Found the problem
This is the gain/phase of a normal chip amp
This is the gain/phase of the MyRefC
First of all, the phase lead the the MyRef at the lower frequency is reduced, which correspond with better bass definition and blend with music, and is what I hear.
However, the phase lag at the high frequencies is lagging significantly. This generally indicates that the tone balance and the image focus will suffer, which is also what I hear.
Both tests were conducted with the same driver as a load with minim impedance around 5.6Ohm.
Lag in voltage phase will lead to even more lag in the current. So there definitely is room for improvement. Mere Swapping of components is not going to solve the problem. However, since I have already ordered some, Might as well just swap them in and see what happens.
This is the gain/phase of a normal chip amp
An externally hosted image should be here but it was not working when we last tested it.
This is the gain/phase of the MyRefC
An externally hosted image should be here but it was not working when we last tested it.
First of all, the phase lead the the MyRef at the lower frequency is reduced, which correspond with better bass definition and blend with music, and is what I hear.
However, the phase lag at the high frequencies is lagging significantly. This generally indicates that the tone balance and the image focus will suffer, which is also what I hear.
Both tests were conducted with the same driver as a load with minim impedance around 5.6Ohm.
Lag in voltage phase will lead to even more lag in the current. So there definitely is room for improvement. Mere Swapping of components is not going to solve the problem. However, since I have already ordered some, Might as well just swap them in and see what happens.
Last edited:
Interesting post!
I'm curious to see if different passives are going to change something...
Has anyone performed measurements in the past? Removal of the cap might further lower low frequency phase, but it is already not the most critical issue. I am wondering whether the new resistors are going to improve the high frequency phase or not.
Simulation using SoundEasy
Well, maybe a little improvement, but not enough.
Simulation using SoundEasy
An externally hosted image should be here but it was not working when we last tested it.
Well, maybe a little improvement, but not enough.
Last edited:
If something could change it would be for the caps, they make a BIG subjective change.
Did you ordered also FKP2s?
Change the type of cap or change the value of the cap?
Which in an attempt to reduce the phase error at upper audio frequencies?
Those gain/phase plots are interesting.
Normal chip: 40Hz to >40kHz for +-10dgegree phase error.
MyRefC: 20Hz to 11kHz for +-10degrees phase error.
Simulation: 20Hz to 34kHz for +-10degrees phase error.
Which in an attempt to reduce the phase error at upper audio frequencies?
Those gain/phase plots are interesting.
Normal chip: 40Hz to >40kHz for +-10dgegree phase error.
MyRefC: 20Hz to 11kHz for +-10degrees phase error.
Simulation: 20Hz to 34kHz for +-10degrees phase error.
As per my BOM.
I don't know if they could change something in measurements but it would be nice to discover it...isnt't it?
To tell you the truth, I already have another in mind, but it's going to take time to check and see when it's available. The reason is that the factory has material from different sources for the same spec that makes difference in sound and measurement. I just ordered two Vishay something of 1uF just to get the order ubove the no shipping range.
I did a more complicated simulation that brought the results closer to the measured data. Compensation seems possible, but I am concerned what might happen above 100KHz because the simulation is pretty much limited. I might just cross my fingers and stick the compensation in.
Cap performance effects cannot be measured because it's an issue more subtle that a simulation can do.
Latest quick and dirty sim with and without compensation.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Quite all are type/brand/model changes, the unique two value changes are C7 (100nF->10nF) and C21 (100nF->22nF)
Can you elaborate?
I did a more complicated simulation that brought the results closer to the measured data. Compensation seems possible, but I am concerned what might happen above 100KHz because the simulation is pretty much limited. I might just cross my fingers and stick the compensation in.
If you're referring to what I've PMed you, it's working in my amp since 3 months without a problem...
Pity...
If you're referring to what I've PMed you, very very interesting...
C34 cannot be implemented in the simple sim model that I use. I use generic OPamp.
C10 causes stability problems in the sim for some reason, so it is not in there.
C32 makes the phase worse.
One of the things I've discovered while working on interconnects is that phase change shifts sonic perception, currently I have not figured out what the rule is, but it happens. So normally, I listen mostly for image focus improvement which almost always is accompanied by improvements in measurements.
Compensation in my sim is just adding a feed forward capacitor (22pF)from pin 2 of LM318 to pin 9 of LM3886. Since my sim is a simple model, it is a bit risky.
C10 causes stability problems in the sim for some reason, so it is not in there.
C32 makes the phase worse.
One of the things I've discovered while working on interconnects is that phase change shifts sonic perception, currently I have not figured out what the rule is, but it happens. So normally, I listen mostly for image focus improvement which almost always is accompanied by improvements in measurements.
Compensation in my sim is just adding a feed forward capacitor (22pF)from pin 2 of LM318 to pin 9 of LM3886. Since my sim is a simple model, it is a bit risky.
Generic opamp... no good....
The My_Ref can't be simulated without a complete model of the LM318 and also of the LM3886.
The compensation network (R39, R43, C34) has a crucial rule in this amp for stability, phase coeherence and dumping factor.
Without that net the sim can't say quite nothing about the amp behaviour.
There's already a FF compensation, C10.
I fear that the simulation doesn't reflect the actual performance of the My_Ref...
This is a risk I will have to take. At least all other things have not changed, and the frequency/phase response matches reasonably well. My instinct tells me that probably (R39, R43, C34) makes the op amp closer to ideal Opamp. So unless we can improve that, those are best left alone. I have no idea what C10 effects. Personally I think does not act like a feed forward. Generally, current always lags behind voltage, the current pump should be a means to allow voltage controlled current to have very close to the same phase as the voltage. So the command voltage should have a phase lead. I think once I understand how the phase is properly controlled, then I will see if damping is causing any audible problem or not. It has to be better than what I listen to now.
Penasa states that the compensation net do:
Post 3201 are real measurements. This is indicated in the graphs where there is a sharp drop and phase irregularity at the high end frequency cutoff.
I need to study the LM318 data sheets a bit, but just based on this description, we may need a better balance between open loop gain, BW, and phase compensation (if that is what R43+C34 does). If this can be done, then maybe we don't need the feed forward cap that works in my sim.
No wonder C10 does not work in my sim. Simple sims do not need this.