MEASURING amplifier clipping

[Hiten]

How about diy clipping indicator like this ? very few parts and can be taken out after measuring.

 

[solderdude]

Such a circuit would be preferable but it can only detect how close the output comes near the voltage rails.
It cannot detect clipping caused by current limiters.
One would also need input from the current limiters if one really is to monitor the output (the External input on the circuit)

It should always be in circuit b.t.w. if you want to have a clipping indicator in that specific amp.

 

[Hiten]

Such a circuit would be preferable but it can only detect how close the output comes near the voltage rails.
It cannot detect clipping caused by current limiters.
One would also need input from the current limiters if one really is to monitor the output (the External input on the circuit)

It should always be in circuit b.t.w. if you want to have a clipping indicator in that specific amp.

ahh. OK.
my friend had a nad 3020 and it had led to detect. Nice feature to have. have not seen any current amplifier having even basic indication. Some Nad power amps had only one led. (Pre out, Main in, clipping indicator all basic features in vintage amplifiers)
regards.

 

[solderdude]

I once made 2 of these things.
Rechargeable battery fed, connect to a speaker and when the set level was reached it flashed (regardless how short the pulse was)
The double side taped paper covered with self adhesive covering did not gracefully survive 40 years on the attic

 

[mhardy6647]

^^^ outstanding.

ahem.
On topic: this thread is reminding me of the (in?)famous Dick Burwen (by way of Bob Carver) "scissor snip" scenario.

source: https://worldradiohistory.com/Archive-All-Audio/Archive-Audio/80s/Audio-1985-10.pdf

As all y'all likely know, Dick Burwen (who, last I knew, was still on this side of the turf and who is, by all accounts, a pretty down-to-earth & approachable guy) practiced what he preached, vis-a-vis having adequate power reserves to accommodate transients & to avoid instantaneous clipping.

http://www.burwenaudio.com/images/Audio_April_1995.pdf

Dr. Burwen (or Mr. Burwen, as the case may be) gets bonus points, at least from me, for that National HRO communications receiver in his stack

see also:
https://worldradiohistory.com/Archive-All-Audio/Archive-Audio/90s/Audio-1995-04.pdf pg. 28
Wherein his 20 kW hifi updates to "digital EQ"

 

[Head_Unit]

minimum LED duration

This is a factor-another way to check for clipping would be get some kind of level meter, and set the "0 dB" or whatever to the calculated voltage rail. BUT without knowing there is some persistence, it's probably just eye candy.

 

[Head_Unit]

the goal of Head_Unit's proposal is obviously not to determine what an amplifier is capable of, but to help people become aware of the fact that his amplifier might sometime enter into clipping

Exactly. If someone can determine that the loudest they play is 20W peaks on a "135W" AVR, no need to even think about an external amp. If they determine they are hitting 120W peaks, well then *maybe* they are clipping and need to investigate further or just buy more amplification. So the idea is more a "ruling out" clipping than an exact determination.

 

[Zapper]

"Music is full of surprises such as explosive crescendos, combinant crests, blah blah."

Not with digital audio. You know precisely how loud the music can be.

This kind of blather has been used for decades to sell high power amplifiers, and it's a lie. It's always been a lie, but the hard amplitude limit of digital audio makes it clear. The recording signal path has never been unbounded. Mics, pre-amplifiers and signal conditioners, and magnetic recording media all had amplitude limits, some of them hard (pre-amp clipping), some of them soft but insurmountable nonetheless (magnetic tape saturation). When LPs where the primary hifi distribution medium there were fairly strict limits imposed during the mastering on groove modulation amplitude and velocity, else ordinary styli would not track. And dynamic compression, both through deliberate signal processing and passive effects such as magnetic or vacuum tube nonlinearity, has long been used to compress the natural dynamic range of music to make it work in the recording medium, be palatable to the listener in the home or car, and to surmount the noise floor in the recording and playback chain.

As a result of these factors and more, the usable dynamic range of LPs was much less than CDs. It's ironic that digital music, with its potentially expanded dynamic range, also mathematically eliminated the myth of surprise, unbounded amplitude peaks in music.

