Sound Quality Vs. Measurements

Status
Not open for further replies.
And a warm welcome to Russ, who is a friend of old. Strictly a hands-on man, Russ doesn't theorize much, he goes out and tries it out for himself.

Good to see you joining us, Russ, I do believe you can contribute with your experience AND daily job, plus the fact that you are always well composed, unlike a few of us here.
 
Class A amp with your speaker impedance? Are you kidding? :D

<edit> No nog here, only cha :)

Rick, it IS possible.

Technically, I mean, but whether Wayne would be willing to foot the bill is another matter.

Think of Marantz' old S series amps. Forced air cooling, like 20 pairs of high power devices per channel, and so forth. Basically, it all comes down to bare muscle, both in terms of the electronics and the bill for the fun.

The saving grace here is that according to his own data, Wayne doesn't need too much voltage headroom. Accomodating 17V peak is not too difficult, you could do it with power supply lines of +/- 25...30V. Even the trannies are not really a problem.

Cooling them is a BIG problem, and unless Wayne can put up with 6 feet long heat sinks, it will need to be forced air cooling. Really high quality, long life forced air cooling (using Papst fan motors) cost around €160 each, or together, about €320 or about US$ 420. Not at all cheap, but also not too expensive for such a difficult task. I think it could be done for about US$ 1.5k as two mono blocks.
 
Have you ever listened to Etymotic Research ER 4S? Tonally, they are almost indistinguishable from AKG 240DFs, the 600 ohm version now discontinued and the most neutral headphone I have ever heard, but I haven't heard many Staxes. I know when I am recording and I take the ER4Ss off to listen to the real thing (orchestra and choir, usually, Royer SF24 microphone) there is no significant difference above around 60 Hz.
Hello Russell, no I have not heard those in-ears....I have heard a few $100+ in-ears, and none impressed me.
I have always found in-ears (in my limited sample set) to be way too in ear canal position dependent to the point of being quite useless really, and seriously lacking in bass.
I sometimes use a pair of AKG K272 HD phones, and these are particularly nice, but don't quite match the Stax phones....closeish, but I prefer the Stax headsets.
Losing the 60Hz and down is problematic to me, and sounds distorted, ditto loudspeakers with fully extended mid/highs but lacking low bass.
If the high end was rolled out similarly they might sound less distorted, but then we are back to lo-fi.
I will keep an eye out for the ER 4S....I might be pleasantly surprised.

Dan.
 
When really cooking 5-7 amps avg with peaks hitting 17-22 amps is what is used , normally what you state is about norm, the KSA200 handles this drive well , then Again it idles at 14 amps ..:) that amp worked better to me after a re-cap and a lowering of the bias to a more sedate 3 amps, it was much better to me at 1 ohm , the high bias sounded better at 8 ohm ..

I'm not interested in more than 50-80 watts class-a , a-a/b amplifiers sounds the best at this Z, intended updates and mods in the next version will improve sensitivity and raise Zmin to 1.5 , this is a big reduction in amps required and testing over the years shows improved sonics from the amplfiers driving , i have found 1.2-1.8 ohm to be the sweet spot for sonics, 1ohm was the best for drive due to the higher sensitivity, so I'm optimistic the improved sensitivity and higher imp will give me the best of both ....

I would like to see a min of 20-25v at 1 ohm from any amp built for this thing .. :)

Rick, it IS possible.

Technically, I mean, but whether Wayne would be willing to foot the bill is another matter.

Think of Marantz' old S series amps. Forced air cooling, like 20 pairs of high power devices per channel, and so forth. Basically, it all comes down to bare muscle, both in terms of the electronics and the bill for the fun.

The saving grace here is that according to his own data, Wayne doesn't need too much voltage headroom. Accomodating 17V peak is not too difficult, you could do it with power supply lines of +/- 25...30V. Even the trannies are not really a problem.

Cooling them is a BIG problem, and unless Wayne can put up with 6 feet long heat sinks, it will need to be forced air cooling. Really high quality, long life forced air cooling (using Papst fan motors) cost around €160 each, or together, about €320 or about US$ 420. Not at all cheap, but also not too expensive for such a difficult task. I think it could be done for about US$ 1.5k as two mono blocks.
 
