the inversion of the audio signal, but rather the frequency response compensation and the TIM doctrine of the original. The opposition and troublemakers forced us to navigate a turbulent side channel of the Rio Grande.
But I don't want to have to explain all the processes in detail now either. Especially because the opposition refuses to provide detailed knowledge.
Nevertheless, a look at the angle, at the rotation of the phase, could prove to be extremely helpful. But let's take a different route and use a trick: (Rf+Rn_) / Rn_ = 30 --> Rn_ = Rf/29 = 68kOhm / 29 = |2k345Ohm|, we keep this value in the back of our minds as a mark.
But I don't want to have to explain all the processes in detail now either. Especially because the opposition refuses to provide detailed knowledge per se. Nevertheless, a look at the angle, at the rotation of the phase, could prove to be extremely helpful. But let's take a different route and use a trick: (Rf+Rn_) / Rn_ = 30 --> Rn_ = Rf/29 = 68kOhm / 29 = |2k345Ohm|, we keep this value in the back of our minds as a mark. Our aim is to fool the system into thinking that a 1st order low-pass filter with an OLG of perhaps 56.66dB is following. For this, Rn_ must be a complex resistance to ground, [R-jXc] ---> R is already known, namely conveniently 100Ohm (this is often or always the case!) ---> -j2k343Ohm, but at what frequency? At least two ways come to mind immediately, based on the graphical representation of the system in the OLG, we look at the phase and move from -225° to the left and immediately plumb to the X-axis (somewhere to the right of sqrt(fcorner^2/10) we find a suitable value and calculate and choose 10nF.
We are now on the home stretch and the somewhat cleverer, but still primitive and intuitive procedure has been outlined. With this trick we have bypassed the 26dB mark, which is inherent in the present concept as a system.
The yellow Post-It was the 5k1 representative that the simulator threw out for us while we were playing with the tool.
#
I hope I haven't gotten too entangled in my narrative structure, otherwise the mathematically correct spelling will have to do.
Please excuse any linguistic inconsistencies, they are due to the complexity of the matter and the translation process.
Best regards,
HBt.
Psst ...
after the static view, everything revolves around dynamic stability - and how do I achieve this? The distortion factor or the inversion and the noise were not really the topic of this thread, nor was the inversion of the audio signal, but rather the frequency response compensation and the TIM doctrine of the original. The opposition and troublemakers forced us to navigate a turbulent side channel of the Rio Grande.
Last edited:
- and the attempt at a highly simplified, linguistic representation (actually of mathematical relationships and operations, perhaps transformations?!).
Now that we know one possibility - and are sensitized to the importance of the location of the poles - we can finally try to salvage the wreck of the sunken Australian Viking ship L.Stellema.
There is still one (or two) point for me, namely the attempt to localize a flaw:
dc-gain totaly not equal to ac-gain.
Well,
since audio amplifiers are not DC amplifiers or measuring amplifiers of all kinds, but AC amplifiers, the attempt to create confusion with this pseudo-argument alone is silly or a little bit stupid. It is enough that the control loop is able to keep all operating points constant and - to regulate /set the output offset to zero.
For this situation it is absolutely not necessary that (casually put) dc-gain = ac-gain. Incidentally, in terms of control technology there is a special term for that.
On the other hand, I have never heard a zero Hertz or even a DC voltage in its lowest bass equivalent
.The other bad thing is the "use the correct bias network ..."[quotation] statement!
1. it assumes that there is an incorrect network
and
2. it obscures the basic task of a base voltage divider of a BJTs basic circuit design ...
Oh, let's cut a long story short:
I can guess what was meant and intended by all this nonsense, but disturbance variables of this kind are not exactly helpful. By and large, the opposition's entire random argumentation chain is completely wrong, they must have failed to understand something elementary.
What drives me up the wall -> is the fact that it's only about a tiny handful of resistors, transistors and very, very few capacitors.
Nothing more and nothing less.
HBt.
Psst
In the meantime, the PCB has been drilled and is waiting to be fitted with components
.
U_Noise approx <20µV (bw=20kHz)
The whole MC12 gimmick only confirms what you /we already know in advance.
On the whole, the concept cannot be thwarted, even if we were to confirm to each other once again that we (including myself) would not do it that way.
Bye,
HBt.
Psst
And what will it sound like? We can't answer this question with any toy in the world, you /we have to build one and listen to it.
The whole MC12 gimmick only confirms what you /we already know in advance.
On the whole, the concept cannot be thwarted, even if we were to confirm to each other once again that we (including myself) would not do it that way.
Bye,
HBt.
Psst
And what will it sound like? We can't answer this question with any toy in the world, you /we have to build one and listen to it.
Last edited:
forgotten
T.I.M. (DIN) is a power of ten better than the corresponding THD (MC12 can of course also determine this) - clearly, target achieved.
factor is 20 (not 30) -> 26dB!
T.I.M. (DIN) is a power of ten better than the corresponding THD (MC12 can of course also determine this) - clearly, target achieved.
typo
factor is 20 (not 30) -> 26dB!
I understand the background to your concerns wahab; I would also be more comfortable if the emitter potential of the LTP were lower, i.e. V_CE were statically slightly higher than 1.45Vdc (maybe >= 3V). Of course, this can also be achieved in different ways, but apparently everything is fully functional as it is (seems to be).
I'm not going to simulate anything now, but tinker, measure and calculate.
I'm not going to simulate anything now, but tinker, measure and calculate.
There is absolutely no reason to do so - strangely enough
.I couldn't resist the temptation after all - with MC12, these little things are no longer a problem at all.
The upper red curves show v_ce_~ and the lower curves in pink show the resulting ~current through the first working resistor 10k.
Differential voltage of the IPS in black at the top, signal source (she is also the running variable) in blue below.
Zener diodes for static shifting of the base potential of the (inner) IPS are not necessary. The real experiment will bring the truth to light, I am quite sure of that.
regards,
HBt.
Appendix B
It was proposed to use a magic 5.1k unicorn resistor to stabilize the amplifier, however that resistor has a negative effect on THD.
Here is the "infamous" zdiode schematic with mirror. The version with mirror is much harder to get stable compared to the basic resistor version.
Gain margin 15 dB
Phase margin 50 degrees
-3dB bandwith 500 kHz
3rd harmonic -138 dB at 10V out without circuit optimization.
No magic unicorn.
It was proposed to use a magic 5.1k unicorn resistor to stabilize the amplifier, however that resistor has a negative effect on THD.
Here is the "infamous" zdiode schematic with mirror. The version with mirror is much harder to get stable compared to the basic resistor version.
Gain margin 15 dB
Phase margin 50 degrees
-3dB bandwith 500 kHz
3rd harmonic -138 dB at 10V out without circuit optimization.
No magic unicorn.
Last edited: