Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Charles Johnson]

Greetings all,

During the not-too-distant past, a few questions regarding amplifiers have formed in my mind and I've decided that it's time to "unload" them and de-clutter my mind a bit (at least, of some potential questions ). To whit:

1) I noticed that Bink said that 2 of the large dummy loads used during the Amp Shootout found their way over to the shop of Third Ear Sound, who used them to "burn in" a rather large number of QSC PL6.0s. That statement has raised at least 2 questions in my mind...

a) What does it mean to "burn in" an amplifier (and why would you want to do it)?

b) How does one go about "burning in" an amplifier?

2) Drawing on the results of Bink's Amplifier Shootout, in which data was gathered and presented, showing the various amplifiers' output at 20Hz, 1KHz, and 20KHz (along with their rated 1KHz and 20Hz-20KHz output), I've decided to test out at least the amplifier model that I'm currently using on my subs (QSC PLX3402) and see how linear the output is as the output level is raised. To this end (and with the help of fellow LABster Phil Ouellette), I've built a little box that will take amplifier-level voltages and knock them down to a level more suitable for my USBPre, thus allowing me to use Smaart's Transfer Function mode to evaluate the linearity of the amplifier wrt frequency. My problem, however, is what to use for source material. If I were to use the Pink Noise track from Bink's handy test CD (which has an approx 12db crest factor, if I'm not mistaken), then I'll be tickling the clip lights on the amplifier before it is outputting at mere 1/10th of its rated output. I could, of course, continue to raise the input level and just let it clip, but I don't know what effect that would have on the measurements. Trying to keep reality in mind, would using pink noise w/ a 6db crest factor be a good choice for source material? Keep in mind that I'm not so much looking to gather data for official publication, but rather just to get an idea of where the roll-off starts (and at what rate). FYI, for dummy loads, I have, with Langston Holland's help, built (2) UglyBuckets (tm), which have ended up having a load impedence of ~4.3 Ohms each (according to my Fluke 12, that is).

3) Regarding amplifier duty cycle, I know that it's common to rate AC current draw at 1/8 and 1/3 duty cycles (with a pure sine wave at full power being a 50% or 1/2 duty cycle, yes?). My question then is this: How would one go about reproducing a duty cycle of, say, 1/8th?

I appreciate your thoughts/comments/opinions/suggestions! Cheers!

Charles "inquisitive as ever" Johnson

PS For the record, one modification, at the suggestion of the aforementioned fellow LABster, that I made to the UglyBuckets I built was to drill (3) holes (I used a 5/16" drill bit, since 1/4" seemed too small and I couldn't get the 3/8" bit out of the holder! ) in the tops for the buckets, so as to prevent any steam pressure (albeit quite unlikely) from building up.

[Gian P. Portanova]

Hi Charles,

1a)  All amplifier mfgr's should already be burning-in amplifiers as one of the final stages of their production cycle.  Burn-in will eliminate the "infant mortality" units from the production run.  These usually will reveal themselves within 24-48 hours of burn-in.  Third Ear is probably going one step further and putting in additional hours, hoping to better prove field reliability.

b)  I suggest you use the EIA RS-426 (specific bandwidth, gated pink noise spec) as your source.  Typical load of 4 ohms should suffice.  You may want to power up/down on an hourly cycle in order to achieve the most thermal cycling.

2)  How can you verify linear response with random noise as your source?  Use a sine wave and sweep from 20Hz to 20KHz.  Do it at specific power levels (100W increments) until you get to max power for your amp.  Bink used three specific sinewaves to verify max power.  

3.  The 1/3 and 1/8 refers to the amp's max output power, not duty cycle.  1/3 max output sinewave power of an amp is equivalent to high compressed audio (music).  1/8 max output sinewave power is equivalent to typical music.

I will assume that your "uglybuckets" are a piece of wire in water?  Please note to keep the load resistor as non-inductive (purely resistive) as possible since this could affect your high frequency measurements!

Have fun....

