Ultimately, isn't the Holy Grail of amplification a true VOLTAGE source?
An amp should ideally be able to deliver its nominal voltage into almost any load within reason, say down to 2 Ohms (a debatable low impedance point) without caring whether the speaker behaves as an 8 or 4 or 2 Ohm load?
Not many manage it, we all know, some are better at it than others for many reasons, but that is the ideal, is it not?
It's the voltage that counts, the associated current is just colateral damage we have to deal with along the way. Again, some more than others, and I am definitely not thinking only about Wayne's 1 Ohm load, commercial speakers often referred to as high quality (in their day) were often downright evil loads. Just to remind you, Otala used AR 3a Improved and Yamaha NS-1000, very highly regarded in their day, as material for investigating TIM and interface problems.
An amp should ideally be able to deliver its nominal voltage into almost any load within reason, say down to 2 Ohms (a debatable low impedance point) without caring whether the speaker behaves as an 8 or 4 or 2 Ohm load?
Not many manage it, we all know, some are better at it than others for many reasons, but that is the ideal, is it not?
It's the voltage that counts, the associated current is just colateral damage we have to deal with along the way. Again, some more than others, and I am definitely not thinking only about Wayne's 1 Ohm load, commercial speakers often referred to as high quality (in their day) were often downright evil loads. Just to remind you, Otala used AR 3a Improved and Yamaha NS-1000, very highly regarded in their day, as material for investigating TIM and interface problems.
Yes, I would agree. The reality, however, is that the power supply has to work 2, 4, 8 times harder to maintain the 'design' voltages, which of course doesn't happen. Sag occurs, and it can be very vicious, particularly nasty spikes of voltage collapse on a transient, peak current demand.
This is what benign speaker loads, very sensitive,
, drivers keep well at bay - Klipsch knew he was on to a good thing ...
This is what benign speaker loads, very sensitive,
, drivers keep well at bay - Klipsch knew he was on to a good thing ...Yes, I would agree. The reality, however, is that the power supply has to work 2, 4, 8 times harder to maintain the 'design' voltages, which of course doesn't happen. Sag occurs, and it can be very vicious, particularly nasty spikes of voltage collapse on a transient, peak current demand.
This is what benign speaker loads, very sensitive,
The power supplies may have to sag, but by how much? If duly oversized right from the start, the sag will be rather small.
For example, my own rails sag by just -3% when working into 2 Ohms, however, this is in comarison with 4 Ohms, since the amp was initially designed for 4 Ohms as its default work load. Also, I use uncommonly large power transformers, 500 VA per side, for a nominal load of 200 Watts into 4 Ohms, again per side. All this backed up by 53,200 uF per side.
Sag also works the other way as well. While I have 28.3V into 4 Ohms, my nomial voltage rises to 28.7V into 8 Ohms; however, I consider these to be small variations not worthy of particular note. In short, they were expected, since there is no electronic regulation of all current gain stages, only the predriver has regulated power lines, the driver and output devices do not and rely only on the capacitors (but I do use distributed filtering, with some caps immediately next to the output devices).
I hate the industry practice of using a relatively small number of output devices, fed by extremely high voltages, but with a very large sag factor when they run into low impedance loads. I believe this to be an exclusively economic function, it's simply cheaper to do it that way and boast large nubers of Watts. These are paper Watts. My ears are not made of paper.
Our HI FI Chioce magazine would have loved your amp . Almost a constant voltage source . I suspect on real music it would drive 1 ohm . If you shunt an 8 ohms speaker with about 1R5 I bet it will still clip at 40 V peak on music . Crest factor of music is 1/6 or 1/3 Techno . Hi Fi or PA that is .
It might actually drive a 1 Oghm load, not like Electrocompaniet (spec for 1 mS, for the love of Christ), more like 35-40 mS, but in peaks only. It doesn't have the wherewithall for constant output, either in terms of output stage SOAR, or heat sink cooling capacity. Repeated peaks into 1 Ohm will either blow the PSU rail fuses, or trigger the overheat protection, or both.
BTW, while not shown on my graphic, it has at its output not one but two power relays. One is rated at "only" 16 A, and while in peaks it will probably pass through about 24 A (just guessing here), I figured two were a better deal, all the more so since they are in series with the output, so two in parallel half their inherent impedance. Besides, they are hardly expensive.
