This is a complex problem. Voltage and current limits in the output load need to be determined.
Voltage is limited by the rails (transformer output).
Output collector current change is not limited by the output transistor slew rates.
I have seen figure of 30V per microsecond for 2n3055.
Some other circuit component must limit overall slew rate.
candidates are, the transformer secondary (impedance) or the smps output caps(max time dependant current) or the output load resistor or the loudspeaker.
In practical terms, if your premise is correct, then increasing the smps output caps should increase the measured overall slew rate of the amp.
There may be a ceiling to the benefit if other components become limiting.
Voltage is limited by the rails (transformer output).
Output collector current change is not limited by the output transistor slew rates.
I have seen figure of 30V per microsecond for 2n3055.
Some other circuit component must limit overall slew rate.
candidates are, the transformer secondary (impedance) or the smps output caps(max time dependant current) or the output load resistor or the loudspeaker.
In practical terms, if your premise is correct, then increasing the smps output caps should increase the measured overall slew rate of the amp.
There may be a ceiling to the benefit if other components become limiting.
No that's not what I'm tryig to say at all. This is not about how fast he amp can respond, or frequency bandwidth, this is about how it handles the volume envelope...for maybe 1/4 or half a second burst. That pick attack or drum hit...the "A" in a synth's ADSR.
ok I take your point, you are thinking about dynamic range for amplitude.
If a complex music waveform has amplitude excursions that exceed the input or output device supply rail, it gets clipped-regardless of caps present.Thats an analogue limitation.
Maybe caps can improve the performance for "NEAR the supply rail" excursions by stabilising it.
If so all forms of distortion should measure lower.
My point being that if an improvement is real then it should be measurable somehow.
I suspect that at normal music listening levels analog amps will occasionally clip transients.
If a complex music waveform has amplitude excursions that exceed the input or output device supply rail, it gets clipped-regardless of caps present.Thats an analogue limitation.
Maybe caps can improve the performance for "NEAR the supply rail" excursions by stabilising it.
If so all forms of distortion should measure lower.
My point being that if an improvement is real then it should be measurable somehow.
I suspect that at normal music listening levels analog amps will occasionally clip transients.
No, the problem is if a short-term current demand exceeds the capability of the supply to replenish the caps and maintain the rail voltage, then the rail voltage drops. The size of the caps determines the rate the voltage drops. If there is a large enough bank of caps, then the small drop causes a negligible dynamic range compression and the peak is "handled."
The measurement difference is:
1) without big bank of caps: no additional short-term current headroom
2) with big bank of caps: significant additional short-term current headroom
My claim is that if they design the power supply and following conventional amplifier to handle a significant continuous rating (which they do) but NO additional short-term current capability, then they can probably add a bank of caps for a reasonable price and little additional wieght, and have siginificant current reservfes.
The measurement difference is:
1) without big bank of caps: no additional short-term current headroom
2) with big bank of caps: significant additional short-term current headroom
My claim is that if they design the power supply and following conventional amplifier to handle a significant continuous rating (which they do) but NO additional short-term current capability, then they can probably add a bank of caps for a reasonable price and little additional wieght, and have siginificant current reservfes.
I dont get how dynamic compression happens when the rail volts drop. Do you mean that clipping happens at the lower rail voltage with some loss of signal data? larger caps could help in that area.
The continuous rating of the smps must have a reserve over the average amp dissipation. Ultimately the ps must source all current. For millisecond bursts some rail volt drop is inevitable. The tradeoff with large caps is a lower drop and a longer recovery. I dont have the math to calculate rail drop and recovery times with "normal" caps compared to large value caps.
Perhaps it is better to accept a shorter time of higher signal degradation during bursts. Even with the rail volts stabilised it is not enough to prevent clipping, which is the main effect of bursts.
Clipping should not happen, unless it is a pre-manufactured (whole) amp where they give you an excessively high output wattage figure, then tell you there's orders of magnitude higher THD too. Otherwise, the PSU current capacity is too small or the amp gain is too high.
We keep thinking modularly and pondering putting much larger capacitors, even a bank of them in place as part of the SMPS "module" but that is seldom space or cost effective and after all if you're picking a SMPS you're often doing it for acoustic reasons (I would suppose) rather than efficiency, while a huge capacitor bank implies that while you have a SMPS, it is acting too much like a linear PSU.
