Hi guys ,
how much power do you think can be pulled out of 12pairs of 2sc5200 or NJW0302G on a 110VDC rail on 4ohms there will ample cooling and short-circuit protection based on hall effect sensors. I am confused because i see profession companies pulling out 800+ watts out of only 7pairs do they not calculate SOA?
I am talking about simple class Ab no class h or g ....
how much power do you think can be pulled out of 12pairs of 2sc5200 or NJW0302G on a 110VDC rail on 4ohms there will ample cooling and short-circuit protection based on hall effect sensors. I am confused because i see profession companies pulling out 800+ watts out of only 7pairs do they not calculate SOA?
I am talking about simple class Ab no class h or g ....
Sure they do. But it is common to use the 10 millisecond pulsed SOA without any temperature derating. Do you feel lucky, punk? 
12 pairs of 0302 would handle a 4R resistive load, but not quite a 60 degree reactive one. You would need to limit the current at full Vce to about 9 amps - which might handle the real current swing you’ll get assuming typical supply sag under load. For 2 ohms or fully reactive 4 ohms you will need more. The 4281/4302 pairs are really what you want here. Don’t try this with the 5200. SOA at 100 volts is nowhere near good enough. Stacking 2 banks of 6 in series would have enough SOA, even with 5200’s, but that arrangement has its own problems. Most people would rather eat glass than deal with those, but it can be done. I have - with 10 pair of 2119x in series/parallel on +/-127 and had flawless reliability at 2 ohms, after chasing down non-destructive but annoying oscillations for months.
A common misconception is that class G or H reduces the SOA requirement. It does not. The peak stresses in the devices are exactly the same as class AB, just reduced in duration, statistically. This reduces the average current requirement and average heat dissipated for almost every real world use, But the devices still see large stresses. The only way to actually reduce them is to use more than two steps on the supply rails. You don’t see that very often, and then only on the very largest and most expensive amps.
12 pairs of 0302 would handle a 4R resistive load, but not quite a 60 degree reactive one. You would need to limit the current at full Vce to about 9 amps - which might handle the real current swing you’ll get assuming typical supply sag under load. For 2 ohms or fully reactive 4 ohms you will need more. The 4281/4302 pairs are really what you want here. Don’t try this with the 5200. SOA at 100 volts is nowhere near good enough. Stacking 2 banks of 6 in series would have enough SOA, even with 5200’s, but that arrangement has its own problems. Most people would rather eat glass than deal with those, but it can be done. I have - with 10 pair of 2119x in series/parallel on +/-127 and had flawless reliability at 2 ohms, after chasing down non-destructive but annoying oscillations for months.
A common misconception is that class G or H reduces the SOA requirement. It does not. The peak stresses in the devices are exactly the same as class AB, just reduced in duration, statistically. This reduces the average current requirement and average heat dissipated for almost every real world use, But the devices still see large stresses. The only way to actually reduce them is to use more than two steps on the supply rails. You don’t see that very often, and then only on the very largest and most expensive amps.
It reduce the voltage across the output devices, which can help. The lower dissipation reduce the temperature, so less derating is necessary.
Sajti
The problem is that it reaches the same maximum stress condition, the instant the high rail switches in. The duration (and frequency) of high stress conditions is drastically reduced, reducing the overall heat load. If you are set on designing conservatively such that it stays inside the DC SOA for all conditions you save nothing, other than maybe 25 degrees of derating - if you use the same heat sink you would have. Nobody ever does. And a common occurrence in the PA world is for the operator to set an aggressive limiter on the subs and slam it such that you’re putting out a sine wave with little or no crest factor. If the limit threshold is set just a bit above the class H trip point, it will sit there and dwell at those high stress conditions far longer than it was ever intended to. Totally making the idea behind it backfire.