Let me put it this way - I will guarantee that its output into 2 Ohms will eventually be limited by the suppply rails and nothing else. If it manages to drive even 1 Ohm loads at whatever power output it can manage, I won't be the one to compalin.
I did what I could to make it as trouble free and as reliable as I could. I used excellent heat sinks, separate for each channel, each one rated at 0.56 K/W (Fischer SK51, if memory serves), of a fine size (300 mm long, 100 mm high, 40 mm deep, of which the base only without fins is a solid 10 mm thick, output and driver trannies bolted directly onto the heat sink, no in-betweens, spread evenly longitudinally so as to make best use of the heat sink). Overheat protection will fire when the sensor is at 65 deg C, meaning the output devices will be 85-90 deg C inside, the limit of full reliability. Assuming the inside of the power trannies to be 85 deg C, according to the data sheet the transistors should still be good for 114 W of dissipation each, and there are three pairs. You work it out.
I suppose I could add a fan in the center, but I honestly don't see the point, as I made it clear this was not a PA device but a home amp. As a home amp, I think it goes some way above and beyond the call of duty.
However, all that is subsidiary, really, the key is how it will ideally not sound, as it should have as little sound of its own as possible. For that, we'll have to wait a bit, I am currently financially indisposed, as I just plonked down some serious dough, about US$ 12k for fitting my son's aparatment, to which he moved the day before yesterday. At almost 28, it was about time, too, but here, things work much differently than in what you would consider normal. Anyway, I am compiling a list of the parts I need, and will proceed in steps, to ease the financial load.
Thankfully, I already do have the two of the three key parts, the heat sinks and the filter capacitors, lacking are only the toroids. I also have two rails of Motorola/ON Semi MJL 3281/1302, each with 20 devices inside. To quote British pilots from the Battle of Britain (my favorite WWII movie):
Tally ho, chaps!
2 Ohms in parallel with 2 uF?
Holy cow! Nigel, The master Blaster! That's sadistic, Nige.
But I'm willing to try ...
No 100 uf . It's horrible . 45 degrees bang in the mid band .
A relay I am using for power engineering is rather nice . 16 A and 80 A peak ( florescent rating , never seen it before ) . If bypassed with a capacitor I see no harm in it and put the feedback loop around it will help. The capacitor also say you have been naughty as a little sound is still heard . 1 uF I guess would be ideal ( preferential path ) .
1-1415899-6 - TE CONNECTIVITY / SCHRACK - RELAY, POWER, PCB, 12VDC,16A | Farnell United Kingdom
A relay I am using for power engineering is rather nice . 16 A and 80 A peak ( florescent rating , never seen it before ) . If bypassed with a capacitor I see no harm in it and put the feedback loop around it will help. The capacitor also say you have been naughty as a little sound is still heard . 1 uF I guess would be ideal ( preferential path ) .
1-1415899-6 - TE CONNECTIVITY / SCHRACK - RELAY, POWER, PCB, 12VDC,16A | Farnell United Kingdom
definition of ideal coil.. i think it should have zero dcr, therefore current fb amp would be ideal for driving it.
for example fostex coil inductance would give xl <2ohms at 5kHz, so at bass freq a ideal voltage amp would work almost to shortcircuit-like load
+ and if speaker had no resonance peak
for example fostex coil inductance would give xl <2ohms at 5kHz, so at bass freq a ideal voltage amp would work almost to shortcircuit-like load
+ and if speaker had no resonance peak
A very interesting point . Current drive is a bit of a naughty way around it . I built by accident a current drive amp . I was being too optimistic when having only collector outputs . It only work well as a guitar amp ( very ) . That it worked at all now pleases me . It was like complimentary feedback paires without the emitter coupled transistors . A big VAS and CCS with loads of feedback . I guess it was running 1 or 2 amps ?
That's the one, Nige, except mine has a translucent plastic cover and tries to look very high tech.
YOU said 2 Ohms and 2 uF, not me, I'm still a flower power kinda guy.
