Power Crest Factor

[ChocoHolic]

Hi Ho!
...back again....
During last time I had again the chance to do some examinations for
my SubWoofer project.
Active system with SMPS and Class-D amp.

During my last examinations I was curious about the power crest factor.
This ratio is important for me because especially the SMPS will have to handle high peak power, but the thermal design does not need to allow 1kW all the time.... and I guess the speakers voice coil has also some sort of privat opinion regarding survivable average power...

I measured some power crest factors in my current sub woofer, in order to get the information about the music program behind all the filters...
I "measured" the speaker signal for 5sec. with 2ks/s.
2ks/s is OK to measure the low frequency signal of a sub, but
5 sec. is just a short sequence from the entire song. I played
around and the values seem to be quite reproducable at various
positions of the songs.
I just measured the voltage and calculated the power.
...yes, you are right the real power might be different due to
the complex impedance of the speaker...
But just for getting a feeling about the power ratio, this method should be OK.

My findings for the power crest factor ...so far:
"Eye In The Sky" by Alan Parson : about 24
"Without Me" by Eminem: about 20
Some ordinary pop song with additional noise from the radio (still searching title....) : about 16


Attached "Eye In The Sky".
Would be interesting to hear which ratios you might have observed.



Cheers
Markus

[ChocoHolic]

Nobody out there, who has done similar examinations???!

[d3imlay]

Do I understand that you're measuring crest factor, as defined as the ration of the Peak to the RMS value of a waveform, of an audio signal? I deal with crest factor with my work on power supplies and inverters. A crest factor of 2.5 to 3 is considered high. You've seen a C.F of 24 with an audio signal? I'm not in a position to dispute your data. I'm just amazed that it could be that high.

[peranders]

Crest factor is something with continuous signals!

[quasi]

I agree with d3imlay. When I was in the UPS / power protection business "Crest Factor" was defined as the ratio of the peak current drawn by the load over the RMS current drawn by the load. If the UPS was powering a simple load then the "sine wave" crest factor of 1.414 would apply and little consideration was given to the UPS "Crest Factor" capability.

However if the UPS was powering a complex load (switching power supplies, wierd lights, some lab / medical equipment) then cosideration had to be given to repetitive peak currents drawn by the load. This could produce crest factors as high as 2 or 3 which meant that the UPS had to be able to deliver repetitive current peaks 2-3 times the RMS value.

A little long winded I know but I wonder whether we are talking about the same thing.

ChocoHolic seems to be talking about a type of music dynamic range capability and has termed it as "Crest Factor". This is outside the definition that I learned. Not saying it's not correct as often the same words are understood differently in different applications.

So ChocoHolic are you defining the power peaks contained in a music sample when compared to the average or are you defining something else?

Cheers

[Workhorse]

So there's a need of crest factorization.

The Crest Factor in audio amplifiers is defined as"

"The ratio of Peak amplitude of Musical Transient in terms of voltage TO continuous average amplitude of musical signal at -3dB point "

The Crest factor ranges from 3 to 20 in professional amplifiers, provided with less power supply sagging.

regards,
kanwar

[quasi]

ChocoHolic's view seems to be ralated to signal sources rather than an amplifiers specifications.

"The ratio of Peak amplitude of Musical Transient in terms of voltage TO continuous average amplitude of musical signal at -3dB point " also seems to relate to the signal rather than the amplifier.

So how can a crest factor be applied to an amplifier?

Is it related to the SOAR of the output stage i.e current capability for 100ms 10ms 1ms etc? Or Workhorse are you talking about the dynamic range of the amplifier that is normally expressed in dB?

I know that some amplifiers are rated for momentary peak current capability, is this the "crest factor" for amplifiers?


Cheers

[JensRasmussen]

I think you need to evaluate the frequency content of the program material also. Depending on where the energy is mostly concentrated you will gain some “dynamical power” by having a higher non loaded rail voltage and more caps. This gives you a lower RMS output power, but it might sound better.

Of cause this is a question of budget, and some might argue that the biggest transformer and the most caps you can fit into the chosen enclosure will be the optimal choice.

A friend of mine Peter John Chapman did an AES paper on the frequency content of different program material, I don’t remember what it is called, but try looking it up.

\Jens

[ChocoHolic]

Hi Ho!
Sorry that I did not look to this thread that long.
I am glad to see that there were coming more answers.
Great.

Yes of course, I borrowed the naming crest factor.
And definition in regard is not very clear.

I defined my power crest factor as:
cfp = max. instantanious power / average power
mathematical this is equal to
cfp = (Umax^2 / R) / ( Urms^2 / R ) = Umax^2 / Urms^2

The voltage crest factor would range between 4.4 - 5.
The same with the current crest factor, if we assume constant
resistive load (stupid assumption).

For the PSU this large ratio means that you will need a supply that
can deliver very high power for some time without sagging. Typical for around 0.1 s ... 0.3 s. But thermally you do not need to design it for continous peak power.
In a regulated SMPS this will give the provision for much smaller
designed inductive components. You will have to be able to handle the full power without saturation in storage magnetics (flyback or boost preregulator) and provide good coupling in forward magnetics. But thermally the demands are much more relaxed compared to continuos power applications. This will allow smaller
core sizes with higher number of turns with thinner wire (less number of strands in HF-litz) as long as the leakage inductances are
acceptable.

@Jens:
Yes , I agree to your sight that the frequency content is important.
In my application (subwoofer) the frequencies are low and peak load durations 0.1s...0.3s are that long, that the required caps would blow the planned geometric size of my subwoofer....

@Peranders:
Would you call music something discontinous or even interruptive?

[DSP_Geek]

Originally posted by ChocoHolic

For the PSU this large ratio means that you will need a supply that
can deliver very high power for some time without sagging. Typical for around 0.1 s ... 0.3 s. But thermally you do not need to design it for continous peak power.

Some music has deep sustained bass - Pink Floyd seems to love shaking the room. The Vangelis soundtrack for Blade Runner is another example: it has several seconds of a 25 Hz (I think, the scope's elsewhere) sinusoid about 1 minute into track 1. If your subwoofer uses a Linkwitz transform to extend F3, that'll push your supply requirements higher than with music.

If you're using the subwoofer for home theatre, then you're looking at some demanding requirements when Ahnuld blows something up or Tom Cruise shoots down a Bad Guy.
http://www.svsubwoofers.com/faq.htm#moviedemos
is a good site for frequency/time/amplitude measurements of movie bass.


Francois.