It's also ironic that in the age of 24/192 lossless digital, most popular music is compressed all to hell and might as well be played in 12 bits.

 

[Piere]

*This is the hard part. The Stereophile 1st CD track 27 might work https://www.discogs.com/release/3692778-Various-Stereophile-Test-CD2
If someone could generate and upload suitable test tone(s) you would be the thread HERO!!
300 or 400 Hz at -20 dB is probably good since that will go into the woofers but without huge excursion, and also I dimly recall most voltmeters should be OK at those frequencies even if they are not good across the whole audio band.

To good old https://www.goldwave.com/ will do for you. It has a build in equation editor that lets you make almost any mathematically described .wav

_________________________

Point is not only when an amp is clipping but how it recovers from clipping. It should do that quick and without storage artefacts. If that is a clean short clipping, it will not impair the sound too much if not too severe. Here a real example of an otherwise fine sounding amp. But don't let it clip!

 

[DonH56]

The way to do a simple circuit to detect clipping without characterizing the amp: make a differential amplifier with adjustable attenuation on inverting input. Split the power amp input signal and apply to non-inverting input. Apply power amp output to attenuator on inverting input. Play the power amp into speaker and adjust attenuator for minimum ac output of diff amp. The attenuated power amp output should almost match the input giving a good null. Now the diff amp will give a large signal when the power amp clips and its output no longer tracks the input signal accurately.

I missed your post... I actually tried that but dropped the idea for reasons that now (40 years later) I do not remember. I think partly because the little clipping box with leads to the output was easier to attach than having to get a splitter to monitor the input, and I had to account for phase shift and delay through the amp with appropriate band limiting and integration. But a proper nulling circuit is a great idea.

Bandwidth/detection also an issue with my circuit above; to see clipping, I had to add an integration cap so the LED would stay on long enough to see unless you were constantly clipping (which often happened). Basically made an envelope detector. I also implemented a peak detector using the old National op-amp application book's circuit and I think that was the final version; it would detect a peak then slowly decay

 

[Piere]

Bandwidth/detection also an issue with my circuit above; to see clipping, I had to add an integration cap so the LED would stay on long enough to see unless you were constantly clipping (which often happened). Basically made an envelope detector. I also implemented a peak detector using the old National op-amp application book's circuit and I think that was the final version; it would detect a peak then slowly decay

Long time ago I build a clipping detector as follows: A comparator simply drove a retriggerable c-mos one shot of 0.1s. Worked pretty well, also for very short pulses.

 

[DVDdoug]

You can make something "similar" to an oscilloscope by putting a voltage divider (two resistors) or a resistor and a pot between the speaker-output and your soundcard input. Then record it with Audacity (FREE). You can zoom-in to see the waveform details and you can probably see if it's clipping and you you can turn up volume to see if you have any headroom.

Be careful with test tones... You can fry your speakers with continuous test tones... A 100W speaker is designed to handle program material hitting 100W on the peaks with a much lower average average. And the tweeter can't handle as much power as the woofer.

Audacity can also generate tones at your chosen frequency.

 

[Head_Unit]

Here you go:

WeTransfer | Send Large Files Fast - Up To 2GB Free

The simple, quick and secure way to send your files around the world without an account. Share your files, photos, and videos today for free.

we.tl

WeTransfer | Send Large Files Fast - Up To 2GB Free

The simple, quick and secure way to send your files around the world without an account. Share your files, photos, and videos today for free.

we.tl

"Sorry, this transfer has expired and is not available any more"

 

[Head_Unit]

Point is not only when an amp is clipping but...how it recovers from clipping.

Excellent point. I recall some AES paper (Matti Otala?) showing how some amplifiers had really ugly behavior upon clipping. One would snap momentarily to DC of the opposite rail voltage!

 

[DonH56]

Long time ago I build a clipping detector as follows: A comparator simply drove a retriggerable c-mos one shot of 0.1s. Worked pretty well, also for very short pulses.

