Dimensioning (pre)drivers for EF stage

Hi,

I want to design an amp for a 2 ohm load. I'm basing my design on the book by Self. What he doesn't explain is how to dimension the (pre)driver section when you parallel multiple output devices. Is there any literature on that? I can only find analysis of EF stages that are pre-designed, not how to figure out the resistances and BJT requirements yourself.
Thanks in advance.

Kind regards,

Remco Poelstra
 
Hi,

Since my supply is limited to +-33V I can not put more than about 500W into 2 Ohm. I don't need that much power, it's just that my ESL's can present a 2 Ohm load, that I want my amp to be able to handle it. I think I will not design my PSU to handle that power for a long time, so actual power output will be lower.
I've made a drawing of what is in my mind at the moment, including component values that I know. I'm not sure at the moment whether I really need the pre-drivers.
Since the beta of the 2SC3519 is about 50, does that mean that the driver section needs to be able to deliver 330mA? (16A in 2 Ohm, divided by beta).
Thanks in advance.

Kind regards,

Remco Poelstra
[IMGDEAD]http://remco.beryllium.net/ef.pdf[/IMGDEAD]
 
are you using a triple as the output stage (pre-driver+driver+paralleled outputs)
or a double (driver+paralleled outputs)?
What is the maximum Vpk into your load?
What is the maximum Ipk into your load?
Can I peak get any worse if the transient is very short term?

Once you know the maximum Ipk to the load, you can work back through the stages using the hFE and curents to determine the loadings for the previous stages.

the beta of the 2SC3519 is about 50, does that mean that the driver section needs to be able to deliver 330mA? (16A in 2 Ohm, divided by beta).
Remember hFE varies with Ic.
hFE @100mA could be 100, and @1000mA could be 90 and at 16000mA could be 10.

You must use the hFE appropriate to the worst case peak current you have determined.
 
Hope this helps...

With a plus/minus 33 Volt supply and a 2 Ohm load, call the peak current 16 Amps. Assume that the 2 Ohms is so low that we can’t do something too much more horrible with a reactive component of speaker impedance. Worst case dissipation in an output transistor is then about 33/2*33/2=272 Watts. If you say the max ambient you want to run in is 50 C, then with an infinite heatsink, you can dissipate 100 Watts. That says you need 3 2SC3519 output devices in parallel. I’ve tried to pick assumptions that fall on both sides of the conservative design line...maybe reaching something not too out there.

That says each of the three transistors on top (or three on bottom) will see about 5.3 Amps of collector c current. Data sheet shows worst case beta of about 30 at 5 Amps. This says the drivers need to put out 16/30=0.5 Amps. You could use MJE15034 and MJE15035’s. Their worst case dissipation is 16V @0.5 Amps, which is within their SOA, although they will need heat sinks.

The MJE15034 and 15E15035 have worst case beta around 100 (wow, that’s pretty good) at 0.5 Amps of collector current. That makes their base current peak at around 5 mA. You could get that from a driver, e.g. the Vas stage, or you could put a third emitter follower in the path.

Now you need only to do something with the turnon/turnoff resistors. They’re typically (for a single transistor) around 100 Ohms. For three transistors, maybe 33 Ohms is pretty good.

Picture is a sketch of the output stage portion...of course, you must add reasonable stuff around it (I used a different output transistor than you called for, but just change the names). The 10 Ohm resistors in the base may need some experimentation. They are there to stop parasitic oscillation of the emitter followers.
 

Attachments

  • tripleout.PNG
    tripleout.PNG
    54.4 KB · Views: 79
This is really getting me somewhere. Thanks for all insights. I've just a few questions: am I correct in assuming that the resistors r5 and r6 have nothing to do with setting the current that the drivers can deliver?
My heatsink has an Rth of 0.7. It's shared with all output transistors, so I can never dissipate 300W, but do I really need to do so? My PSU will limit the output power and the amp will switch of at 75C heatsink temperature.
I don't really see why the worst case distortion of the drivers is 16V at 0.5 mA. Why do they only see half the supply voltage?
Since my VAS is running at 7mA I think I can do without the pre drivers as only 5mA is needed to drive at full power. Maybe it introduces some extra distortion, but at such output powers I don't think that's a problem.
 
