Very high power audio amplifiers

[Mark Hennessy]

Hi Craig,

Apologies for missing this message - the forum software isn't very good at alerting the admins/mods to queued messages. Anyway, future messages should appear straight away now.

You're absolutely right about "Watts RMS". Ultimately, it's a marketing term, which implies nothing more than RMS voltage/current measurements. I suspect that most designers are similarly frustrated with this - certainly, the ones I know are. However, "Watts RMS" is somewhat more snappy than "Continuous Average Sine Wave Power"

Having said that, sine waves are totally unrepresentative of real audio signals. Just compare the PMR of sine waves (3dB) to typical music (20dB or more). In power terms, 20dB is obviously a factor of 100, so a Quad 405 running on the verge of clipping might only be delivering 1 watt on average to the load. Realising this leads you to see that peak power is actually more relevant than continuous average sine wave power - and an amplifier designed to play all day at its rated power is actually seriously over-engineered for typical programme material. And clipping on peaks is a lot more common than people realise...

Obviously, that changes for subwoofer amplifiers and heavily compressed scenarios.

At least "Watts RMS" is a known, repeatable measurement, along with various others like IHF dynamic power. Unlike, however, PMPO, which is totally arbitrary. About 20 years back, we took delivery of some PC speakers with a 500W PMPO power rating - all from a wall-wart containing a 6VA mains transformer

Power ratings for loudspeakers are even more amusing, and have to make all manner of assumptions about the programme material. Ideally, manufacturers should provide a plot of power vs frequency, which is something you approximate at the design stage anyway. Obviously, at low frequencies mechanical limits are problem, but things get thermal as you go up in frequency (unless the smaller drivers are crossed over too low, in which case they can struggle mechanically as well). But given the amount of equalisation that takes place in a crossover - in addition to the basic filter function - the overall picture is pretty complex. Ultimately, you just give up and quote a suitably "round" number that'll look promising on the spec sheet

The Wikipedia article is quite reasonable: https://en.wikipedia.org/wiki/Audio_power

All the best,

Mark

[Craig]

Hi Mark

The problem is that extreme music compression is sadly becoming the norm, very well summed up here https://www.youtube.com/watch?v=u9Fb3rWNWDA , the generic term for which is the "Loudness war". The exceptionally small dynamic range, with pegged VU meters and "managed distortion" means that the peak to average power is now very small indeed.

Loudspeaker power handling is another "interesting" topic. Particularly since good driver manufacturers (such as SEAS) publish rather comprehensive power handling data. That seems to go out the window once a manufacturer puts these drivers in a box. I suspect that is to prevent users blowing up an expensive speaker, and then arguing that they used an amplifier that met the quoted power handing.

The most representative signal to use for power handling, and a mimic of "real" music is a pink noise signal in which one can adjust the crest factor. What seems an eternity ago now I was CTO of Wharfedale, during which time we acquired FANE. The survival test for a FANE driver was to mount it in a concrete bunker and feed it with pink noise at the rated power level (which could be 1kW peak and 500W average) for 24 hours, after which it not only had to survive, but also have unchanged acoustic response. These were used at rock concerts and had to survive all kinds of abuse.

I'm currently listening to these http://www.linkwitzlab.com/LX521/Description.htm actively crossed over and EQ'd dipoles. I'm using Douglas Self Blameless power amps, 2x100W and 2x50W per channel. My build is the ninth one down here http://www.linkwitzlab.com/LX521/PhotoGallery.htm.

Thinking of Self, at the end of his book on Power Amplifiers he considers power transistor dissipation for real world loudspeaker loads, which makes for interesting reading. It needs a serious rethink about what an amp actually needs regarding thermal management when driving a loudspeaker.

