some DVM's have a min/max setting such as Fluke 87, I am not sure how fast they are, but I have used them to measure transients when I used to work and customers were always suprised how high line spikes would go or how low the line would drop. If they didn't believe the meter then I would ask for the fancy one with the printout and time display. ie one spike or dropout @ 8:13AM = system crash. A scope is good too.
Hi Guys
I scoped my amp output the other day while blasting music. I expected to see a 1W average level for the loudest SPL I could stand. It was about as expected. My DVM is not fast enough to capture peaks.
Most modern music is highly compressed to make the most use out of the limited output of the medium. 2V is a big signal but you only get distortion past 16-bits all as ones. The LED meters on the CD players just go solid to the max for many discs. Older discs and other music styles will have bouncing LEDs and a bouncy signal on the scope. With pop the scope just displays a constant amplitude envelope of about 5Vpp. The dynamic music displays a lower average level but has peaks that are mostly twice the average.
For dynamic headroom purposes, you could allow a 4x voltage space for transients. This suggests that if you wanted to build a low-watt amp, with very good performance at 1W and below - since you can't listen to it that loud for very long before you won't be hearing anything! - then 4x V produces 16x P, so 16W for assurance against clipping. You could build this easily with +/- 24V rails and set it up to be class-A to eliminate XO distortion. You could take it a step further and make it class-G with lower rails at +/- 12V. Transition would occur at about 9Wrms, so not very often.
The nice thing about modern class-G amps is that the rail transition does not invoke distortion when the drivers are supported by the high rails all the time and the low-V diodes are shottky. Operating class-A maintains very low THD and the low-V helps keep dissipation in check.
Have fun
Kevin O'Connor
londonpower.com
I scoped my amp output the other day while blasting music. I expected to see a 1W average level for the loudest SPL I could stand. It was about as expected. My DVM is not fast enough to capture peaks.
Most modern music is highly compressed to make the most use out of the limited output of the medium. 2V is a big signal but you only get distortion past 16-bits all as ones. The LED meters on the CD players just go solid to the max for many discs. Older discs and other music styles will have bouncing LEDs and a bouncy signal on the scope. With pop the scope just displays a constant amplitude envelope of about 5Vpp. The dynamic music displays a lower average level but has peaks that are mostly twice the average.
For dynamic headroom purposes, you could allow a 4x voltage space for transients. This suggests that if you wanted to build a low-watt amp, with very good performance at 1W and below - since you can't listen to it that loud for very long before you won't be hearing anything! - then 4x V produces 16x P, so 16W for assurance against clipping. You could build this easily with +/- 24V rails and set it up to be class-A to eliminate XO distortion. You could take it a step further and make it class-G with lower rails at +/- 12V. Transition would occur at about 9Wrms, so not very often.
The nice thing about modern class-G amps is that the rail transition does not invoke distortion when the drivers are supported by the high rails all the time and the low-V diodes are shottky. Operating class-A maintains very low THD and the low-V helps keep dissipation in check.
Have fun
Kevin O'Connor
londonpower.com
Hi,
That is about right, most good recordings spend >80% of the time
at voltage levels <20% of maximum, closer for poorer recordings.
12dB is probably as high as it gets for "rock", but not other stuff,
you really need about 20dB IMO to cover most quality recordings,
and even more for "audiophile" hairshirt stuff.
rgds, sreten.
Last edited:
An oscilloscope shows voltage, not power, correct?
You would have to multiply the voltage by the current, at each instant, to see power, since the impedance is not constant.
Hi Guys
Yes, power may vary with frequency especially with multiway systems and with complex signals.
How much voltage is required to drive the speaker is the most important parameter as that drives the decision for supply rails and everything else. Assuring sufficient current for the load is not a problem, particularly if you are designing for a modest power as I intend, and as it seems might be the intention of the original poster.
The cost of having double or quadruple the nominal load current when considering a 16-20W amp is inconsequential. For an 8R load and a 16Wrms output, 32Wpk, you need 16V at 2Apk. If you have a speaker that dips to 4R then you need 4Apk, for 2R then 8Apk. These are all small numbers, easily handled by a single pair of MJL3281/1302. If you were going class-A, more pairs should be added.
Similarly, where a 20VA PT would support the split-rail 16W into 8R, doubling or quadrupling the VA will not add greatly to the cost. Production amps usually use twice the watts as the VA rating, since they don't know what load someone might use.
For myself, I know the load presented by my speakers to the amp and it is benign. I could skimp on the caps and PT size and not notice. Knowing the load makes the amplifier design simple and sane.
Filtering is based on the ripple current values, and the actual capacitance just follows suit. Most people just throw 10mF per rail and assume this is enough. On the other hand, some go crazy and have 100mF per rail which is completely unnecessary and might even increase hum.
note m=milli
Have fun
Kevin O'Connor
londonpower.com
Yes, power may vary with frequency especially with multiway systems and with complex signals.