What resolves this? Your SMPS feedback could come from where rail voltage droop matters, at the amp output transistors as global feedback rather than local on the SMPS board. In some cases a mere piece of wire will replace the idea of a huge, expensive, added bank of capacitors with its own complications on inrush current and SMPS stabilization time before a low voltage shutdown.
We keep thinking modularly and pondering putting much larger capacitors, even a bank of them in place as part of the SMPS "module" but that is seldom space or cost effective and after all if you're picking a SMPS you're often doing it for acoustic reasons (I would suppose) rather than efficiency, while a huge capacitor bank implies that while you have a SMPS, it is acting too much like a linear PSU.
What resolves this? Your SMPS feedback could come from where rail voltage droop matters, at the amp output transistors as global feedback rather than local on the SMPS board. In some cases a mere piece of wire will replace the idea of a huge, expensive, added bank of capacitors with its own complications on inrush current and SMPS stabilization time before a low voltage shutdown.
If we turn our amps down to a whisper, then we will not get clipping. But that ain't gonna happen. At normal music listening levels, the music dynamic range means clipping will happen with high attack rates. with analogue amps
Most overdriven amps run out of voltage rather than current. <full stop>The SMPS voltage control loop is fast given > 40KHz sample rates. Massive caps on the output of SMPS regulated supplies are plain silly and a waste and they probably will cause it to become unstable and blow up, maybe yer amp and speakers too. there is no free lunch doing this,You would be better off making sure there are smaller caps near the amps OP stage for rise and fall time of slewing signals where wiring inductance maybe the limit.
rather than wondering , Test an Amp 1st find out what the slewing current requirements are. Then attach a SMPS which either supports that current or a smaller one and see what mayhem you can cause adding massive caps.
rather than wondering , Test an Amp 1st find out what the slewing current requirements are. Then attach a SMPS which either supports that current or a smaller one and see what mayhem you can cause adding massive caps.
Last edited:
Only if it's not designed properly and although some of the designs we see in the forums make me blink a couple times in disbelief, most of them aren't going to have clipping at "normal music listening levels".
Although some things about audio are either snake oil or user preference when it comes to what sounds better, having rail voltage stay high enough to support input signal x amp gain is drifting dangerously close to pure science, it can be measured and determined if PSU voltage is high enough and/or amp gain is too high.
Where does your argument end? No matter how much capacitance you have, you would still reach a point where it won't work if amp gain or rail voltage is too low. If you can't raise PSU voltage to handle the peaks at the volume level you want, because the amp itself can't handle that higher supply voltage, you have built an amp incapable of the volume level you claim is clipping. Otherwise, the small drop in rail voltage from not having a massive amount of capacitance is trivial on a SMPS, if it weren't the SMPS would shut down being out of regulation so even in best case, a huge capacitance addition would only cover a small increase in peak output, then if the design problem weren't fixed, it's in the same situation where it would clip.
Last edited:
Large supply filter/storage caps are a good idea in a SMPS as it is in conventional 60Hz rectifier supplies.
Where they really shine is in free running unregulated SMPS. This can be done in off-line converters to get rid of big steel core transformers, and depending on the input rectifier filter size, can contribute to limiting of 120Hz ripple more or less. Some car amps used to look at the line for regulation in PWM converters, a case where output filter size probably can only contribute to increased output stability. Large output filters rather than large input filters keeps the pulse load off the converter semiconductors, and can do a much more efficient job of averaging the load on the B+ than a Farad tacked right on to it.
I don't think there's any reasonable argument here about amplifier clipping and stuff. Of course the capacitance can't really change that more than a fraction of a dB for some very limited period of time. The trouble to argue about and solve is the ability of an amplifier to produce extremely high accuracy in the output when the supply is dancing around. It's pretty well known that this doesn't work so well. Just ask the shunt regulator opamp supply gurus, especially those who claim to notice serious increase in bass performance from very high supply stability... and this is just signal level stuff where the power input is essentially constant anyway.
