Again, the VOLTAGE is NOT very important. In fact, it MUST sag, otherwise NO signal could come out of the amplifier.
The output music signal is CURRENT. And the capacitor voltage must sag, or else no current would be discharged, to make the signal.
More voltage sag at the caps might make a very slight increase in non-linearity. But PSRR of amplifier usually makes it negligible. Clipping at very high output level is usually the only significant problem related to voltage sag.
Voltage sag being able to be larger (i.e. lower capacitance) MIGHT tend to make current signal from caps more agile, which might give more-accurate transient response, which might be perceived as sounding better. OR, that could be hogwash because more capacitance might more than compensate for that effect and still be better sounding. [Needs research.]
gootee I think you're doing circles.
If you want the benefit of what allows sag, use several smaller capacitors... but overall capacitance doesn't have to be particularly low.
Look at basically every amplifier ever made, it would appear they already know this stuff... Typically there are stages, large reseviors, and smaller ones near the actual amplification (transistor, switcher, whatever).
I suspect there's one more factor at play here, which I'll refrain from mentioning
If you want the benefit of what allows sag, use several smaller capacitors... but overall capacitance doesn't have to be particularly low.
Look at basically every amplifier ever made, it would appear they already know this stuff... Typically there are stages, large reseviors, and smaller ones near the actual amplification (transistor, switcher, whatever).
I suspect there's one more factor at play here, which I'll refrain from mentioning
I agree this area needs a lot more research giving some datapoints. In my experience more capacitance always sounds better provided the layout isn't comrpomised. Having more caps provides a lower impedance for PSU currents - the flip side of this is it makes the circuit even more sensitive still to slight errors in the layout.
Just last week I was playing with adding crazy amounts of capacitance to my DAC's power supply. Just tacking on a big capacitor array to an existing circuit provided improvements in some areas (dynamics) but made others worse (imaging). Re-working the grounding layout (only a change of 1 or 2 mm of soldered connection) fixed up the imaging problems without losing the dynamics.
But the level of sag depends on the size of the caps. Or use regs. So you can lower sag very much.
the regulator lowers the peaks much more than it lowers the sags.
The regulator may not lower the sags at all when the mains supply voltage falls, it will still lower the peaks.
I'm actually very surprised it is not possible to prove the effect of different components used in an amplifier.
The only thing I see is theoretical assumptions, sims and subjective listening experiences (including mine).
My profession is Injection molding process development specialized in molding of CD, DVD, CDR etc.
I have learned in the 20 years plus working in this field that its impossible to design a injection molding process only on theoretical calculations.
Well, for maybe 95% its is possible but the last 5% that is missing in the theory determines if the process will work or not.
Many times I have tried to improve the process based on theoretical ideas being sure it would work positively and had to found out it did not work at all.
One time someone that studied for 6 years on the university specialized in injection molding thought he could easily improve the process resulting in a train wreck, the machine produced completely NOT.
In my case it was luckily easy to measure the result of the experiments done (theoretical or not) because the quality of the end product was easy to measure.
In this case of experiments with different components in an audio amplifier the results stay very subjective, therefor I think its hard to prove what is best and I think the discussion will probably continue for ever
The only thing I see is theoretical assumptions, sims and subjective listening experiences (including mine).
My profession is Injection molding process development specialized in molding of CD, DVD, CDR etc.
I have learned in the 20 years plus working in this field that its impossible to design a injection molding process only on theoretical calculations.
Well, for maybe 95% its is possible but the last 5% that is missing in the theory determines if the process will work or not.
Many times I have tried to improve the process based on theoretical ideas being sure it would work positively and had to found out it did not work at all.
One time someone that studied for 6 years on the university specialized in injection molding thought he could easily improve the process resulting in a train wreck, the machine produced completely NOT.
In my case it was luckily easy to measure the result of the experiments done (theoretical or not) because the quality of the end product was easy to measure.
In this case of experiments with different components in an audio amplifier the results stay very subjective, therefor I think its hard to prove what is best and I think the discussion will probably continue for ever
Joerytech,
It's funny you say that. I was realising that an OpAmp is no more than a SERVO SYSTEM. When fine tuning a process, I never have to analyse the servo in the system, just tell it what to do and hope it doesn't reach its limits of operation.
You change a process and you get a result. It might not be what you expected and you try to explain what just happened, but proving your explanation is a lot more difficult. These amps are the same.
A simulation is far more than a theoretical calculation. Its a copy of your process that you can run offline. With a simulation your only concern is if the model is correct, in a real process it's if you can measure it correctly and don't blow it up.