Finder have a nice one with 30/16 A rating . I uses them and love seeing though the cover as the truth of how well they are working . I tried putting them in oil as it is said the arc is due to air . Not a scrap of difference . Great to see through the device when a reactive load . I usually use a 2000 watt vacuum cleaner to do tests .
One thing about relays . Read the specs very carefully . 80/30/16A coupled with <400 VAC of the TE relay is exceptional , as is the price at $1.50 .
Miniature Relay SPDT 16A
One thing about relays . Read the specs very carefully . 80/30/16A coupled with <400 VAC of the TE relay is exceptional , as is the price at $1.50 .
Miniature Relay SPDT 16A
In a normal power amplifier the voltages rails are an extra signal input, of the worst kind; which one doesn't "see" if feeding with a simple, constant sine wave. Easy enough to set up a second amplifier to monitor what's really happening on the rails, allowing one to 'listen' to the behaviour when driving hard into a load with real music - and it's a mess ...
For some strange reason,
, the more messy the voltage rails become, the worse the sound coming from the amplifier driving ... all them fancy PSRR ideas seem to lose their clout at this point ...
Getting good sound is a juggling act - you're balancing all the factors which degrade the sound in various ways, such that their subjective impact is minimal ... or you should be, . If a component delivers excellent sound as is, but is very suceptible to interference effects then do a clean room treatment on it externally, baby it, so it can do its best, as is.
Likewise internally - a circuit may be intrinsically very resistant to 'damage': the supporting subcircuits may be a rubbish heap, electrically speaking - but you can get away with it because the topology of the crucial elements, that directly processes the audio, is very robust.
The overall process is then to do whatever is necessary to get to an adequate SQ state: finesse the power supply so it's cleaner, or make the amplifying circuitry 'tougher', or perhaps pre-process the mains being fed to the unit so the PS, as is, has an easier time of it - any, or all, or a combination may be necessary to get one there ...
When 'difficult' recordings start to come to life, start making musical sense rather than just being an endurance test, is the best guide as to how far you've come, what progress has been made ...
. If a component delivers excellent sound as is, but is very suceptible to interference effects then do a clean room treatment on it externally, baby it, so it can do its best, as is.Likewise internally - a circuit may be intrinsically very resistant to 'damage': the supporting subcircuits may be a rubbish heap, electrically speaking - but you can get away with it because the topology of the crucial elements, that directly processes the audio, is very robust.
The overall process is then to do whatever is necessary to get to an adequate SQ state: finesse the power supply so it's cleaner, or make the amplifying circuitry 'tougher', or perhaps pre-process the mains being fed to the unit so the PS, as is, has an easier time of it - any, or all, or a combination may be necessary to get one there ...
When 'difficult' recordings start to come to life, start making musical sense rather than just being an endurance test, is the best guide as to how far you've come, what progress has been made ...
Last edited:
That's it, Nige, exactly like that, only by Schrack. In effect, I suspect it comes from a common source, but I can't prove it.
In a normal power amp you are really listening to the PSU ....
Yeah, I'll buy that. Especially so after I have experienced what a decent change of the large caps can do all on its own. And then some, after reattaching the suspect amp to an external power supply done right; on occasion, I was surprised at the benefit in sound, telling me what a poor power supply it has.
I have spent a lot of time looking at service schematics to see how various people do it. For example, it's obvious that in its 1985-1995 series of integrated amps, Sony makes it a point to use a large and a small cap in parallel, the "small" cap usually being about half the value of the big one. Tried it and found it to be pretty good on average.
Even a quick look at capacitor specs will show that two 10,000 uF caps will do better than a single 22,000 uF cap, all else being equal.
Fiddling around with what is effectively a Zobel attached to the PSU lines can do more for the sound than I'd care to try to explain. A little experimenting is required, but it's not much and it's easy, while what happens to your amp's mid and treble range is no joke, it is a very real benefit.
Thinking about your input stage PSRR should be done when designing. Unfortunately, due to purely economic reasons, input stage CCSs are usually reduced to simple resistors only because that's cheap.
Last but not least, it's always a good idea to divorce your input stage and VAS PSU lines from those feeding the currentgain stage. Far too many benefits, the only downsides being once again the associated cost in money and PCB real estate. But soundwise, this approach wins hands down.
- Status
- Not open for further replies.