First of all, a duly obliged thanks to all who took some of their time to reply.
My foremost concern is to keep the existing physical circuitry as intact as possible. Therefore, I am aiming at the minimum number of changes which correct the present problems of crackling and crossover ringing. The latter, is caused by parasitics from the output stage having a toll on global negative feedback. Consequently, to correct the serious issue of crossover ringing, which was quite appreciable under LTSpice, I initially excluded the output stage from global negative feedback. The result, was as expected, the removal of the crossover ringing and an extremely clean crossover. The distortion figure increased, as the output stage, which is an emitter follower, because the drivers are connected as a totem pole. Since, negative feedback is inherently good, I tried to partially include the output stage in the loop weakly enough for the crossover ringing distortion not to happen, and this succeeded.
The negative feedback is now taken from the output point through a 100k resistor and from the middle point of the VAS. This point was achieved with two parallel RC networks connected in series with the point joining them forming the feedback point. This network was connected in parallel with the VBE multiplier. The RC value is 1k and 33uF. The equivalent insertion of the resultant DC resistance of 2k in parallel with the VBE multiplier, resulted in an easier ability to find a bias for the right quiescent current.
Finally, I will try further to correct crackling by replacing the large, 2200uF, capacitor at the base of the negative feedback path. Large electrolytic capacitors are primarily intended for power supply smoothing where minute leakages of a few microamps is immaterial, but in a feedback loop that is problematica as it is a source of unwanted signals. In this case, these signals are a characteristic of the thermodynamical properties of the electrolyte molecules and solvent. These days there are 'dry' electrolytics which I think should be immune from such thermodynamical noise.
I am attaching a schematic which, as I wrote, keeps the original circuitry as intact as possible. This amplifier already gave me two years of service with beautiful music.
My foremost concern is to keep the existing physical circuitry as intact as possible. Therefore, I am aiming at the minimum number of changes which correct the present problems of crackling and crossover ringing. The latter, is caused by parasitics from the output stage having a toll on global negative feedback. Consequently, to correct the serious issue of crossover ringing, which was quite appreciable under LTSpice, I initially excluded the output stage from global negative feedback. The result, was as expected, the removal of the crossover ringing and an extremely clean crossover. The distortion figure increased, as the output stage, which is an emitter follower, because the drivers are connected as a totem pole. Since, negative feedback is inherently good, I tried to partially include the output stage in the loop weakly enough for the crossover ringing distortion not to happen, and this succeeded.
The negative feedback is now taken from the output point through a 100k resistor and from the middle point of the VAS. This point was achieved with two parallel RC networks connected in series with the point joining them forming the feedback point. This network was connected in parallel with the VBE multiplier. The RC value is 1k and 33uF. The equivalent insertion of the resultant DC resistance of 2k in parallel with the VBE multiplier, resulted in an easier ability to find a bias for the right quiescent current.
Finally, I will try further to correct crackling by replacing the large, 2200uF, capacitor at the base of the negative feedback path. Large electrolytic capacitors are primarily intended for power supply smoothing where minute leakages of a few microamps is immaterial, but in a feedback loop that is problematica as it is a source of unwanted signals. In this case, these signals are a characteristic of the thermodynamical properties of the electrolyte molecules and solvent. These days there are 'dry' electrolytics which I think should be immune from such thermodynamical noise.
I am attaching a schematic which, as I wrote, keeps the original circuitry as intact as possible. This amplifier already gave me two years of service with beautiful music.
Last edited:
The NFB resistor network senses the output from two distinct points, namely, the output itself and the midpoint of the VAS. Since, these voltages follow each other very closely, the two resistors which sense these voltages behave as if they are in parallel. In fact, the close loop gain is that obtained with a 10k resistor at the output.
I tried various combinations for these two resistors and I think, I will stop with an 18.3k resistor connected to sense the midpoint of the VAS and a 22k resistor to sense the output. The equivalent parallel resistance is 10k. With this setup and these resistance, I am getting very good distortion figures without crossover ringing and an extremely clean crossover. Before making the changes to the physical circuit, I will do more exploration of the output to confirm my observation further.
Thanks to this forum, this amplifier will become a real jewel for me. This forum taught me what expensive tuition could do for a lot of money.
Thanks to ALL.
I tried various combinations for these two resistors and I think, I will stop with an 18.3k resistor connected to sense the midpoint of the VAS and a 22k resistor to sense the output. The equivalent parallel resistance is 10k. With this setup and these resistance, I am getting very good distortion figures without crossover ringing and an extremely clean crossover. Before making the changes to the physical circuit, I will do more exploration of the output to confirm my observation further.
Thanks to this forum, this amplifier will become a real jewel for me. This forum taught me what expensive tuition could do for a lot of money.