Back when I was programming robots it would have been great to have been able to do simulations. I could have tested the programmes sitting behind my PC and not expose the machines to the risk of a stupid programming mistake.
It's funny you say that. I was realising that an OpAmp is no more than a SERVO SYSTEM. When fine tuning a process, I never have to analyse the servo in the system, just tell it what to do and hope it doesn't reach its limits of operation.
You change a process and you get a result. It might not be what you expected and you try to explain what just happened, but proving your explanation is a lot more difficult. These amps are the same.
A simulation is far more than a theoretical calculation. Its a copy of your process that you can run offline. With a simulation your only concern is if the model is correct, in a real process it's if you can measure it correctly and don't blow it up.
Back when I was programming robots it would have been great to have been able to do simulations. I could have tested the programmes sitting behind my PC and not expose the machines to the risk of a stupid programming mistake.
Last edited:
I understand you, been there done that. From what I recall, the vast majority of proponents of using small capacitance rely on their ears and subjective feelings, while those who advocate big capacitance tend to base their opinions in EE theories and formulae. Who is right? I tend to agree with the second ones, but who am I to judge what are you (or you 'believe' you are) hearing?
As a side note, could it just be that due to LM3886's high PSRR it just become unnoticeable having an small amount of capacitance?
Now that regulators have been metioned, hacing in mind all pros and cons exposed on this thread, isn't a regulator placed really close to the IC pins more appropiated than big capacitors? I mean, you could theorically get away with an small capacitance.
Hello, i have one question.
When you say to prove you mean to measure ?
Thanks a lot and kind regards,
gino
Hello again and i must admit that this is a very important aspect but also quite beyond my ability to understand.
I have a basic question by the way.
If you want to improve the current ability of an amp (assumed, of course, no current limiting circuits in it)
Would you replace the transformer or the PS caps with higher rating parts ?
I mean would you double the VA or the uF, so to speak ?
This is my very old and recurrent dilemma ...
Thanks a lot and kind regards,
gino
Last edited:
If a sim would be the same as the actual process we would not need injection moulding specialists or skilled CNC craftsmen.
Moldfow, an injection molding simulator will give you an idea of how a new mold basically will function but the end product depends on a lot more parameters as machine type its used on, the quality of materials the mold is made from and a lot of tolerances in process parameters.
In many cases a moldflow sim will not represent the reality and often the the mold has to be modified to give the desired result.
In the case of the injection molding process this is sometimes difficult but in the end its easy to check because the end result is a very objective measurable result.
What I start to find out is that this is not the case with the result of the design of an audio amplifier.
As long as only the subjective listening experience is the only way to qualify the end result, endless discussions will continue
As far as I understand regulators can not handle high peaks so will reduce the dynamics of the output.
I have no experience with sims of injection-moulding but it sounds like the modelling of mould and material are the problem. It's even difficult to keep thing stable for real. I thought you would see the parallel to a very fast Servo system.
Dear Mark, I have not enough knowledge of an audio amplifier to be able to compare the chip amp to a very fast servo system.
I know how a servo systems work but I think that even in this case its only one of the components in a vast bigger system that will produce the end result.
Last edited:
Hi and thanks a lot for the kind and valuable reply.
In my humble experience the best sounding amps i have listened they had all very "robust" power supplies.
Some speakers put a big stress on amps.
Thanks again and kind regards,
gino
Hello again and i must admit that this is a very important aspect but also quite beyond my ability to understand.
I have a basic question by the way.
If you want to improve the current ability of an amp (assumed, of course, no current limiting circuits in it)
Would you replace the transformer or the PS caps with higher rating parts ?
I mean would you double the VA or the uF, so to speak ?
This is my very old and recurrent dilemma ...
Thanks a lot and kind regards,
gino
As I might have already mentioned, no one should be fooled into thinking that a transformer/rectifier/capacitor power supply is simple. The circuit looks simple but, with a dynamic load current, its operation is not simple.
It depends what you mean by "improve the current ability". There are several things that the amp and power supply must be able to do, in terms of providing current:
- send current to load, with accurate amplitude timing, upon demand
- provide enough average and peak power to the load, at all times, upon demand
- Avoid clipping
To your question: If your transformer is rated for enough average Volt-Amps, then you only need to worry about having enough capacitance. For a robust system, I like to size the transformer so that it has VA = 2X the rated max output power per channel, of the amplifier, times the number of channels that will be powered by the transformer. The capacitance and the rated max average sine output power are redundant quantities. i.e. If you know one of those then you can calculate the other one, assuming that clipping can be allowed to start at the point when the rated max output power is reached. I have already posted the equations and a spreadsheet for calculating a minimum required capacitance for a class AB amplifier, given the other relevant system parameters (which are the transformer secondary's rated max RMS current at max average output power, the worst-case rectifier voltage drop at max average output power, the AC Mains frequency, the voltage rating or ESR of the capacitors, the amplifier's dropout (Vclip) voltage, and the nominal load resistance).