If you really CAN get 0302’s for 80 cents (and not fakes) it may in fact be the bargain - if the only limiting factor is dollars. But I would certainly double check that because typically they are a bit over half the cost of 2119x if you get the, from the same source. The MJW2119x really is king for SOA if the limiting factor is the number of devices you can fit on the heat sink before you run out of length. Some people have an aversion to single digit fT output devices, hence the 0302. Beta is lower, but you just use a 0302 for a driver instead of a 15035, and do not be afraid to use an EF3. With the number of devices in parallel just to get SOA, beta droop at high current just isn’t an issue. Peak currents won’t get above 8A in each device, and with most loads will stay under 5.
I question the class G scenario as the upper devices are typically run as linear pass transistors once the required voltage gets close to the lower rail, the high rail operating condition has typically only a volt or so of Vce across the lower device, with the remainder dropped across the upper one.
There are rail switching class G designs out there, but meh, sounds like a pain in the **** to pull off, and as you say, no real SOA advantage!
I have actually seen a design aimed at highly inductive loads where the innermost rail was actually 0V, done as a 'three rail' design with the upper two rails being somewhere in the +25, +50 region.
There are rail switching class G designs out there, but meh, sounds like a pain in the **** to pull off, and as you say, no real SOA advantage!
I have actually seen a design aimed at highly inductive loads where the innermost rail was actually 0V, done as a 'three rail' design with the upper two rails being somewhere in the +25, +50 region.
They will burn anyway.
Speaker is a non-resistive load and worst case your design need at least 2/3 of maximum current to be available at 0 Volts at the output.
Let's check.
110/4 =~30 Amps at max.
So you need to deliver around 20 Amps worst case at 0 Volts output.
SOA at 110 V ~0,5-0,6 A, so 12 pairs could provide 7-8 Amps being realistic.
Now, you can calculate class H/G rails drop benefit.
Yes maybe it will burn away , but strange how the renowned crown xls802 survived with 100+ rails and only 6pairs of mjl transistors
Last edited:
20 amps worst case, but 10 being typical which you would have to support. Which you could almost pull off. Set the zero volt current limit at 10, add an extra pair or two, and it would for the most part work. Especially considering the supply will probably drop to about 80 when heavily loaded. The maximum that occurs half way up the voltage swing with a resistive load *will* happen regardless of how well behaved your speaker is and there’s no way to cheat that. Assuming a 3 ohm DCR, that’s a kW of short term dissipation which even a class H or G amp would have to deal with. And that NEEDs to be temperature derated because it can happen 50 milliseconds at a time, repetitively. For reactive load driving the voltage and current are out of phase, limiting the time where both are high together. And the impedance magnitude minimum can not occur at maximum phase mathematically. If neither were true people would be blowing out their amps a whole lot more than they do.
And I would not call the XLS802 “renowned”. This was about the time even Crown started making stuff cheap. Very cheap compared to Macrotech, and not in the same league.
And I would not call the XLS802 “renowned”. This was about the time even Crown started making stuff cheap. Very cheap compared to Macrotech, and not in the same league.
Exactly, but this is DIYA where people spend their time building phono pre-amps and tube amps, and circuits that constantly blow up and require rebuilding over and over again. It's a style thing like exotic cars etc.
Probably true. The biggest AB I would consider would run on +/-80 putting out 400-ish per channel a 4. This requires a supply which drops no lower than +/- 62 volts when loaded with 28.2 amps, and 6 or 7 pairs of metal 2119x ( 8 to 10 plastic). Above that, the amps get exponentially harder to build. Anything more and the speakers are at risk so the power Isn’t needed anyway.
PA use these days requires larger amps, but for that some compromises in the ultimate sound quality can be accommodated. D,G,H, or TD are appropriate. Face it, you will never hear the distortion from increased dead time in huge power class D or the commutation distortion in 4-step class H when you’re slamming the limiters and driving loudspeakers to distortion. Tweeter drivers at 110dB per watt can run well with 200 to 400 watt AB’s if you’re worried about critical high frequencies.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.