Thanks to ALL.
The network connected to the VAS look like some form of bootstrap on the collector. Bootstraps are positive feedback circuits. If you hover over the SPICE toolbar moving left to right you will find one "Pick Visible Traces" select the resistors you mention so these are both in the same take and see how close the wave forms match.
Definitely, I would not use positive feedback on a transistorised amplifier circuit. I only want negative feedback, and since the PD midpoint of the VBE multiplier very closely follows the output voltage, I connected a resistor to use that point for negative feedback. I have built self contained amplifiers with only an input stage and a VAS, and the output was sensed from the VAS, and none oscillated, or better they stayed stable with the right Miller Capacitor. I do not want a bootstrap, especially one sporting positive feedback in a transistorised circuit.
Please, note as a consequence of the two pronged negative feedback network the crossover ringing was cured and the bias was easier to set than before. The reason for the latter is the added equivalent 2k resistance across the VBE multiplier.
Please, note as a consequence of the two pronged negative feedback network the crossover ringing was cured and the bias was easier to set than before. The reason for the latter is the added equivalent 2k resistance across the VBE multiplier.
The current phase matches perfectly for both resistor R28 and R50. The waveforms are exactly the same but with slightly different currents. This corroborates further my hypothesis that these resistors are effectively in parallel and their parallel resistance is what determines the closed loop gain. As written earlier, I will study the diagrams more and then I will do the physical changes.
The fact the achieved crossover is very linear and clean is what is motivating me to do the changes. This amplifier will be a jewel, but I will test it under a protective circuit until I am more than certain it is well behaved: amplifiers can be treacherous!
The fact the achieved crossover is very linear and clean is what is motivating me to do the changes. This amplifier will be a jewel, but I will test it under a protective circuit until I am more than certain it is well behaved: amplifiers can be treacherous!
If you think about your output stage with 4 transistors in each output half the changing currents will generate a magnetic field as these turn on and off at the crossover point. One should ensure the collector and emitter leads to each of these device are in parallel to minimise the area within such fields. Also using large value capacitors in the nfb coupling to earth arm poses a greater risk of changing magnetic field inducing switching artifacts into the inverting input terminal as they make better aerials - especially if the closed loop gain is as high as you have made this. If you reduce the closed loop gain as I did in my simulation you minimise the risk since you can get away with 100uF instead of 2200uF. You have diodes in parallel to limit voltages across whatever capacitor you put there and lower voltage ones are more compact.
Simulations are strongly suggesting the issue of crossover ringing and other ringing is resolved completely. Nevertheless, I will do more study of the graphs to be more than sure this is what is actually happening. Regarding changing the positioning of the output pairs, I kept the original circuitry from the factory (Wharfedale model MS1500), the VAS and input stages were sawed off as these were surface mount circuitry and I had full acess to the power supply, driver and output stage.
Testing for Self Oscillation:
In my search for more direct methods to test an amplifier for stability, I thought of the idea of investigating the base currents of the differential pair. The phases of these currents are what causes stability and instability, and knowing the phases should help. It will also help is the Miller Capacitance is removed for the first test which renders the amplifier unstable and compare the results with when the Miller Capacitance is reinstroduced. I found there are very appreciable differences in the graphs and that instability should be first reached at as low as about 300kHz.
The test that I used is an AC Test from 10Hz to 100MHz.
Testing for Self Oscillation:
In my search for more direct methods to test an amplifier for stability, I thought of the idea of investigating the base currents of the differential pair. The phases of these currents are what causes stability and instability, and knowing the phases should help. It will also help is the Miller Capacitance is removed for the first test which renders the amplifier unstable and compare the results with when the Miller Capacitance is reinstroduced. I found there are very appreciable differences in the graphs and that instability should be first reached at as low as about 300kHz.
The test that I used is an AC Test from 10Hz to 100MHz.
I take it you are referring to C1 =100pF. The actual level of capacitance the non inverting half of the LTP will be loaded by will be far greater since it will be amplified by the gain factor of the VAS.
You need enough current in the LTP to be able to do that. What you have may be enough after all your hardware has been working you could simulate with R8 reduced from 270R to 220R.
In domestic use I doubt you will be driving extreme loads like 2R//1uF as per your simulation.
You need enough current in the LTP to be able to do that. What you have may be enough after all your hardware has been working you could simulate with R8 reduced from 270R to 220R.
In domestic use I doubt you will be driving extreme loads like 2R//1uF as per your simulation.
There is a thread you might like to explore which might help sort out your LTP mainly resistor values - this following on from a paper by Samuel Groner on Self's Audio Design Handbook. There is a pdf of the paper and more to consider from comments in the thread. See https://www.diyaudio.com/community/...ifier-design-handbook-by-douglas-self.183146/