In the case of a converter with output voltage feedback, I think you can design/redesign with that in mind to maintain stability with high output capacitance. In the end you'll just make the converter a little less hyper and allow the caps to do their business.
Where they really shine is in free running unregulated SMPS. This can be done in off-line converters to get rid of big steel core transformers, and depending on the input rectifier filter size, can contribute to limiting of 120Hz ripple more or less. Some car amps used to look at the line for regulation in PWM converters, a case where output filter size probably can only contribute to increased output stability. Large output filters rather than large input filters keeps the pulse load off the converter semiconductors, and can do a much more efficient job of averaging the load on the B+ than a Farad tacked right on to it.
I don't think there's any reasonable argument here about amplifier clipping and stuff. Of course the capacitance can't really change that more than a fraction of a dB for some very limited period of time. The trouble to argue about and solve is the ability of an amplifier to produce extremely high accuracy in the output when the supply is dancing around. It's pretty well known that this doesn't work so well. Just ask the shunt regulator opamp supply gurus, especially those who claim to notice serious increase in bass performance from very high supply stability... and this is just signal level stuff where the power input is essentially constant anyway.
In the case of a converter with output voltage feedback, I think you can design/redesign with that in mind to maintain stability with high output capacitance. In the end you'll just make the converter a little less hyper and allow the caps to do their business.
Last edited:
only those who don't know how to use feedback, don't. BTW it keeps "stuff from dancing around". you could also slow down a SMPS and the loop BW and add tons of C, but why. There is nothing magical about dynamic current in amplifiers to need tons of C.
Of course there's no magic, and feedback can't make energy come from nowhere.
Tons of C produces dynamic current in cases where feedback can't. In signal amplifier supplies, feedback rules (some sort of regular, just another amplifier of sorts anyway), only because Tons of C at that scale is a few hundred uF, which is routinely used before the regulator.
Tons of C produces dynamic current in cases where feedback can't. In signal amplifier supplies, feedback rules (some sort of regular, just another amplifier of sorts anyway), only because Tons of C at that scale is a few hundred uF, which is routinely used before the regulator.
There is no need for super high instantaneous current unless yer driving a large shunt cap with an amp. Then the amps transistors would be the weak link rather than the PS.
What's super high? It makes better sense when talking in terms of fidelity. Maybe your opamp supply has 50mOhm impedance, and you think it needs that for good bass even though the load impedance is 10kOhm.
Now look at the power amplifiers. Is the supply rejection 30dB worse than the signal amplifiers? Happens often. What is the equivalent power amp supply impedance, to match the signal amp performance, when the load is 8 ohms at 50 volts?
Just taking 10k by 8 gives you 1250. Say you have 300uF ahead of your opamp regulators (and we know that 300uf on an opamp is Not for high frequency performance). 1250 times 300 gives you 375,000 uF.
What should the power amp supply impedance be at 50Hz if the signal amp supply impedance is 50mOhm? Can you get it with some kind of regulator and 4700uF?
Why don't we just use 10uF on the opamps? Is it just because no "audiophools" would buy it?
Now look at the power amplifiers. Is the supply rejection 30dB worse than the signal amplifiers? Happens often. What is the equivalent power amp supply impedance, to match the signal amp performance, when the load is 8 ohms at 50 volts?
Just taking 10k by 8 gives you 1250. Say you have 300uF ahead of your opamp regulators (and we know that 300uf on an opamp is Not for high frequency performance). 1250 times 300 gives you 375,000 uF.
What should the power amp supply impedance be at 50Hz if the signal amp supply impedance is 50mOhm? Can you get it with some kind of regulator and 4700uF?
Why don't we just use 10uF on the opamps? Is it just because no "audiophools" would buy it?
Most of your questions can be solved by using good wiring practices, proper supply decoupling, and regulated supplies AND they will be light years ahead of a standard unregulated 60 Hz supply. The PSRR problems are by and large due to large signal PS modulation causing dynamic sag in unregulated rails coupling into preceeding stages. Typical op amps esp chip amps don't have seperate power and signal supplies so they require more C than simple scaling would suggest.
Last edited:
We seem to agree that clipping is only marginally improved by high cap current reserve.
Is clipping happening in the majority of analogue amps? I think yes.
