Class D (or similar switching amplifiers) is probably the only architecture with a direct link between PSU ripple fraction and audio output ripple fraction. For everything else there is not such a link; the most we can say is that output ripple is probably proportional to PSU ripple but the ratio is unlikely to be 1.simon7000 said:
Here is a 'proof by construction' (hinted at in my previous post).
1) Assume an amp with perfect PSRR in the output stage, but very poor PSRR in the input stage. Assume 1% ripple on a 50V supply, so the input stage sees 0.5V on its supply which it passes straight through to the next stage. Assume a gain of 50 from here to the output, so 25V of ripple at the output. Peak output is 50V so we have generated 50% ripple at the output from 1% on the PSU.
2) Now assume an amp with perfect PSRR in the input stage, but quite good natural PSRR of -20dB in the output (maybe it has a pentode or BJT collector output). So our 1% ripple (0.5V) appears as 0.05V at the output: 0.1% ripple wrt full output.
So 1% ripple from the PSU can produce 50% or 0.1% ripple at the output. One proviso, which gives you some wriggle room: full output means just below clipping, including any ripple-based clipping. Once you get hard clipping then the amp has become a switcher and then 1% PSU ripple turns into 1% output ripple.
Thanks DVVI see the voltages climb as the speaker sensitivity drops.
Are the voltages above what you measured directly from my test tone?
No, I'm afraid they are not, this was measured in 2002.
Perhaps you'll be interested to know that we also measured voltages at a sound pressure level we agreed was becoming uncomfortably high for any longer period listening. We tried varieties of music types, as I listen to just about everything but classical and jazz.
My own 1041 needed just 20 Vpeak to start ratlling my windows. The JBL required some 24 Vpeak, and the AR needed 27 Vpeak, but let me remind you that its lower impedance also required more than twice the current delivery on average, and a lot more at around 220 Hz, where its impedance cutve drops down to just under 3 Ohms, with a phase shift of some -46 degrees. They sound really good, but they tax the amps withou shame.
That someone was me, thanks very much for the attribution
."Half-bridge" = hype?? Say what? Around here a half-bridge is simply a push-pull arrangement of devices in a single-ended configuration. Typically half-bridge and full-bridge are from the vocabulary of switchmode power supplies and switchmode amplifiers. And since N channel DMOS always outperforms PMOS for the same silicon area, much effort has been made to use the same polarity device for the upper and lower switching duties, which has led to a lot of very clever circuitry for the upper device ("high-side") drivers. But I digress.
Your doubts about class D sounding good may be a surmise based on limited experience with specific systems and a lack of understanding as to how they work. I would agree that class D, particularly open-loop class D based on internal digital-domain modulators, are unlikely to really equal the sonics of a no-holds-barred nonswitching amplifier. And it may be mind-boggling to see the magnitude of the out-of-band spuriae after the typical two-pole LC output filter on an oscilloscope, but if done properly it really is out of the audio band. The opportunities for mischief based on signals that get downconverted to in-band signals need serious consideration, to be sure.
As well, if there is no loop closure around the output filter the inductor and capacitor will contribute distortion, and will as well entail an increase in the output impedance, which in turn will change the frequency response at the load. At an AES talk someone asserted that people listening to a particular switchmode amp remarked that it sounded somewhat tube-like, which led me to bark out that surely this was due to the highish output impedance interacting with the load impedance.
One of the benefits of the Bruno Putzeys UcD approach is that there is feedback around the output filter. BTW I receive no promotional considerations from Hypex
Brad Wood
Another observation relevant to power supply ripple/noise and perceptible output spuriae, for both conventional linear and switchmode amps: we are usually far more sensitive to intermodulation distortion of PS and desired signal than simply a "breakthrough" of the bandlimited sawtooth ripple typical of rectified mains supplies, and this is even more obvious when the IM products both fall into frequencies of high aural acuity and are harmonically unrelated to the music. This would be yet again more obvious if we still paid attention to just-intonational tunings, as within equal temperament there are no pure intervals except octaves, but it's still pretty apparent.
No problem. If you need anything else, just ask. I like your test.
Yessir, he surely does. With Wayne, it's much more about the Amperes than the Volts.
While I would normally be sceptical about such arrangements, I remember the now long gone AR 3a Improved and the Yamaha NS 1000 Monitor. Both were very difficult loads to drive, so much so that Otala used them as examples of how tough loads can become, but both were speakers you never forget once you have heard them properly driven. Quite outstanding, both of them.