Craig

[Mark Hennessy]

Hi Craig,

Yes, I've been boring everyone with the Loudness War for the best part of a decade now. I find it ironic that the biggest group of people to bring it to public attention were Metallica fans - not exactly a stereotypical "audiophile" audience https://www.youtube.com/watch?v=DRyIACDCc1I

Of course, it's existed in many forms for decades. E.g. Phil Spector's "Wall of Sound" production, which helped tracks stand out in fixed-gain jukeboxes. Radio is another classic example. But many commercially released CDs have got especially bad over the last 15-20 years. Bob Orban is on record saying that if you want the best results from his Optimods, you need to feed them material with a "natural" dynamic range - hyper-compressed stuff will just sound awful (and he's right). IIRC, that quote come from Bob Katz's book "Mastering Audio" - an excellent read

There is an end in sight. The loudness war came about because we've historically been in an environment where peak level was king (the need to not over-modulate transmitters, analogue tape, etc), and where the overwhelming type of monitoring used in broadcast was the PPM, not the VU. Naturally, CD has 0dBFS as a notional maximum level (though it's possible to slightly exceed that, of course). In this environment, compression was used to make your mix stand out while not exceeding the maximum allowable peaks - hence complaints about adverts being louder, etc.

Today, we are moving towards an environment where we equalise gain according to loudness. Loudness has always been tricky to measure accurately, but standards today do a reasonable job. Broadcasters are using R128. But what's really made the difference, however, is mainstream services like iTunes and YouTube (and no-doubt many others) also adopting this approach.

I run a series of demonstrations* to show why this works well. In short, a hyper-compressed mix might sound impressive initially, but if you bring the gain down, it just sounds flat and boring. If you play a better track at the same loudness, it sounds more impressive. In a loudness-equalised world, you need to make your mix more dynamic to make it stand out.

* The main tracks I use for this are "Money for nothing" (original version, not the re-master), and a horrible bit of Red Hot Chilli Peppers (Scar Tissue). The former has a PMR of ~30dB, and the latter is less than 10dB (assessed very informally from the level meters in Audacity, so probably not accurate).

Back to hi-fi amplifiers, yes, I did make the point that hyper-compression changes the numbers. However, I think it can be safely ignored when thinking about hi-fi. Almost no-one I know who takes hi-fi seriously listens to that stuff, and those who do listen to it quietly because it doesn't sound better when turn up the volume, and you certainly wouldn't want to be in the room when it's on the verge of clipping (whereas - for the sake of context - "Money for Nothing" would sound pretty reasonable on the same setup at the same "verge of clipping" gain setting).

The problem comes about with the classic "teenage party". As there are compelling reasons for manufacturers to skimp on heat sinks and mains transformers on an amplifier expected to be used for hi-fi (not PA or studio use), then this is a problem. The solution is simple: thermal cut-outs on the heat sinks, and/or cooling fans. I always include the former in my designs, avoiding fans if I possibly can.

By coincidence, I'm currently working through chapter 16 of the latest edition of Self's power amp book. I have all the editions at home, but I think that topic has been expanded for the latest edition (I'd have to check previous ones). The point he makes about PDF is interesting, and is probably unique research. I've been successfully following his principles for 20+ years (though must admit that the humble LM3886 is what I use for anything less than 60 watts these days, providing that's acceptable to the Marketing department! Superb little chips, providing they're not clipping).

Yes, I'm a fan of SEAS drivers, and have used them a fair bit over the years. They're very good at OEM stuff, and very consistent from batch to batch. I've got mixed feelings about the open arrangement they have with the spider - good for airflow and noise, but I worry about ferrous debris finding its way to the gap. Luckily, the BBC-inspired loudspeakers I use are intended to be used with the grilles on (plus I have young kids, so I've not removed grilles since they came along!). Congratulations on the Linkwitz project - they look absolutely amazing . I've followed Linkwitz for years, but have never got around to trying any of his designs - in some ways they are rather at odds with the "BBC" school of loudspeaker design that I'm so familiar with, but I'd love to hear some examples one day.

When did Wharfedale buy FANE - late '80s? Afraid I've only seen the '60s "transistor radio" range of FANE loudspeakers, which were pretty good in that context IME. I know virtually nothing about musical instrument 'speakers - another game entirely! FANE seem to be doing OK today, happily. While the mainstream has left these shores, there's actually quite a chunk of good stuff happening in the UK, with folk like ATC and Volt and Precision Devices making high-end drivers...