Last edited:
Last edited:
ML2
Classé Omega, both 1st and 2nd ed. (v2 btl monaural=> http://www.diyaudio.com/forums/atta...504-sound-quality-vs-measurements-ooh-pen.jpg )

(l'Ampli Fou Fou, by Hephaïstos aka Gerard Perrot, '90, l'Audiophile magazine. my ksa50 mono turd, handles 700W dissipation at 5 amp bias level, rail regulation circuitry from the omega)

There you go, of course it can be done, and obviously is being done.

In the end, it all comes down to space (due to enormous passive heat sinks required if one doesn't want any fans) and of course cost.
 
When really cooking 5-7 amps avg with peaks hitting 17-22 amps is what is used , normally what you state is about norm, the KSA200 handles this drive well , then Again it idles at 14 amps ..:) that amp worked better to me after a re-cap and a lowering of the bias to a more sedate 3 amps, it was much better to me at 1 ohm , the high bias sounded better at 8 ohm ..

I'm not interested in more than 50-80 watts class-a , a-a/b amplifiers sounds the best at this Z, intended updates and mods in the next version will improve sensitivity and raise Zmin to 1.5 , this is a big reduction in amps required and testing over the years shows improved sonics from the amplfiers driving , i have found 1.2-1.8 ohm to be the sweet spot for sonics, 1ohm was the best for drive due to the higher sensitivity, so I'm optimistic the improved sensitivity and higher imp will give me the best of both ....

I would like to see a min of 20-25v at 1 ohm from any amp built for this thing .. :)

For the hell of it, I slapped together a rig in my simulator, just to see what happens.

Turns out you can get away even with only 5 pairs of Motorola's MJL 3281/1302 trannies, HOWEVER, because they would each dissipate 4 Amps at 33V, each would dissipate 132 Watts, and 5 pairs would disspiate 5 times that much. In principle, they could do it, but would require literally mammoth heat sinks.

Switching to Motorola's MJ21193/21194 TO-3 250W devices (which, BTW, were initially designed for audio output stages) would relax things a bit only because they handle high temperatures better. It's no wonder such monsters use almost exclusively TO-3 trannies. Still needs lots of heat sink real estate, though.

I simulated a model with dual power supplies, fully regulated for the input stages and VAS stages, fully complementary, cascode on cascode (meaning the input is a cascode and the VAS is a cascode), with a Locanthi three stage output, meaning a pair of predrivers, drivers and output stages, consisting of 5 pairs of output devices. The front end is fed by +/- 39V, and the rest by +/- 33V, unregulated (i.e. relying on multiple capacitors).

Because of the low output stage voltage, the cost goes down because we can safely use 40V rated capacitors. This is also more practical, since for the same size as a 63V capacitor, the lower voltage version will have greater capacity (e.g. instead of 10,000uF/63V, same can will take 15,000uF/40V, or even 18,000uF/40V).

All in all, it will do 1 Ohm no problemo, but in steady state delivery, the 20-20,000 Hz THD reading is 0.3% for a 20 Vrms output. When the drive is reduced to Wayne's 17V peak, THD drops down to 0.07%. Bias at 1.1A. All easily found off-the-shelf trannies.

I will obviously have to run a few passes on that, there's no way I'll accept more than 0.1% into 1 Ohm worst case. Currently, it's below 0.03% into 4 Ohms and 0.01% into 8 Ohms.
 
For the hell of it, I slapped together a rig in my simulator, just to see what happens.

Turns out you can get away even with only 5 pairs of Motorola's MJL 3281/1302 trannies, HOWEVER, because they would each dissipate 4 Amps at 33V, each would dissipate 132 Watts, and 5 pairs would disspiate 5 times that much. In principle, they could do it, but would require literally mammoth heat sinks.

Switching to Motorola's MJ21193/21194 TO-3 250W devices (which, BTW, were initially designed for audio output stages) would relax things a bit only because they handle high temperatures better. It's no wonder such monsters use almost exclusively TO-3 trannies. Still needs lots of heat sink real estate, though.

I simulated a model with dual power supplies, fully regulated for the input stages and VAS stages, fully complementary, cascode on cascode (meaning the input is a cascode and the VAS is a cascode), with a Locanthi three stage output, meaning a pair of predrivers, drivers and output stages, consisting of 5 pairs of output devices. The front end is fed by +/- 39V, and the rest by +/- 33V, unregulated (i.e. relying on multiple capacitors).