Djoffe has made an assumption that a 2ohm speaker is no more difficult to drive than a 2r0 resistor.

This assumption is completely unrealistic.

I can not recommend you follow his advice on peak current design for your VAS & output stage.

A double EF output stage cannot drive a real 2ohm speaker. The impedance seen by the VAS is far too low and that ruins the VAS performance as well as requiring enormous current swings in the VAS current.
 
Well, the load will be fairly capacitive I suspect, being an ESL, but the amp should work with general loudspeaker loads.
Vpeak is of course 33V , as the power supply can not deliver more, but how does the capacitive load increase Ipk? I only suspect a phase change.
I suppose his basic reasoning is correct. I'll use 3 output devices, since I think that eases power handling, and if I understand it correctly, you advise to keep the pre-drivers in to minimise VAS loading?
 
The same place?

Andrew,

We may be in the same place, although through a fortuitous mistake on my part. I made the too optimistic assumption, you believe, that the 2 Ohm resistor isn't worse than a speaker...I also made a mistake in my worst case power calculation on the 2 Ohm load on +/-33Volts. Worst case dissipation is with output at 16 volts, when output current is 16/2=8 Amps, so total worst case power is just 128 Watts, not 250 ish watts. I had recommended a triple, so it would just be 40 Watts per output device into plain 2 Ohm load...

So...maybe this happy mistake of a factor of 2 leaves us with the extra margin that deals with the 2 Ohm reactive load...

I can't excuse my calculation mistake, but at least in this case, we may have sidestepped Murphy's law and come to a safe place with the triple?

Dan
 
Favorite Test Case?

Does anyone have a favorite worst-case test case for SOA and reactive loads?
AndrewT has rightly said that it's too optimistic, even with a 2-Ohm load, to say things can't get much worse than just the 2-Ohms. That brings up the natural question, has someone cooked up a worst case test case that seems to correlate well with real life stresses?
 

wg_ski

Member
2007-10-10 5:21 pm
Capacitive loads tend to make stability issues and switchoff distortion more apparent. You may want to listen to prototypes built several ways - one with a triple and another with a buffered VAS (run at high current). You do need the extra current gain stage - it's just a matter of where it goes. Driving a capacitor (ESL) you may find it more advantageous to have it upstream of the Vbe multiplier where it can't contribute to "class B" type distortion or local instabilities in the OPS.
 

wg_ski

Member
2007-10-10 5:21 pm
Does anyone have a favorite worst-case test case for SOA and reactive loads?
AndrewT has rightly said that it's too optimistic, even with a 2-Ohm load, to say things can't get much worse than just the 2-Ohms. That brings up the natural question, has someone cooked up a worst case test case that seems to correlate well with real life stresses?

1.4 ohms fully resistive, AND 4 ohms fully reactive, for most speakers.

ESLs aren't really going to draw killer currents. Minimum impedance is at HF, where the peak to average ratio is staggering. You need to supply high peak currents w/o distortion, but not for very long. At low frequency where all the sustained power is you drive an open circuit. A typical SOA limiting circuit won't kick in instantly. You normally want it to limit in 1 to 5 milliseconds which protects sustained overload and short circuits, but lets single-event reactive currents through without limiting. The 100us SOA of most any power transistor is square, and the 1ms is many times the DC, even derated for temperature. Very short, relatively infrequent current spikes outside the DC SOA are NOT what causes amplifiers to fail. Too much bass (many woofers in parallel) and too much temperature do :D
 
Does anyone have a favorite worst-case test case for SOA and reactive loads?
I use David Eather.
Bensen turned this paper into an Excel spreadsheet.
I modified that spreadsheet to include BJTs and to de-rate to elevated Tc.

As a guide many would recommend that the amplifier must safely pass at least twice the current that a nominal resistive load would demand while the drive voltage should still be substantial.
I go further and ask that the amplifier passes 3times the current that a nominal load would demand and that the output voltage should fall less than 2dBV when passing this current.
Note, that "at least twice" and "drive voltage be substantial", gives a designer a lot of freedom on where his pencil shall strike the SOA limits.
Inform, decide, prototype, test, listen, modify your design decisions if necessary.