The trick was finding an amp capable of doing them justice. My reVox integrated A78 amp choked on the AR at anything over room volume, and the transients sounded, well, bland. It just couldn't cope, nor could your usual suspects (Kenwood, Pioneer, Sansui, Technics, etc). It was only when the reVox A740 power amp came that we (owner and myself) could really hear the AR come to true life.
I remember congratulating myself silently for buying AR 5, which were not as difficult and were not 4 but 8 Ohms - and a lot cheaper than the 3A Imoproved, way over my student budget, but admittedly better than my 5.
True, but not very relevant. If you have 0.5V hum on a 1V signal, which was the first example I gave, then that is 50% - even if it arises from 1% ripple on the supply rail.simon7000 said:
Yes. Hum is easy to hear, but I suspect our brains can fairly easily 'tune it out' and ignore it. Hum IM is more difficult to consciously hear, but I suspect that it may be responsible for some cases of 'muddy sound'.bcarso said:
Do really don't understand both issues?
If you have a 50 volt supply with .5 volts of ripple the ripple only is not passed to the next stage unless you are really trying something weird. If you just have a variable resistor powered through a fixed resistor you will have some loss of ripple along with the DC level. Now if you pass that to the next stage the ripple will decrease if it is an equal gain inverting stage. Only if by some not very common method you use a second non-inverting gain stage will the ripple increase.
HOWEVER ALL AMPLIFIERS THAT WOULD HAVE ANY RIPPLE PROBLEM USE ADDITIONAL POWER RAIL FILTERS ON THE INPUT STAGE! The question was about the main filter bank of capacitors and the effect on power output sags.
As to the second comment. the ear is not very sensitive to 50 or even 100 hz. signals. It is most sensitive to ones around 3,000 hz. What was said is that the intermodulation products up in the higher region sound really bad and are more easily detected than the hum. The hum would be masked by the music in actual use.
If you have a 50 volt supply with .5 volts of ripple the ripple only is not passed to the next stage unless you are really trying something weird. If you just have a variable resistor powered through a fixed resistor you will have some loss of ripple along with the DC level. Now if you pass that to the next stage the ripple will decrease if it is an equal gain inverting stage. Only if by some not very common method you use a second non-inverting gain stage will the ripple increase.
HOWEVER ALL AMPLIFIERS THAT WOULD HAVE ANY RIPPLE PROBLEM USE ADDITIONAL POWER RAIL FILTERS ON THE INPUT STAGE! The question was about the main filter bank of capacitors and the effect on power output sags.
As to the second comment. the ear is not very sensitive to 50 or even 100 hz. signals. It is most sensitive to ones around 3,000 hz. What was said is that the intermodulation products up in the higher region sound really bad and are more easily detected than the hum. The hum would be masked by the music in actual use.
Last edited:
That someone was me, thanks very much for the attribution
Sorry that was a bit rude . I was doing something else and being lazy . The half bridge is hype in so much as people say it like it has positive significance , doubtless Hypex didn't intend it as a mark of quality . As you say it can be driven differentially to overcome the " possible " weakness of half bridge . I can see how it relates to switch mode . A slightly unfortunate association I feel .
My pro audio friend John has great affection for his Class H amps . However increasingly they use D for the weight reduction . He did say D seems less reliable and can not be fixed as easily . New modules mostly . His class H amps usually loose one transistor and by then have shut down . A cheap and fairly instant repair .
My own feeling having been around this industry for years and knowing the big fish is to see class D as a cynical move to take all the business . In all fairness the chip makers publish PCB's so we can get it right . It is not the same as doing it ourselves . When photography came along many professional artists gave up . Impressionism flourished . I see Audio as being in a phase like this . I suspect many truly professional people will soon give up . Leaving the more eccentric people like myself to continue the seemingly old fashioned branch lines . A bit of a sad tale if I am right and I am fairly certain I am . Increasingly my work is of the impressionist type . I am doing that next week . I insisted on including a very low distortion option . The idea is to mimic a tape recorder in compression . Nothing original in it . Just a nice little job and useful to me for my own use . The device in question was common at the BBC but did not include optional compression . It has to drive 100 metres of cable without degradation . It has a slight resemblance to video transmission . I will give it 6 MHz bandwidth as it will do no harm . Then we can use it for other things if we choose .
Last edited:
Is this forum's quote function so difficult to use that I have to fix it in 2 out of three posts????
AR 3a was a three way speaker from its first day, using a 12" bass driver, a 38 mm dome midrange (or similar) and a 19 mm dome tweeter.
My own AR 5, the next model down the line, was the same, except it had a 10" bass driver, but was declared not as a 4 but as an 8 Ohm load.