Because of the low output stage voltage, the cost goes down because we can safely use 40V rated capacitors. This is also more practical, since for the same size as a 63V capacitor, the lower voltage version will have greater capacity (e.g. instead of 10,000uF/63V, same can will take 15,000uF/40V, or even 18,000uF/40V).

All in all, it will do 1 Ohm no problemo, but in steady state delivery, the 20-20,000 Hz THD reading is 0.3% for a 20 Vrms output. When the drive is reduced to Wayne's 17V peak, THD drops down to 0.07%. Bias at 1.1A. All easily found off-the-shelf trannies.

I will obviously have to run a few passes on that, there's no way I'll accept more than 0.1% into 1 Ohm worst case. Currently, it's below 0.03% into 4 Ohms and 0.01% into 8 Ohms.

Basically what i had in mind with 7 pr outputs instead and yes mammoth sinks with slow speed fans to cool internal. I still deem the fans necessary turned on thermally to keep continuos play a non-issue. Ps will have regulation for input vas stage and 1kva / ch for outputs with 40v supply . If you get the chance With your sims play around with the bias and feedback to see how much it lowers Thd.

Me also thinks it will need more storage than you are suggesting and anyway i have stock of 90v caps already which i did intend to use with the lower voltage , i'm thinking around 240-280k / ch ...

System is currently bi-amped and eventually i would like to tri-amp , making the panels passive bi-amp ...

I'm happy this is finally getting some thought here :)
 
Last edited:
Rick, it IS possible.

Technically, I mean, but whether Wayne would be willing to foot the bill is another matter.

My remark wasn't meant to suggest it wasn't possible when custom designed, rather that it would never be found commercially off the shelf. No manufacturer in their right mind would go into production with a balanced classA amp designed to drive 1ohm. It would simply be too wasteful of power for the majority of users who had higher impedance speakers.

Let's do a little math - for domestic purposes, an amp's input power is going to be limited by what's possible on a 13A ring main (in UK). So a little over 3kW. In USA I think the maximum power input possible is a little lower due to the lower mains voltage. Of course if the amp were to have an installed supply it could be considered an electric cooker, but let's not go there for now.

So with 3kW to play with, what could we do? CCS-loaded classA is at best 25% efficient so if we burned all that juice it would allow 750W. But the PSU is never going to be 100% efficient - let's assume we used SMPSUs - they might get to 85% efficiency - so perhaps 600W.

For 600W into 1R load the standing current needs to be 34.6A - in balanced this is split between two output stages. The maximum voltage swing is therefore 34.6V and so into a nominal 8R speaker the output power's 75W.

Since this beast takes the power of a 3 bar electric fire as you've noted it'll need some fairly serious forced-air cooling. Or its own refrigeration plant. But sure, its feasible, just commercial suicide for any manufacturer.

<edit> I see a.wayne's max current is 22A so that relaxes things quite a bit - the max RMS into 1R becomes 240W.
 
Last edited:
Because of the low output stage voltage, the cost goes down because we can safely use 40V rated capacitors. This is also more practical, since for the same size as a 63V capacitor, the lower voltage version will have greater capacity (e.g. instead of 10,000uF/63V, same can will take 15,000uF/40V, or even 18,000uF/40V).

Since a.wayne only needs 22A peaks into his 1R then the total swing is 44V. In balanced out there's no need for balanced rails - single ended will be more economical on both caps and regulators. This is power supply invariant - that's the point of the design - so a minimal number of caps is required, With a linear supply, just enough to allow the regulation not to drop out. With a switcher, just enough to filter out the HF hash (with series inductors).
 
My remark wasn't meant to suggest it wasn't possible when custom designed, rather that it would never be found commercially off the shelf. No manufacturer in their right mind would go into production with a balanced classA amp designed to drive 1ohm. It would simply be too wasteful of power for the majority of users who had higher impedance speakers.

Let's do a little math - for domestic purposes, an amp's input power is going to be limited by what's possible on a 13A ring main (in UK). So a little over 3kW. In USA I think the maximum power input possible is a little lower due to the lower mains voltage. Of course if the amp were to have an installed supply it could be considered an electric cooker, but let's not go there for now.

So with 3kW to play with, what could we do? CCS-loaded classA is at best 25% efficient so if we burned all that juice it would allow 750W. But the PSU is never going to be 100% efficient - let's assume we used SMPSUs - they might get to 85% efficiency - so perhaps 600W.