How much voltage is required to drive the speaker is the most important parameter as that drives the decision for supply rails and everything else. Assuring sufficient current for the load is not a problem, particularly if you are designing for a modest power as I intend, and as it seems might be the intention of the original poster.
The cost of having double or quadruple the nominal load current when considering a 16-20W amp is inconsequential. For an 8R load and a 16Wrms output, 32Wpk, you need 16V at 2Apk. If you have a speaker that dips to 4R then you need 4Apk, for 2R then 8Apk. These are all small numbers, easily handled by a single pair of MJL3281/1302. If you were going class-A, more pairs should be added.
Similarly, where a 20VA PT would support the split-rail 16W into 8R, doubling or quadrupling the VA will not add greatly to the cost. Production amps usually use twice the watts as the VA rating, since they don't know what load someone might use.
For myself, I know the load presented by my speakers to the amp and it is benign. I could skimp on the caps and PT size and not notice. Knowing the load makes the amplifier design simple and sane.
Filtering is based on the ripple current values, and the actual capacitance just follows suit. Most people just throw 10mF per rail and assume this is enough. On the other hand, some go crazy and have 100mF per rail which is completely unnecessary and might even increase hum.
note m=milli
Have fun
Kevin O'Connor
londonpower.com
Hi,
The proper way of looking at voltage levels in dB
(20log[Vout/Vin]) automatically converts to power.
rgds, sreten.
If you have two ch memory oscilloscope, you can see voltage/current in a real way, while send music to loudspeaker.
i'm sure you find a surprise, depeand enveloope you listen.
current: one shunt eg. 10mR in series at gnd to speaker.
Voltage: gnd after shunt (this need a differential iso probe) or measure with gnd_probe to gnd and probe on +speaker. in this case you see voltage without DDP.
is good that you calculate and use all formulas, after this measure (not beefore)
i'm sure you find a surprise, depeand enveloope you listen.
current: one shunt eg. 10mR in series at gnd to speaker.
Voltage: gnd after shunt (this need a differential iso probe) or measure with gnd_probe to gnd and probe on +speaker. in this case you see voltage without DDP.
is good that you calculate and use all formulas, after this measure (not beefore)
The method suggested in post # 2 is very good, provided that you understand the method and do it correctly. What you must not do, when following this method, is attempt to adjust the volume of the system while listening to the test tone. The test tone is for measurement only, you can in fact completely disconnect your speakers while playing the test tone (if you don't want to subject your speakers or ears to a somewhat loud monotone); it is not to be used as an acoustic reference level. Many, many people posting into that thread completely misunderstood that important point, and as a result they failed to see the merit in this method. It is a very clever and accurate method. The only caveat is that the result is valid only for the recording(s) that you used to set your reference volume level in the third step.
Hi Guys
A steady test tone is not exactly like dynamic music with peaks. The original Q was about determining how high the peaks were. The best way is to scope it or use a fast meter or RTA that can hold the highest reading. From there you calculate the power based on the nominal impedance of the speaker system.
For such a determination, it does not matter that much if one knows about crest factors, etc, as music does not conform to those rules. There are rough rules of thumb about peak size to average level. I've seen lots of music signals on scopes over the past few decades, so I have a good idea what those rules of thumb are.
Besides, to play a steady test tone through the speaker at a loudness you "believe" to be the same as the music peaks is pretty sketchy. Odds are that the steady tone will be played at a lower level than the real music peak, becuase your ears will tire of it almost immediately. The other thing is that removing the load after hearing the test tone level to make the measurement in silence will give a slightly higher reading than with the load. This has to do with the regulation of the output. Although most amps try to be perfect voltage sources, they are not and removal of the load almost universally causes a jump in output level. Discontinuous loads are a major cause for amp failure in live situations.
I don't believe the original poster was trying to establish an exact figure for the peaks, just a ballpark value.
Have fun
Kevin O'Connor
londonpower.com
A steady test tone is not exactly like dynamic music with peaks. The original Q was about determining how high the peaks were. The best way is to scope it or use a fast meter or RTA that can hold the highest reading. From there you calculate the power based on the nominal impedance of the speaker system.
For such a determination, it does not matter that much if one knows about crest factors, etc, as music does not conform to those rules. There are rough rules of thumb about peak size to average level. I've seen lots of music signals on scopes over the past few decades, so I have a good idea what those rules of thumb are.
Besides, to play a steady test tone through the speaker at a loudness you "believe" to be the same as the music peaks is pretty sketchy. Odds are that the steady tone will be played at a lower level than the real music peak, becuase your ears will tire of it almost immediately. The other thing is that removing the load after hearing the test tone level to make the measurement in silence will give a slightly higher reading than with the load. This has to do with the regulation of the output. Although most amps try to be perfect voltage sources, they are not and removal of the load almost universally causes a jump in output level. Discontinuous loads are a major cause for amp failure in live situations.
