I've already touched on O/L in my ravings about stability. On complex load drive, I've had long discussions with Peter Walker on 'what is a loudspeaker' I want a '8R' amp to drive 6R4 at up to 60 degrees phase angle. This is difficult for BJT O/Ps but easy for MOSFETs. Simple current limiting 'sounds better' than fancy VI SOA stuff.
I don't want my amps to Slew rate limit but in fact if the amp will give full power at 40kHz, any slew limiting after that is 'inaudible'. A 20kHz band limited signal has a maximum slew corresponding to 40kHz full power (on a 20kHz full power square wave) Anything more than that is really an O/L situation and recovery is more important.
If the output slew limits on O/L, it gives LESS high order harmonics than if zillion V/us is available. I've some DBLT results on an amp with deliberate slew limit.
Soft clippers, even very crude '1 resistor and 2 diode" do sound better in DBLTs but I'm not sure how to quantify this. I need to look at Bob Cordell's Klever Klipper more closely. 🙂
I don't want my amps to Slew rate limit but in fact if the amp will give full power at 40kHz, any slew limiting after that is 'inaudible'. A 20kHz band limited signal has a maximum slew corresponding to 40kHz full power (on a 20kHz full power square wave) Anything more than that is really an O/L situation and recovery is more important.
If the output slew limits on O/L, it gives LESS high order harmonics than if zillion V/us is available. I've some DBLT results on an amp with deliberate slew limit.
Soft clippers, even very crude '1 resistor and 2 diode" do sound better in DBLTs but I'm not sure how to quantify this. I need to look at Bob Cordell's Klever Klipper more closely. 🙂
When an amp clips, the PSU ripple 'modulates' the signal. This is very audible on small Class A amps cos they have really yukky ripple on conventional PS and they overload often. Powering eg a JLH 10W Class A, with a bunch of laptop PSUs does result in better sound.
I apologise for not actually coming up with 'numbers' to add to Bonsai's list except for the '<0.01% under ALL conditions' and stability testing with 1nF & greater.
I've tried to simply report the results of DBLTs and have stated my prejudices when I stray from this.
I've tried to simply report the results of DBLTs and have stated my prejudices when I stray from this.
Amplifiers (class A or otherwise) can be made not to do that. And the power supply can be made to have low ripple too. Cheaping out on power supply caps (and the transformer) doesn’t help the situation. It can make the difference between an amp totally going to crap with the first hint of a clip light to one that still sounds clean driven 6dB into clip - repetitively.
Wanna explain how to do that on a small Class A amp? Bigger caps & better transformer will give a 'lower' measured ripple but it is also spikier. I have my own ideas on how to do that eg my 200W/8R amp & 90dB/W @ 1m speakers ... some of which I've DBLTested.
Interesting comments, some of them, but hard data to support the claims would be needed. Otherwise we are in the field of anecdotal stories, similarly as in the times of Blowtorch threads.
That's not my experience at all. About 30 years ago, when I was still young and still occasionally played reggae at volumes high enough to disturb my neighbour (unintendedly, of course), I used an oscilloscope to check what peak power I needed. It turned out to be about 2 W into 8 ohm per channel using 90 dB at 1 W, 1 m dynamic 8 ohm loudspeakers. As I also liked classical music, I also tried a pretty good recording of the Bolero.
Taking 10 dB of headroom, I built myself a two times 20 W amplifier with a clipping LED: clip it for a couple of microseconds and the LED turns on for about 1 second. It never does during normal use. I've only seen it light up when I put the volume much higher than normal to check that it works, and one time with a home-made nature recording with very soft bird sounds and lots of subsonics from wind blowing into the microphone.
Pano's test also shows that the majority of people never use more than 30 W per channel, see https://www.diyaudio.com/community/...h-voltage-power-do-your-speakers-need.204857/ You need to read the first couple of posts to understand what the measured voltage means.
Depends on what you record and at what volume you play it back. I've also played home-made recordings of my partner's then chorus over my two times 20 W amplifier and had no problems with clipping, not even during the applause. I used ESL-63 loudspeakers back then.