[John Roberts {JR}]

I am always apprehensive when consumers try to "confirm" specifications. While an educated consumer is a good consumer there are a lot of subtleties involved in testing, especially at higher power levels.

As you've already been advised the proper tool for measuring frequency response is swept sine wave. This can be unnaturally stressful to the UUT if not careful as few units will encounter full power 20kHz tones in use.

Yes, burn-in is to catch infant failure, and more importantly in the case of power amps, proper heat sink assembly (are devices properly torqued, insulated, greased and ready to rock?).

Repeating a burn-in after receipt can confirm that nothing was affected during shipment. Amplifiers are relatively heavy and as a recent thread has suggested many carriers are less than gentle.

Regarding duty cycle, as I've pondered openly here before, AFAIK there is no standard for this. The closest is some specific noise sources used for loudspeaker testing but this subject is not well understood. The fact that different markets have different needs further complicates a standard which will be difficult enough to resolve let alone build in enough range of measurement to generate meaningful metrics for matching amplifiers to applications.

Good luck and have fun... you will surely finish smarter than you started.

JR

[Charles Johnson]

Gian wrote

1a)  All amplifier mfgr's should already be burning-in amplifiers as one of the final stages of their production cycle.  Burn-in will eliminate the "infant mortality" units from the production run.  These usually will reveal themselves within 24-48 hours of burn-in.  Third Ear is probably going one step further and putting in additional hours, hoping to better prove field reliability.

Yes, I pretty much assumed that it was an attempt to weed out any "infant mortality" units, however, after "Googling" the subject and finding references to "burning in" amplifiers for (up to) 100 hours (as was suggested by one site), I figured I'd ask here. Granted, most of the sites that turned up as results were audiophile pages, but audiophiles have been known to be right about some things from time to time

Quote:

b)  I suggest you use the EIA RS-426 (specific bandwidth, gated pink noise spec) as your source.  Typical load of 4 ohms should suffice.

Ah - and where might I obtain such an audio file (or the specifications needed to make it)?

Quote:

You may want to power up/down on an hourly cycle in order to achieve the most thermal cycling.

By thermal cycling do you mean powering down for a certain length of time to allow the AUT (amplifier-under-test) to cool down a bit before resuming the "burn in" period?

Quote:

2)  How can you verify linear response with random noise as your source?  Use a sine wave and sweep from 20Hz to 20KHz.  Do it at specific power levels (100W increments) until you get to max power for your amp.  Bink used three specific sinewaves to verify max power.

Well, my thinking behind using pink noise as the source material was that it would allow me to view the 20-20K response of the amplifier in Smaart at the same time. To elaborate, let's say that I used pink noise for my source material and raised the power output to 100W and the transfer function graph was flat. Let us then say that I progressively raised the levels in 100W increments and as the output levels increased, the amplifier was able to maintain linearity down to a certain frequency, where the output level wasn't what it should theoretically be. The graph would then show this. Now, I will grant you that using 20-20K pink noise isn't exactly indicative of real world situations, but then if I were to only test one frequency at a time, the amplifier can devote all of its resources to that frequency, can it not? Perhaps for a subwoofer-dedicated amplifier, 20 to, say, 200Hz pink noise might be more appropriate?

Again, as I said before, I'm not so much interested in exactly how much output voltage was being delivered at a certain frequency for a given input (though I certainly wouldn't mind knowing ), but rather the overall performance (in terms of linearity). If the amplifier maintains, say, 1000W from 50Hz on up, but gently trends down to 900W at 30Hz, then we're not talking about that big of a deal. If, OTOH, it trends down to 600-700W at 30Hz, well...for an amp driving subs, this could be a problem, particularly depending upon the impedence curve of the cabinets!

Quote:

3.  The 1/3 and 1/8 refers to the amp's max output power, not duty cycle.  1/3 max output sinewave power of an amp is equivalent to high compressed audio (music).  1/8 max output sinewave power is equivalent to typical music.

I take it then that "1/8 max output" refers to, shall we say "regular multiplication" (no logarithms involved), so 1/8 max output of a 1000W amp would be 125W, yes?

