Very interesting! I wish I had seen this earlier. UA-cam is too busy trying to get me to watch other videos which are of zero interest. A few observations from someone who does this kind of thing all the time with in-water piezoceramic transducers where we use a BVD (Butterworth Van Dyke) model which uses a parallel capacitor and a set of RLC series strings which define the resonance points. One difference in the way that this is handled is that the priorities are different. With in-air acoustics, I guess people are interested in either fidelity or some pleasing type of sound, while in-water acoustics are driven by maximizing radiation efficiency in some defined band. This means the resistances in the model need to be divided into a portion that is heating up the transducer as opposed to a portion that is moving water. As an added complication, the portion that is moving water can't be linear. I think this is similar in-air, so I guess you are actually generating a small signal model around some operating point. 1) Analog devices use to have a low-cost development card for doing impedance measurements over a band. It's not like having one of the professional units, because limited dynamic range means you need to futz with a few components to keep things in range. 2) If you have a measured Bode plot and a model, you can use a genetic algorithm to mindlessly converge on a good approximation for the elements that minimize the error as defined by your scoring function. I used to do this for beamforming phased array element localization. You need to be able to randomly adjust component values and then run your scoring function. If it's an improvement, you use the perturbed values. Otherwise you use the pre-perturbed values and try again. At the beginning, it's easy to find improvements. At some point, you stop finding improvements and you're done. If you plot the new Bode curve each time you find an improvement, it looks like magic as it converges. 3) The simplest implementation of a generalized BVD model uses many parallel RLC strings to model the multiple resonances found in any in-water transducer along with the bulk capacitance. It seems that you could have a similar model with two resonances, one in band and the other above the band of interest, along with a 'bulk resistance' to get the right impedance at low frequencies. I had some interest in air acoustics several decades ago, but was discouraged by too many discussions with people who were primarily interested in acoustic characterizations that only they could hear, or thought that things like speaker cables were a major part of the solution space. In sonar systems, fidelity is never even mentioned, and the focus is on either putting as much power as possible in the water, or receiving signals over the largest possible dynamic range with the highest sensitivity. There are still lots of arguments about the best way to achieve these goals, but at least I don't have to argue with morons who are talking about how much air there is in their copper cables!
Thanks for the input. I haven't done any work with underwater acoustics, certainly not high power sonar apps. It is interesting how the design targets are so different. For audio, fidelity is everything (at least for music applications, public speaking is a little different). And I totally get your final commentary about "things only they can hear". The tweak hi-fi world is little more than well-heeled and gullible individuals being fleeced by grifters. Seriously, I've heard people extolling the virtues of special replacement AC power cables costing hundreds of dollars for their audio power amps. Apparently, it never occurs to them that behind that AC face plate is maybe 40 feet of Romex leading to their breaker box.
Please note that you are using a current generator. It is ok to find parameters but in real word you are driving the loudspeaker with a transistor amplifier which has a very low output impedance or a tube amp which has an impedance equal to your loudspeaker. Therefore because the low output impedance is in parallel of your LC bump the low frequency bump will be attenuated.
Sure, but the point here is to use the current generator as an easy means to derive the model. Once you have the model, you can now use it to explore with greater accuracy the capabilities of your amplifier. As an example, I just posted a video on the effect of using a more realistic reactive load for a power amplifier versus the typical resistive load used typically. That video shows what happens to the load line of a class A amplifier. I will be posting another video in the next couple of weeks showing the effects on a class B amplifier. Here's a link to the class A load line video: ua-cam.com/video/leRAqhsMoTI/v-deo.html
Why Z parameter is difficult to measure for a transistor? In Engineering circuit analysis by Hayth, I read that it is difficult to measure Z _21 parameter by open circuiting output terminal. It says it affects the bias. But they are saying short circuiting can be done. It's in the first paragraph under the heading 'Hybrid parameters '
Very interesting! I wish I had seen this earlier. UA-cam is too busy trying to get me to watch other videos which are of zero interest.