I measure peak power levels of 100mW with 83dB/W speakers. The amplifier is essentially idle. 🙂
Ed
Ed
These DBLTs were done in da previous Millenium. You can decide for yourself if da old beachbum is senile or lying 😳
The Blowtorch thread is different. There da old beach bum is issuing a challenge to da pseudo Messiah. As JC didn't take him up on it, we probably won't ever get hard data 😎
I actually meant the sine wave power corresponding to the peak voltages I measured. The peak momentary power is of course twice the sine wave power, as sine waves have peak voltages of √2 times their RMS value.
I'm very aware of Pano's neat test. I suppose I listen at higher levels than most and beach bums have few neighbours 🙂
In da digital age, with precise FS levels, we know exactly what may come out of the speaker terminals on commercial recordings.
I used to have a list of vinyl with large dynamic range. Bob Cordell had a list of CDs in this century. Alas, one of his examples, a Ricky Lee Jones, track has been re-mastered and recent issues no longer show the large dynamic range.
In general, 'pop' including reggae stuff, may be loud but doesn't have much dynamic range ie loud bits & soft bits. There are more classical recordings with loud & soft bits. There was a Stravinsky Firebird under ?? and also the War Requiem both the old Decca under Britten and also the EMI under Rattle. I don't think any version of Bolero makes this list.
But if you are making your own recordings, there are certain things which are almost impossible to record & play without overloading the microphone, recorder & amplifier.
My most head banging example is "The Garage Door" by Mike Skeet using my microphone and a Sony PCM-F1 . To come anyway near 'realistic' levels (it's a demo of the dynamic range of real life and stereo), you will overload a 1000W/channel amp and 90dB/W @ 1m speakers. The speakers have to behave well under such abuse bla bla too of course and need VERY extended LF response and power.
Less head banging and more common (??) would be drum rim shots or clap sticks which will also overload mike, recorder, & amp at quite modest levels. But as I said, if the mike, recorder & amp are 'well behaved', the overload shouldn't be 'audible'.
On The Garage Door, 200W/channel & 90dB/W @ 1m has overload IMHO at the threshold of audibiltiy ... but 1000W/channel will still overload.
A more difficult example is a VERY GOOD unaccompanied choir of about 12 - 20 singers. It is possible to have levels adding 6dB instead of 3dB if the singers sing sufficiently 'in tune' and your peak meters go wild even at quite modest average levels. The peak levels are high enough to cause Intermod inside your ears; a sort of 'crackling'. Here, the effects of overload are very audible.
I'm sorry if these ravings don't point towards simple 'numbers' or tests for Bonsai's list. 🙁
Please let me return to this - it seems not be only a question of servo loop recovery, but, as mentioned, a combination of LF time constants of the complete circuit. I have just tested another design, which is DC coupled and has a DC servo as well. The results are without traces of LF recovery.
First, IEC60286-3 recovery test definition:
and the results with 20Hz and 1kHz bursts
Does weighing belong in an objective measure? Thought I read that but cant find it now - good.
Cost - I dont think it should be one of the describing quality factors. Articles using the AFOM could choose to do such a comparison / calculation. And plot trend per year and cost, and.... after all, you said "...assessment of overall amplifier quality" which cost isn't.
//
Cost - I dont think it should be one of the describing quality factors. Articles using the AFOM could choose to do such a comparison / calculation. And plot trend per year and cost, and.... after all, you said "...assessment of overall amplifier quality" which cost isn't.
//
TNT, interesting point. I would hope that by selecting the right measures, the weighting would automatically come out of that. Your point about tracking long term amplifier trends is s good one. AFAIK, we can only assess these trends from anecdotal evidence or going and measuring old equipment, which due to the ravages of time, may not be performing at their best.
I do think amplifiers in general have gotten better in some aspects as knowledge has improved, but there remain areas for improvement. There is far too much focus on distortion in the objectivist camp and not enough in the subjectivist side for example.
Re the cost aspect. How would one address claims that a 10 W amplifier with 3-5% distortion costing $19000 is a ‘good amp’ or a 200W monobloc at $55 000 each when a Benchmark ABH2 at a fraction of that cost gets rave subjective reviews and puts in a stirling set of measurements? The idea of the cost per watt spec is simply to make sure a reasonably informed purchaser sees what they are paying for over and above the cost per watt to produce a decently engineered product. Interestingly, if you read the comments in Stereophile when some of these outrageously expensive amps are reviewed, they are often critical of the cost, and commentators point to more reasonably priced amps that received better subjective reviews and better measurements. If you are paying more than $30/watt, AFOM is simply proposing you are paying a premium beyond that required to produce a well engineered product.