Quote:

I will assume that your "uglybuckets" are a piece of wire in water?  Please note to keep the load resistor as non-inductive (purely resistive) as possible since this could affect your high frequency measurements!

Check out http://srforums.prosoundweb.com/index.php/m/38971/107/?SQ=a5 8f3885087c270f8949d4625b3f8593 for a description Water-cooled, 13,500W dummy loads Remarkably enough, I didn't draw any stares when I was measuring the hot water elements at Home Depot...though I did get a reaction when I went to check out, as the gentleman at the register said, "You must have some hot water system at your house!" I explained what I was doing and it turned out that he attended the same college I'm (slowly) working on getting my BA in CompSci at and that his roommate hand-built his entire stereo and speaker system, so we got along quite well whilst he was scanning everything and ringing up the total!

I appreciate your help! Cheers!

Charles Johnson

[John Roberts {JR}]

The audiophile reference to burning-in is an example of the too often practice of taking a small anecdotal observation and extrapolating to absurd lengths.

There is a mechanism in semiconductor junctions called self-annealing where noise levels can lower, beta (current gain) increase, and a few other closely related parameters improve, just from operating in circuit.

Apparently some "golden ears" discovered that a certain piece of esoteric gear got quieter after being turned on for a period time. What they failed to realize since this annealing is typically insignificant, that they were probably correcting for the residual effects of another fault mechanism.

For self annealing to be evident, the junction needs to be distressed in the first place. A common way for this to occur is if a semiconductor junction is allowed to reverse zener (6.5-7v backwards across most b-e junctions). What our over enthusiastic audiophiles had was a faulty design that probably trashed the semiconductor junctions during turn-on or turn-off. After running for a while (their burning-in, my self-annealing), they recovered to quiet, normal operation. The design flaw could have been corrected with a penny diode across the junction to clamp it if reversed but they prefer to add this to the audiophile folk lore.

JR  

[Gian P. Portanova]

Yes, I pretty much assumed that it was an attempt to weed out any "infant mortality" units, however, after "Googling" the subject and finding references to "burning in" amplifiers for (up to) 100 hours (as was suggested by one site), I figured I'd ask here. Granted, most of the sites that turned up as results were audiophile pages, but audiophiles have been known to be right about some things from time to time

***JR explained this one for you! 100 hours seems like a waste of your time, electric, and cooling fan life!



Ah - and where might I obtain such an audio file (or the specifications needed to make it)?

***Try  www.comm-2000.com.  I think for your application, standard pink noise would suffice!


By thermal cycling do you mean powering down for a certain length of time to allow the AUT (amplifier-under-test) to cool down a bit before resuming the "burn in" period?

***Yes.



Well, my thinking behind using pink noise as the source material was that it would allow me to view the 20-20K response of the amplifier in Smaart at the same time. To elaborate, let's say that I used pink noise for my source material and raised the power output to 100W and the transfer function graph was flat. Let us then say that I progressively raised the levels in 100W increments and as the output levels increased, the amplifier was able to maintain linearity down to a certain frequency, where the output level wasn't what it should theoretically be. The graph would then show this. Now, I will grant you that using 20-20K pink noise isn't exactly indicative of real world situations, but then if I were to only test one frequency at a time, the amplifier can devote all of its resources to that frequency, can it not? Perhaps for a subwoofer-dedicated amplifier, 20 to, say, 200Hz pink noise might be more appropriate?

Again, as I said before, I'm not so much interested in exactly how much output voltage was being delivered at a certain frequency for a given input (though I certainly wouldn't mind knowing ), but rather the overall performance (in terms of linearity). If the amplifier maintains, say, 1000W from 50Hz on up, but gently trends down to 900W at 30Hz, then we're not talking about that big of a deal. If, OTOH, it trends down to 600-700W at 30Hz, well...for an amp driving subs, this could be a problem, particularly depending upon the impedence curve of the cabinets!