Many amps have LED display for output reaching rail volts.I am not convinced that a short high attack current burst would be visible to the eye (wonder if anyone has tried a fast sensitive photomultiplier) In a test rig (transistor amp)you might see a stepup in the overall distortion as the gain is raised from zero (due to bursts reaching the rail volts).I don't know how much clipping contributes to overall distortion, but I imagine it could have a knockon effect on any of them.If the overall amp distortion is 0.01%, then for a sine wave, that is the rough (very) equivalent of one peak in every 10000 maxed out.
At 10kHz thats one a second.
You can dynamically reduce the stage gain to eliminate clipping, but this butchers the music dynamics anyway.
I wonder if at normal listening levels, current bursts are exceeding rail volts and if rail volts could be raised by special transistors say to 240v. would the bursts be audible anyway.
Is clipping happening in the majority of analogue amps? I think yes.
Many amps have LED display for output reaching rail volts.I am not convinced that a short high attack current burst would be visible to the eye (wonder if anyone has tried a fast sensitive photomultiplier) In a test rig (transistor amp)you might see a stepup in the overall distortion as the gain is raised from zero (due to bursts reaching the rail volts).I don't know how much clipping contributes to overall distortion, but I imagine it could have a knockon effect on any of them.If the overall amp distortion is 0.01%, then for a sine wave, that is the rough (very) equivalent of one peak in every 10000 maxed out.
At 10kHz thats one a second.
You can dynamically reduce the stage gain to eliminate clipping, but this butchers the music dynamics anyway.
I wonder if at normal listening levels, current bursts are exceeding rail volts and if rail volts could be raised by special transistors say to 240v. would the bursts be audible anyway.
Last edited:
^ The display in amps seldom if ever, specifically displays output reaching rail volts. Can you provide an example where it does? You might mentally note a coincidence between the level displayed and when a particular amp design clips since an amp with excessively high gain per PSU voltage approaches its limits as signal level rises, but using signal level monitoring, not signal vs voltage comparison monitoring which would be needed to know if level reaches rail voltage limits, an amp can put that meter display *into the red* without clipping.
If you say the "majority" of analog amps includes all those sold at retail, most of which built to have high wattage : price ratio, then yes many of those can be made to clip. If you count only DIY amps, it's up to the designer if they try to push the amp until it clips. Just because an electronics superstore sells some supposed "500W" amp for $120 USD that clips, that doesn't compel a DIY amp builder to try to match that result.
If you say the "majority" of analog amps includes all those sold at retail, most of which built to have high wattage : price ratio, then yes many of those can be made to clip. If you count only DIY amps, it's up to the designer if they try to push the amp until it clips. Just because an electronics superstore sells some supposed "500W" amp for $120 USD that clips, that doesn't compel a DIY amp builder to try to match that result.
Last edited:
Sorry for sloppy terminology, they work by comparing input and attenuated output signal and feeding a differential amp. However the end result is the same.
I agree we should be able to design out clipping. A digital storage oscilloscope
should give us the input "signal burst" maximum volts for input and output headroom. This will depend on the accuracy of the signal conditioning between the sound source and the input
I have never seen any figures for burst maxima but I would be surprised if they are less than 5X the "metered" rms measured values (pure guesswork)
Found some measurements done on transistor & valve amps for transient bursts.It seems to indicate that output signal bursts are several times the rms measured max output for these amps. Notice that the transistor amp was running peaks right up to the rail and would not go beyond for bursts.
The valve amps however went beyond the same max output voltage, albeit with increasing dynamic compression. This shows that valva amp bursts are at least twice the maximum non clipping output.
The transistor amp had the same input signal so was presumably missing this wave information
Of course this could be a design deficiency in the transistor amp.
www.audionote.co.uk/articles/tech_amp_peak_01.shtml
Audio Note
The valve amps however went beyond the same max output voltage, albeit with increasing dynamic compression. This shows that valva amp bursts are at least twice the maximum non clipping output.
The transistor amp had the same input signal so was presumably missing this wave information
Of course this could be a design deficiency in the transistor amp.
www.audionote.co.uk/articles/tech_amp_peak_01.shtml
Audio Note
Last edited:
- Status
- Not open for further replies.