A few observations from someone who does this kind of thing all the time with in-water piezoceramic transducers where we use a BVD (Butterworth Van Dyke) model which uses a parallel capacitor and a set of RLC series strings which define the resonance points. One difference in the way that this is handled is that the priorities are different. With in-air acoustics, I guess people are interested in either fidelity or some pleasing type of sound, while in-water acoustics are driven by maximizing radiation efficiency in some defined band. This means the resistances in the model need to be divided into a portion that is heating up the transducer as opposed to a portion that is moving water. As an added complication, the portion that is moving water can't be linear. I think this is similar in-air, so I guess you are actually generating a small signal model around some operating point.
1) Analog devices use to have a low-cost development card for doing impedance measurements over a band. It's not like having one of the professional units, because limited dynamic range means you need to futz with a few components to keep things in range.
2) If you have a measured Bode plot and a model, you can use a genetic algorithm to mindlessly converge on a good approximation for the elements that minimize the error as defined by your scoring function. I used to do this for beamforming phased array element localization. You need to be able to randomly adjust component values and then run your scoring function. If it's an improvement, you use the perturbed values. Otherwise you use the pre-perturbed values and try again. At the beginning, it's easy to find improvements. At some point, you stop finding improvements and you're done. If you plot the new Bode curve each time you find an improvement, it looks like magic as it converges.
3) The simplest implementation of a generalized BVD model uses many parallel RLC strings to model the multiple resonances found in any in-water transducer along with the bulk capacitance. It seems that you could have a similar model with two resonances, one in band and the other above the band of interest, along with a 'bulk resistance' to get the right impedance at low frequencies.
I had some interest in air acoustics several decades ago, but was discouraged by too many discussions with people who were primarily interested in acoustic characterizations that only they could hear, or thought that things like speaker cables were a major part of the solution space. In sonar systems, fidelity is never even mentioned, and the focus is on either putting as much power as possible in the water, or receiving signals over the largest possible dynamic range with the highest sensitivity. There are still lots of arguments about the best way to achieve these goals, but at least I don't have to argue with morons who are talking about how much air there is in their copper cables!
Thanks for the input. I haven't done any work with underwater acoustics, certainly not high power sonar apps. It is interesting how the design targets are so different. For audio, fidelity is everything (at least for music applications, public speaking is a little different). And I totally get your final commentary about "things only they can hear". The tweak hi-fi world is little more than well-heeled and gullible individuals being fleeced by grifters. Seriously, I've heard people extolling the virtues of special replacement AC power cables costing hundreds of dollars for their audio power amps. Apparently, it never occurs to them that behind that AC face plate is maybe 40 feet of Romex leading to their breaker box.
Thank you!
I was today years old when I learned 'Small' is a name and not an adjective relating to Thiele parameters
Please note that you are using a current generator. It is ok to find parameters but in real word you are driving the loudspeaker with a transistor amplifier which has a very low output impedance or a tube amp which has an impedance equal to your loudspeaker. Therefore because the low output impedance is in parallel of your LC bump the low frequency bump will be attenuated.
Sure, but the point here is to use the current generator as an easy means to derive the model. Once you have the model, you can now use it to explore with greater accuracy the capabilities of your amplifier. As an example, I just posted a video on the effect of using a more realistic reactive load for a power amplifier versus the typical resistive load used typically. That video shows what happens to the load line of a class A amplifier. I will be posting another video in the next couple of weeks showing the effects on a class B amplifier.
Here's a link to the class A load line video: ua-cam.com/video/leRAqhsMoTI/v-deo.html
Why Z parameter is difficult to measure for a transistor?
In Engineering circuit analysis by Hayth, I read that it is difficult to measure Z _21 parameter by open circuiting output terminal.
It says it affects the bias. But they are saying short circuiting can be done.
It's in the first paragraph under the heading 'Hybrid parameters '