Yup, another good idea. Back then I used a TTL one-shot but same idea. Probably a better way than the more complex integrator, but my circuit provided a bit of intensity modulation of the LED (brighter when clipping was more frequent) which was sort of cool. Not always a good thing as if I wasn't careful you could still miss a fast clip.

 

[DonH56]

Excellent point. I recall some AES paper (Matti Otala?) showing how some amplifiers had really ugly behavior upon clipping. One would snap momentarily to DC of the opposite rail voltage!

Some early op-amps would invert when overdriven. Some power amps had similar behavior but it was resolved pretty quickly (either by fixing the amp or going out of business ). Overload behavior is why many of us advocate for such testing. The (a) problem for Amir @amirm is he really does not want to blow up an amplifier he doesn't own. And probably not one he does... Destructive testing gets expensive!

These days, at least the past decade or two I have looked, the problem seems to be more oscillatory behavior when heavily overdriven, and not always way up in the supersonic region.

 

[Sokel]

"Sorry, this transfer has expired and is not available any more"

Here's some fresh ones then:

WeTransfer | Send Large Files Fast - Up To 2GB Free

The simple, quick and secure way to send your files around the world without an account. Share your files, photos, and videos today for free.

we.tl

 

[Piere]

Be careful with test tones... You can fry your speakers with continuous test tones... A 100W speaker is designed to handle program material hitting 100W on the peaks with a much lower average average. And the tweeter can't handle as much power as the woofer.

Good point. Especially tweeters cannot handle more than around 5 watts average continuously! And hearing of many of us has degraded just in the higher regions! Most domestic amps cannot spit out max. power continuously, even in a dummy load. They will fry also then. They are designed for program material that contains much less average power.

Best way to test for clipping is to use sine burst with a 1:10 duty cycle, or better 1:100 duty cycle: 10 cycles on - 90 or longer cycles off. A 1:10 burst at low frequency is also a good measure how solid the PSU performs!

 

[Scytales]

View attachment 274703

"Music is full of surprises such as explosive crescendos, combinant crests, blah blah."

Not with digital audio. You know precisely how loud the music can be.

This kind of blather has been used for decades to sell high power amplifiers, and it's a lie. It's always been a lie, but the hard amplitude limit of digital audio makes it clear.

I think you are mislead. While digital signals do indeed have clearly defined dynamic range boundaries, you fail to recognize that the upper limit of this dynamic range (the 0 dBFS or +3.1 dB SA-CD for that matter) could very well be aligned on a very high analogue level (be it electronic or acoustic) when the gain of the amplification chain and the sensitivity of the loudspeaker have been factored in.

Conversely, you could very well align the upper digital level to a very very low analogue output level at the end of the chain. In this case, all you would hear is silence or muted sounds even with the digital level at maximum modulation.

The fact is that hi-fi means reproducing sound at the correct level. The correct level is the natural level of a performance.

In the end of the 80s, the French manufacturer Cabasse sponsored a recording of a pianist who played Chopin's and Schumann's works on a Bösendorfer Imperial grand piano. The microphone type and set up were disclosed, as well as the condition needed to reproduced this recording in hi-fi. With the 2 microphones placed roughly at 2/2.5 m from the geometric center of the piano, the reproduction of the forte of this very recording with 93 dB/1m/1W sensitivity loudspeakers at their natural levels (those of the piano in the room) required at least 300 W of peak power, the manufacturer said.

 

[Head_Unit]

Be careful with test tones... You can fry your speakers with continuous test tones

Best way to test for clipping is to use sine burst with a 1:10 duty cycle, or better 1:100 duty cycle:

Yes, though the duty cycle would mean you couldn't use a voltmeter easily-I got one with "peak hold" then started wondering how good the peak hold actually was. Hence the qualification of using like 300-400 Hz waves, which should avoid such accidents as the power would go to the woofers, and the -20 dB scenario which is easy for the math. I suppose even -20 dB would be pretty loud, though any competent woofer should handle that long enough to make a measurement.