🙂
I do think amplifiers in general have gotten better in some aspects as knowledge has improved, but there remain areas for improvement. There is far too much focus on distortion in the objectivist camp and not enough in the subjectivist side for example.
Re the cost aspect. How would one address claims that a 10 W amplifier with 3-5% distortion costing $19000 is a ‘good amp’ or a 200W monobloc at $55 000 each when a Benchmark ABH2 at a fraction of that cost gets rave subjective reviews and puts in a stirling set of measurements? The idea of the cost per watt spec is simply to make sure a reasonably informed purchaser sees what they are paying for over and above the cost per watt to produce a decently engineered product. Interestingly, if you read the comments in Stereophile when some of these outrageously expensive amps are reviewed, they are often critical of the cost, and commentators point to more reasonably priced amps that received better subjective reviews and better measurements. If you are paying more than $30/watt, AFOM is simply proposing you are paying a premium beyond that required to produce a well engineered product.
🙂
Last edited:
@kgrlee By 'small' I assume you mean "push pull". It seems the closer you get to class b, the worse the PSU load gets: negative impedance modulated by a nasty rectified audio signal. The lightness of the load has to be balanced against the grunge injected onto (and reflected from) the rails. This is a whole area of design I didn't really want to get into, but it seems unavoidable.
For reasons of Feng Shui at least, bulky transformers and capacitor banks would be a last resort for me. Yet at the same time, buffering the "audio-modulated linear regulator" with another linear reg to filter the noise from dirty OTS switching supplies seems dumb. So a proper option would be low noise SMPS design... It doesn't even have to be well regulated, just low noise.
An interesting twist could be feeding the audio signal forward to stiffen or lift the rails if needed -- an unholy fusion of some different classes. There are so many opportunities where people already have A/D + D/As, where the MCU in-between can tell the power supply to ramp up to prevent clipping in advance, or ramp down to save power. As a 'lite' alternative, the amplifier could be fully analogue and still use a processor to control the rails based on the averae audio level. A bit better than the glitchy auto-mute some amplifiers have.
Ultimately, unless the system runs on batteries, you're stuck converting mains power into a DC-ish voltage, which then powers a super-clean audio-modulated regulator. I could be a bit miffed if I'd developed this artistic/engineering masterpiece, and some reviewer decided to input 4 hours of laboratory test signals at "max power", concluding: "gets red hot and shuts down mid-test. 2 stars."
For reasons of Feng Shui at least, bulky transformers and capacitor banks would be a last resort for me. Yet at the same time, buffering the "audio-modulated linear regulator" with another linear reg to filter the noise from dirty OTS switching supplies seems dumb. So a proper option would be low noise SMPS design... It doesn't even have to be well regulated, just low noise.
An interesting twist could be feeding the audio signal forward to stiffen or lift the rails if needed -- an unholy fusion of some different classes. There are so many opportunities where people already have A/D + D/As, where the MCU in-between can tell the power supply to ramp up to prevent clipping in advance, or ramp down to save power. As a 'lite' alternative, the amplifier could be fully analogue and still use a processor to control the rails based on the averae audio level. A bit better than the glitchy auto-mute some amplifiers have.
Ultimately, unless the system runs on batteries, you're stuck converting mains power into a DC-ish voltage, which then powers a super-clean audio-modulated regulator. I could be a bit miffed if I'd developed this artistic/engineering masterpiece, and some reviewer decided to input 4 hours of laboratory test signals at "max power", concluding: "gets red hot and shuts down mid-test. 2 stars."
‘There was a Stravinsky Firebird’
It’s Claudio Abaddo’s DG LSO recording. I have it on vinyl. Fantastic but the recording does clip. Incredible dynamic range.
As general comment, speaker sensitivity, room size/acoustics and amplifier output have to be considered as a ‘system’. A 15W amplifier should be used with sensitive speakers if one expects realistic sound levels, or, the listening space must be small enough.
This is where component matching is important, but I don’t think the AFOM proposals we are discussing here should stray into that area as it is really very subjective.
This is where component matching is important, but I don’t think the AFOM proposals we are discussing here should stray into that area as it is really very subjective.