Using less capacitance than that does not necessarily mean that the amplifier won't sound as good, or that it will sound better. In this context it only means that it will have a lower RATED maximum output power level, where clipping starts. Also, since the calculations assume a worst-case load current, most real music will still be fine, even if the volume knob is turned up to the max output power position.
I like to design for the mathematical worst-case, so that performance during "normal" cases is far from any physical limitations. So I like to size the capacitances and transformers so that they will at least not cause any gross distortion (e.g. clipping) ALL the way up to the point where the user commands maximum rated output power, for which I also assume the worst-case signal in terms of capacitor voltage dip (a constant DC current at the peak output level that's implied by the max rated power). It might seem like a bit of overkill, compared to some designs. But I like "bulletproof".
The caps discharge current when music plays (see the link in my signature; the PSU caps MAKE the music). That makes their voltage sag lower. When a charging pulse occurs [every 1/(2 x fmains) second], the rectifier turns on and the transformer secondary sends current to the caps. The rectifiers are usually only "on" for about 20% to 40% of the time (depending largely on capacitance value; shorter times for larger values), so the rectifier current pulses' peak amplitude is usually much larger than the load current amplitude, so that the total transformer secondary and rectifier current, over time, is the same as the total load current, over time (plus the losses due to not being 100% efficient). [Nevertheless, note that it is typical for the capacitors to not be fully recharged with only one charging pulse, which can make most of the "simple" calculation methods, including mine, partially invalid in certain cases, in the direction of requiring too little capacitance.]
The peak rectifier current is even larger than the "usually ... 20% to 40%" implies (i.e. 5x to 2.5X), because the amplifier might be only 67% efficient, with the rest being dissipated as heat. So that alone makes the ratio of rectifier peak current to average output current more like "7.5X to 4X".
You don't need to worry too much about the peak currents, because most parts are rated for their average or RMS current capability, even though they are able to handle much larger peaks, for short times.
All of the above really only covers the handling of large low-frequency signals, and not clipping unexpectedly. For "accuracy", we also need to think about the higher frequencies, including the frequencies needed internally only, by the feedback system. It will have to be sufficient, here, for now at least, to say that we should plan to be able to provide a minimum of 10X the highest audio frequency, for the internal feedback's operation. That high-frequency current must come from our capacitors, accurately, on demand.
In terms of electronics in general, something like 250 kHz is not considered to be a high frequency. In fact it is considered to be quite low. However, to get the desired accuracy and precision, we still need to be very careful with the physical implementations of our circuits. The enemies are series inductance and resistance. They need to be as low as possible. Otherwise, response speeds get slower, in a frequency-dependent way, and fast edges tend to get blurred, which I imagine could affect (or destroy) soundstage imaging.
The parasitic inductances are mostly caused by the physical sizes of conductors, i.e. their lengths. That includes the parasitic inductances of components such as capacitors. "Never, ever" use axial-lead capacitors, for reservoir or decoupling duty. Use the smallest lead-spacing parts with the lowest ESR. And use the absolute-minimum PCB trace and wire lengths (and the fattest widths). [Power and ground planes are usually best, if done correctly.]
Parasitic resistances are similar, in that longer conductors have more resistance. Minimize conductor lengths and use low-ESR (equivalent series resistance) capacitors.
Both the inductance and resistance of a capacitance can be reduced by using multiple smaller caps in parallel, instead of one larger one. The overall capacitor cost is usually lower, that way, too. And also, the effects of failed and out-of-tolerance caps are diminished and thus reliability is improved. Also, the charging current pulses are divided among multiple parallel caps, making it possible for each of them to have a longer lifespan due to less ESR-dissipation heating, depending on how they are rated. It's a rare potential for a "win-win-win-win" situation. Take advantage of it.
As we add more and more capacitance, the benefit of adding each uF starts to diminish. If capacitance was closer to zero cost and zero size, we would probably always want to use as much as possible. But in the real world, the cost and size usually cause us to pick some point, above which "it's not worth it" to add more. Usually, at some point before THAT point, we could do more good by using a better layout, anyway, for example.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.