I don't believe the original poster was trying to establish an exact figure for the peaks, just a ballpark value.
Have fun
Kevin O'Connor
londonpower.com
You want to know power, but it's easier to know volts. From there you can figure your peak power.
Here is how to do it:
Voila! You now have your peak power at the volume setting where you measured. As long as you don't clip the amp at that setting, the number is valid.
Here is how to do it:
- Set your volume level where you normally do. Leave it there. (You may have more than one normal level, you can test any or all of those)
- Play one or both of the test tones linked to here.
- Measure the tone with an AC voltmeter. A 2 volt scale will most likely be best.
- Multiply the voltage you have measured by 4 (12 dB)
- You now have your peak voltage at that volume setting.
- Use Ohm's law to calculate your peak power from your speaker's impedance. W = V2 / R. If you don't know the impedance, use the DCR, as that is the lowest it should ever go.
Voila! You now have your peak power at the volume setting where you measured. As long as you don't clip the amp at that setting, the number is valid.
My normal listening level is 80 dB at 1 meter, so on average I'm using less than 1W of power, but I want to figure out how much power the speakers are drawing during the fast transients during real music. What is the best way?
I think it's a very good way to specify the listening level at 1 meter than at the seating place. An average level of 80 dB SPL at 1 meter could be adopted as a reasonable standard for the dimensioning of home systems.
If we consider that sensitivities of loudspeakers (for 1 W, that is 2.83 V for 8 Ohm at 1 meter) of 84 dB as low, 90 dB as midle and 96 dB as high, the average power for 80 dB SPL that hey have to handle is respectively 0.4 W, 0.1 W and 0.025 W.
I hold the following data from a french professional guy who owns a real reference system at home, able to more 150 dB SPL :
The highest dynamics he ever met on (very rare) DVDAs or SCACDs are 68 dB (+/-34 dB).
He says it is very difficult no manage such dynamics in home systems, they need a dedicated room and very powerful loudspeakers.
Best dynamics he ever met on (very rare too) CDs are 52 to 55 dB (a little more than +/-26 dB ). This means that the highest level you may encounter wiht CDs at 1 m is 108 dB for an average listening level of 80 dB at 1 m.
With a sensitivity of 89 dB/W (more exactly : 89 dB SPL /2.83 V/1 m) the amp power needs to be a bit more than 100 W (more exactly 30 Vms for 8 Ohm) to handle the peaks.
The fact is that many instruments have very high dynamics, however to make well balanced records, they can't be mixed all together with respecting the full dynamic of each one, a bit of compression is always required.
Beware that as drivers are mostly voltage driven, converting V to W leads to errors.
This is specially important in the region below 100 Hz where the impedance of the drivers shows peaks much higher than 8 Ohm.
Amps do not need to handle much power for long, they need to hande high voltage peaks for short, about +/-43 Vpk for he above example.
All this leads to an other question : do drivers used at home ever significantly suffer of thermal compression ?
I think it's a very good way to specify the listening level at 1 meter than at the seating place. An average level of 80 dB SPL at 1 meter could be adopted as a reasonable standard for the dimensioning of home systems.
If we consider that sensitivities of loudspeakers (for 1 W, that is 2.83 V for 8 Ohm at 1 meter) of 84 dB as low, 90 dB as midle and 96 dB as high, the average power for 80 dB SPL that hey have to handle is respectively 0.4 W, 0.1 W and 0.025 W.
I hold the following data from a french professional guy who owns a real reference system at home, able to more 150 dB SPL :
The highest dynamics he ever met on (very rare) DVDAs or SCACDs are 68 dB (+/-34 dB).
He says it is very difficult no manage such dynamics in home systems, they need a dedicated room and very powerful loudspeakers.
Best dynamics he ever met on (very rare too) CDs are 52 to 55 dB (a little more than +/-26 dB ). This means that the highest level you may encounter wiht CDs at 1 m is 108 dB for an average listening level of 80 dB at 1 m.
With a sensitivity of 89 dB/W (more exactly : 89 dB SPL /2.83 V/1 m) the amp power needs to be a bit more than 100 W (more exactly 30 Vms for 8 Ohm) to handle the peaks.
The fact is that many instruments have very high dynamics, however to make well balanced records, they can't be mixed all together with respecting the full dynamic of each one, a bit of compression is always required.
Beware that as drivers are mostly voltage driven, converting V to W leads to errors.
This is specially important in the region below 100 Hz where the impedance of the drivers shows peaks much higher than 8 Ohm.
Amps do not need to handle much power for long, they need to hande high voltage peaks for short, about +/-43 Vpk for he above example.
All this leads to an other question : do drivers used at home ever significantly suffer of thermal compression ?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.