Acoustic Research contributed to the world of audio with two inventions they had patents for - sealed ("acoustic suspension") type of enclosure and dome type drivers.
The golden age of AR were the 60ies, when everybody who was anybody in classical music owned AR speakers, from Herbert von Karajan all the way down the line. Undisputed king. During the 70ies, the competition got much tougher, and they decayed during the 80ies.
I believe my 94 was from the last series of quality speakers, but I am aware that in its "mk.2" form, it was a completely different speaker, with cheap and nasty drivers, in fact cashing in on the name. After that, they were gone.
Last edited:
Simon, what happens to ripple if one uses fully electronically regulated power supplies?
Specifically, in my case, I plan to run everything but the output stage from fully regulated power supplies, only the output transistors will be fed "raw" DC from the caps.
Assuming I don't botch it up somewhere, as I see it, ripple should be of no concern to me - right?
Acoustic suspension was interesting . As a patent dodge I suspect it got renamed infinite baffle ? If I understood correctly the principle of acoustic suspension was to have the sealed box support the cone rather than have a stiff suspension ? A 10 to 1 ratio if I remember correctly of air to surround ( tightly sealed box ) ? I had a pair of AR'3 and loved them . The rolled off top said to be a reaction to early transistor amps ( I don't know if it's true ) . Later I had some Alison speakers which were similar , I think he was a development engineer for AR ? US speakers that I had . Dalquist DQ 10 , Dynaco A25 ( wonderful ) . Klipsche Forte , La Scala , Forte 2 ( love them ) . I was to meet Ed V , never did as my friend who was going to take me to see him died . Ed and I both lived in towns called Woodstock .
Dvv . I was thinking . You could run everything regulated up to the time loud music kicks in . It could be controlled by a simple thermostatic switch in fact ( the ones with 6.3 mm / 1/4 inch terminals ) . The switch / relay could kick in raw DC when you wanted to party . That would keep the capacitor layout optimum . You could then be posh for the regulated and bargain capacitors for the raw stuff . It would need a bit of thought . The raw ones could be charged ready to go .
Dvv . I was thinking . You could run everything regulated up to the time loud music kicks in . It could be controlled by a simple thermostatic switch in fact ( the ones with 6.3 mm / 1/4 inch terminals ) . The switch / relay could kick in raw DC when you wanted to party . That would keep the capacitor layout optimum . You could then be posh for the regulated and bargain capacitors for the raw stuff . It would need a bit of thought . The raw ones could be charged ready to go .
Yes, Roy Allison did work for AR for a while, and then spinned oof his own company.
Yes, the basic principle was to use the air trapped inside the sealed box as a spring, rather than stiffer suspensions. The net result was a bass which rolled off earlier, but with a much more gentle slope and with a lot less phase funnies the bass reflex can't get rid of. Unfortunately, the acoustic suspension or infinite baffle or sealed speakers were also less efficient than their BR bretheren, and required bigger to much bigger enclosures for the same bass response, which eventually brought them to their present status, that of a dying breed.
Well, that's an idea. However, I will not use it just now, I'll do the amp under development as is, but I will try for it in the next project I have in mind. Quite simply, I'd like to have some time playing with any new idea before I rush in to implement it.
But it's well worth thinking about, methinks.
No. Two different implementations, which did not mixed their names.
Infinitive baffle, is the older (and more generic).
Acoustic suspension, around middle 1950 by Edgar Villhcur, implemented and marketed by Henry Kloss
Yes. The boxed air is to act like a spring to the cone (and cone is to be supported with a compliant suspension)
George
I would say "Bingo!" But as that may not translate properly, it means you win, you covered everything.
As long as your driver stages have the extra voltage to drive the outputs fully, you should get what you are looking for.
The downside is that when some designs done that way clip, they can have a nasty recovery and create messy noises which are easier to hear than a soft clip that is masked by harmonic distortion.
You should look at what happens to the stored charges during clipping, where they go and how things are restored. Most folks only analyze the linear behavior of their designs. If the amplifier is big enough and your source is only digital such as CD's then there is much less chance for clipping. If you play records then things such as record warp can actually cause the system to clip without you being aware of the source. Things just don't sound right.
There is also an issue with the initial attack of electronic music. As this is modulated by the inband musical energy you can again get LF clipping behavior.
Best of luck in your project.
Please explain the difference between infinite baffle and acoustic suspension.
I ask because I am not aware of any, it was as Nigel said a matter of copyrights, but as far as I am aware, both use exactly the same principle and calculations.
Henry Kloss, who later started up KLH and eventually Advent, had no copyright problems simply because he was the co-owner of the original patent.
- Status
- Not open for further replies.