For 600W into 1R load the standing current needs to be 34.6A - in balanced this is split between two output stages. The maximum voltage swing is therefore 34.6V and so into a nominal 8R speaker the output power's 75W.

Since this beast takes the power of a 3 bar electric fire as you've noted it'll need some fairly serious forced-air cooling. Or its own refrigeration plant. But sure, its feasible, just commercial suicide for any manufacturer.

<edit> I see a.wayne's max current is 22A so that relaxes things quite a bit - the max RMS into 1R becomes 240W.

Thats 22A per channel ......
 
Yep - I was (mentally) designing a monoblock amp. Or were you meaning you're bi- or tri- amping?

The caveat here is I reckon that the 1ohm impedance isn't going to be maintained across the whole frequency band - so you'll end up voltage starved in practice if the amp is designed to give only a 22V swing...:p
 
@Abraxalito,

It seems we have a bit of DC offset between us. Our formulations are a bit off.

I know you know it can be done. I meant it could be done for what would still be, given the purpose, sane money. You know, without mortaging the missus.

Wayne's bark is often a lot worse than his bite. :D I figured that out to, that in fact we weren't taking about kilowatts of output into 1 Ohm, I suppose because I'm used to working out what I need into 2 Ohms, and that in peaks only. So I subconsciously must have balked at my usual math, 100W/8R=200W/4R=400W/2R=800W/1R.

Having fiddled with it a bit yesterday, it turned out we don't actually need like 60 power devices, and so forth, but the key problem is cooling it all. It is further made worse by Wayne's desire to run the output stage with a very high bias.

So, as far as I can see, there will have to be some fans, one way or anoter, because the alternative is using two mighty monoblocks, with tons of heat sinking. All these heat sinks would probably sink the deal financially. And he's need a fork lifter to move them around.
 
Nope dont have to have high bias , already said the ksa200 was reduced from 200 watts / ch class-a to abput 20-25 watts . I only wanted for you to play around in your sims to see if it makes a difference. I wont need more than is necessary ...

100W@ 8 ohm was my original target ....
 
Last edited:
So what does the impedance do below 300Hz? That's most likely the area where the amp has to work hardest....

Its rising imp is due to xover impedance, so it's bi-amped for the 20-250 range and i do agree the 25 volt limit is a problem unless you go full active and no passive to roll out the mids , then again it's not driving frequencies below 200-250hz ...
 
Last edited:
@a.wayne There seems to me some misunderstanding - I'm responding to your 'Let me know when I can get one of those' remark. 'One of those' being a balanced, CCS loaded classA. In such an amp the bias current determines the maximum output current - if you need 22A peak output then the bias is equally 22A coz the CCS load is set to that value.

I think going full active has to be the optimum solution here, then it allows each amp to be tailored fairly closely to its respective driver in terms of impedance and hence bias current.
 
Wayne, can you put up with a 3 dB (halving) power loss between 2 and 1 Ohm?

Down from nominally 100/200/400W into 8/4/2 Ohms to nominally 50W into 1 Ohm?

I've managed to get 6 pairs of MJL devices to deliver that, still needing mighty heat sinks, with a total bias current of about 1 Amp, which should suffice for low level listening in pure class A, but without wildly overheating. Nevertheless, the sheer number of output devices mandates two heatsinks per channel and a fan to switch (manual or automatic) on when working at high power outputs or into 1 Ohm loads.

It should be able to deliver short term peaks with a THD factor of 0.1% or less into 8/4/2 Ohms of 182/350/660W, which is kinda powerful for us mere mortal folks. :D :D :D

Your choice of input filter -3 dB points at 200, 150 or 100 kHz by simply installing caps of 470, 680 or 910 pF. NO SWITCHES IN MY INPUT STAGE, I know how you think, the instant you hear there's a choice you want it, plus something extra.

3 level LED power indictaor is an optional extra. Green LED for signal present, amp working, orange LED for nominal 0 dB VU and red LED for clipping warning, saparate for each chnnel. Free of charge for you, a sexy blue LED for power on, just so everybody knows you have a High End device, no blue LED, no High End. :p

That's my "Current Factory" Version 3.0.
 
Last edited:
Status
Not open for further replies.