***I'm not aware of any amp that limit the level AND bandwidth simultaneously (unless there is an output transformer).  Low freq performance is dependant on the power supply.  So if your amp can do 1000W at 20Hz, I'd say it can do full bandwitdh power at 1000W.  Just keep in mind that amps are "typically" designed for ~1/3 power of continuous sinewave.  The full power spec is for your audio peaks!



I take it then that "1/8 max output" refers to, shall we say "regular multiplication" (no logarithms involved), so 1/8 max output of a 1000W amp would be 125W, yes?

***Yes!  If now everybody could spec there SR systems with 1/8 to 1/3 power in mind.  Speakers are rated on the RMS/PEAK convention, aren't they?



Check out http://srforums.prosoundweb.com/index.php/m/38971/107/?SQ=a5 8f3885087c270f8949d4625b3f8593 for a description  Water-cooled, 13,500W dummy loads  Remarkably enough, I didn't draw any stares when I was measuring the hot water elements at Home Depot...though I did get a reaction when I went to check out, as the gentleman at the register said, "You must have some hot water system at your house!"  I explained what I was doing and it turned out that he attended the same college I'm (slowly) working on getting my BA in CompSci at and that his roommate hand-built his entire stereo and speaker system, so we got along quite well whilst he was scanning everything and ringing up the total!

I appreciate your help! Cheers!

***If I get chance, I'll check it out!

[Charles Johnson]

JR wrote

I am always apprehensive when consumers try to "confirm" specifications. While an educated consumer is a good consumer there are a lot of subtleties involved in testing, especially at higher power levels.

Putting myself in the shoes of a manufacturer, I can understand your concern, particularly since you have essentially no control over the test process by the consumer, who can then publicize his/her possibly incorrect results. With that said, however, as I have stated, I'm not so much interested in the max output of the amplifier (though I wouldn't mind knowing it ), but rather the linearity as the output level is increased.

To expound upon my stated interest, a quick look at the results from Bink's Amp Shootout (which was done in a pretty scientific fashion, from what I've read) shows that many (most?) of the amps do not meet their 20-20KHz spec (well, at least on the low end of the spectrum). Bink's test points were at 20Hz, 1KHz, and 20KHz, and while they offer a tantalizing glimpse of how the DUT performs, it is still merely a glimpse - though I should note that this is through no fault of Bink's, since reality dictated that, due to the sheer number of amps involved, the test criteria had to be limited. Anyhow, back to the issue at hand - looking at the results for the lab.gruppen fP 6400, we see a 20Hz test result of 1327W, a 1KHz result of 2310W (which met both the 1KHz and 20-20KHz specs), and a 20KHz result of 648W. (Note: I'm taking these results from the spreadsheet that gives a summary of all of the amps tested.) Clearly, the LF (and VLF) output falls off, but my questions are: Where does the decline start? How quickly does the power (er, voltage) output drop off? Let's say that I need to deliver 2000W @ 4 Ohms to a subwoofer. Well, the 20-20K spec on the fP 6400 says that I can, but the sweet spot at 20Hz was only 1327W (and, interestingly enough, raising the input voltage further caused less output). For subwoofer-duty, this could be a problem, n'est-ce pas? However, let's say we tested the amp and discovered that it outputted, oh say, 2000W down to at least 50Hz and then sloped down a bit before plunging down at 30Hz. In light of that data, the answer to the question of the amp's suitability for subwoofer-duty is much different than earlier, wouldn't you say? If, OTOH, the output started to steadily drop off from, say, 2000W @ 150Hz on down to the 1327W @ 20Hz mentioned above, then the answer to the question is a bit different, wouldn't you agree? BTW, for the record, the reason I'm being picky about the linearity of the output wrt the input is that for subs, if the bass player hits a note that just happens to coincide with a point at which the subwoofer's impedence skyrockets from, say, 4 to 30+ Ohms, then you're going to want/need all the power you've got!

Quote:

As you've already been advised the proper tool for measuring frequency response is swept sine wave. This can be unnaturally stressful to the UUT if not careful as few units will encounter full power 20kHz tones in use.

You've piqued my curiosity here - why would 20KHz sine waves present a problem to an amplifier? Speculating idly, I'm guessing that perhaps it could induce some sort of oscillation if a circuit was poorly designed?