I think that most hi-fi loudspeaker manufacturers are careful when wording their warranties to specifically cover "misuse". If an overheated voicecoil is the problem, then that's misuse, whatever the amp. When quoting power handling, phrases like "unclipped programme" are good, as that gets away from sine wave power. But it's a minefield. A speaker with a properly set up crossover with full BSC should be able to absorb 6dB more at 1kHz than at 100Hz (ignoring the mechanical limits), but when the tweeter starts to come in at 2-3kHz... I usually say "it's interesting that power ratings are such nice, round numbers"! But away from hi-fi, things become more difficult, and certainly for PA use, you have to be rather more confident in your ratings as these things are expected to work continuously at those powers for many hours at a time - and to cope with "accidents" that happen on a stage.

It's a fascinating subject

[Craig]

I know Money for Nothing well - in fact I usually use the final and Title track. I have the original vinyl, the CD version (ripped to NAS, as well as the CD), and the streamed version via Tidal HiFi. It is an interesting comparison. The vinyl and early mastered CD are very close, but the later remixes are compressed, even though Tidal sends via uncompressed format - so the problem is in the master that they have used. For the worst in compressed music (and it is a real shame, because she is a fine vocalist) is Christine and the Queens - it is just maxed out with almost no dynamic range at all, and is essentially unlistenable to.

Regarding Self, it is only the 6th Edition that contains his treatment of device power dissipation. He invents Power Partition Diagrams, which are interesting in their own right - but the impact of reactive loads is of major interest. With a 45 degree phase angle (typical of real loads) the power transistor dissipation increases by 50%, and at 90 degrees (which is somewhat extreme) the device dissipation triples.

Linkwitz's LX521 is not for the faint hearted. If you don't have good woodwork skills and machine tools you can figure on about £5k by the time you have the woodwork, the drivers, crossover and a whole bunch of power amps. A bit of an expensive act of faith! Good job they sound good. His LX-mini however does pretty much the same job sound quality wise, but with not as much bass output, for a fraction of the price. The recommended way is to use chip amps and a MiniDSP crossover for that - but you can be more adventurous is you wish.

Yes - Fane was around 1990. It was part of a group that had a stock market listing, and Wharfedale reversed into it to get itself the listing . Monster loudspeakers manufactured in the worst factory I have ever seen. Voice coils were wet wound with epoxy, by women wearing protective gloves, and the winding lathes covered in semi-gelled epoxy, and the coils then baked in a bank of domestic ovens in a room lagged with asbestos. It was awful in every way that you could imagine.

[billh]

We had a monster as an applications lab item at Ferranti in the late 60's. The output was too high to risk using speakers so it was run into electric fire elements - it boiled kettles.

The output transistors got hot too, we used to spit on them to see if they were up to temperature.

Can't remember too much about it but I know the power supply was off 3 phase mains @440v.

New member here.
Ferranti's Gem Mill had 3 or 4 valve amps in the cellar powering vibration tables for testing semiconductors. They were driven by audio oscillators varying in frequency. The tables had effectively speaker motors mounted underneath. The amps were not Ferranti's build they were called "Kilowatt Star" and were 6ft rack mounted . I managed to buy one of these amps from the scrap heap- it cost me 50p (the standard price for any item found there) and got it home in my Hillman Imp van, the amp weighed about 1/4 of a ton. Some details of the amp , I remember - the HT transformer  was 3000v at 700mA, the rectifiers were mercury vapour and OP valve drivers were EL37. I never dared to power it up, it was too scary! Eventually, I sold it for a fiver, so a good profit margin. I should have kept the EL37s though.
Around 1970, Ferranti published a book " Transistor HI FI amplifiers " , designs were developed  in Gem Mill application lab and used transistors in the ZT range, e.g. ZT3055. When I started at Gem, there was still some valve production on the ground floor- the equipment for this looked Victorian ( it wasn't!) what with gas jets and vacuum pumps , also there was a lot of Russian 807s tested for some contract, many of these were rejected , they were very poorly made.
While all this ancient valve stuff was going on , we upstairs were assembling huge quantities of very modern (then) TTL and later on , microprocessors,ULAs   and ZN414 radio chips. 
Bill