Quote:

Good luck and have fun... you will surely finish smarter than you started.

That's the idea

Seriously, while the amplifiers that I am using are performing superbly, my thinking is that if I can break out the test gear and experiment some during down times, I can learn more about what's going on "under the hood" and what questions to ask and what to look for when spec'ing and/or purchasing amplifiers in the future

I appreciate your input! Cheers!

Charles Johnson

PS In re-reading this, I noted that you stated that "there are a lot of subtleties involved in testing, especially at higher power levels." - perhaps you could elaborate on this a bit?

[Michael 'Bink' Knowles]

Charles, maybe the amps such as the fP6400 would have met rated power if my test tones weren't continuous. Perhaps if I tested with short bursts and silence between then the results would have been different for the 20Hz and 20kHz trials. However, I was interested in continuous power ratings and for my purposes with live music in mind, continuous means at least a second or two.  

Sine tones aren't stressful because of potential oscillation. They are stressful because they require the most from the amp.

-Bink

[John Roberts {JR}]

I'm on my way out right now so I can't write you a full response. I invite you to read the many posts generated around the time of Bink's amp testing where we discussed some of these subtleties. I also seem to recall a problem with (all) Bink's 20kHz test results. Also well discussed at the time.

Regarding why would amps have trouble at 20kHz... OK, well designed modern amps don't..... but. Some mosfet output stages with poor driver stage designs didn't turn off as fast as they turned on and would get into a mode called "mutual conduction". The outputs were simultaneously trying to pull up and down... not a good thing in a power amp. I have also seen mutual conduction in a early bipolar attempt from a new entry to the power amp business.  

A second HF issue is that for stability most amplifiers use a RC or LC to ground at the output. These components while tuned for a corner frequency well above 20 kHz will have to dissipate power when the amplifier in banged hard at 20kHz. This is not an issue for normal music because music doesn't do that.. but inexperienced testers can stress designs without realizing. Testing power bandwidth to 50 or 100 kHz may seem interesting but be careful. If you fry the output trap you may break the amp and it wasn't the amp's fault. As it is these parts are routinely over designed for the application just for this reason.

Linearity vs. power output is routinely plotted and often included in better reviews. If you have access to some of the better European magazines they publish quite complete specs. I don't need to read German to appreciate a good set of charts and graphs.

There are too many got'chas to anticipate them all and list them again... this has been covered. I am no longer a power amp manufacturer so I have no horse in this race. That said I believe there are flawed assumptions being made that many amps don't meet specs or perform as well as claimed when in fact they typically equal or better their published specs.

JR

[Gian P. Portanova]

Hi Charles,

In addition to what Bink and JR have responded, you must keep in mind that "audio" power amps are not designed to operate continuously at full power.  Hence, the 1/3 and 1/8 power spec (again).  

You are correct to assume (I hope) that the amp will do full power for short bursts.  This is how amps are typically designed.  In fact, it may do more than rated power for short bursts.

If you do a 1W continuous sinewave sweep (10Hz to 100KHz), you will find out what the amp is designed to do for bandwidth.  YOU will probably be limited in your test equipment and will not be able to measure this wide bandwidth.  So, if the amp if flat from say 20Hz to 20KHz @ 1W, then you know the amplifier section should be capable of the bandwidth.  When you measure full power, now the power supply comes into play.  For typical linear supplies, low freqs will depend on transformer regulation and storage capacitance.  For switchers, it will depend on the current (amperage) capability of the switching parts (FETs?).  Under heavy load, a linear supply transformer will sag while a FET will just fail/damage.  Therefore switchers need additional protection circuitry (in the same manner that the amplifier section needs protection).  With this in mind, you can probably guess why the FP6400 (switching supply) could not hold up for continuous 20Hz.

If you want to verify linear performance, why not use tone bursts (say 10 - 20 points between 20Hz to 20KHz), measure the volt peaks with a scope, convert to power, and extrapolate your response curve!


Good luck!

[ProSoundWeb Community]