Need some help with Class A JLH power parameters

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I am going to be building a JLH amplifier. I have read most everything online I can find but I am still confused about a few things. I'll try and list my questions in a way that makes sense but feel free to elaborate with your answer if there is more that should be said.

1) How do you determine the output power of the amp? The way I understand it the available power of the transformer will dictate the output of the amp up to the maximum heat dissipation the output transistors can stand. Is this correct? And if so can you tell me how much transformer is required to safely get the most out of:
A single pair of MJ15003
A dual pair of MJ15003

2) Assuming an 8ohm speaker is being used. What should the voltage and current be set to for the above examples?


Lets start with this and see what happens.

Thanks.


Jeff
 
Jeff,

You have got it in a nutshell in (1) provided that you understand the Safe Operating Parameters of the transistors in question.

We normally approach a design from a requirement side, in other words what is the requirements for an amplifier not how much can I get from...

I have just completed a stereo JLH amplifier capable of about 12 watt (rms) into 8 ohms resistive load using a pair of MJL4281A transistors per amp. One thing I must emphasise is that you will have to get rid of about 45 watt of thermal energy per channel which is no joke.

For my JLH I use a 30V DC regulated supply and the bias current is set to 1.4 amp. The heat sink is rather large and when running it is hot to touch, in other words you cannot hold it for more than about 10 seconds.

I do not use the fan, because it is too noisy, but with the fan obviously it is ice cold.
 

Attachments

  • 15122011068.jpg
    15122011068.jpg
    739.3 KB · Views: 479
In a nutshell, you should consider is that half the supply voltage will be dropped across each output device, so together with the bias current, it should fall inside the DC safe operating area of the device, de-rated by 1.5 watt/deg C temperature rise above 25 degrees centigrade.

Lets assume that you can control the heat to a maximum of 60 degrees centigrade. That would mean 60 - 25 = 35 deg. De-rate the device by 52 watt.

The device at 60 degrees is de-rated to 197 watt. A class A amplifier is about 25% efficient, therefore the peak audio power that you could expect from the device would be one quarter of 197 watt or 49 watt.

The output voltage swing into 8 ohms to realise this power would need to be about 20 volt and current would be 2.5 amp.

Most of the folk here would probably agree that this is very close to the absolute maximum of the devices and two devices would be much safer in the real world with varying ambient and varying mains voltages, etc.

In my humble opinion a 50 watt JLH amp would require a 45V DC supply 2.5 amp bias current and at least two pairs of output devices as well as an enormous heat sink.
 
The power of the amplifier is defined by the supply voltage and the load.
The transistors should be chosen to deal with supply voltage and the predefined minimum speaker resistance on the output.

The transformer should provide the maximum current the supply voltage will put out into the load plus a bit for losses into the heatsink. For class AB 60% of power goes to load and 40% is lost in heat in the heatsink.
 
JLH with shunt regulated drive current, and a pair of Schottkys, can shape
square law vs square law (the difference of such curves is a straight line).
Like Class A, no transistor switches fully off. Like AB, dissipates a lot less.

Three or four component shunt can reduce wasted heat a hundred Watts.
Normal JLH class A was never a straight line vs straight line to begin with.
Both curves bent severely by beta droop, and in a heat wasting direction.
My suggestion merely reverses the direction of that curvature...
 
Last edited:
Krell KSA50 is an over biased class AB amp and 1.9^2 x 8 = 28.8 watt. The only real class A would be single ended I guess.
I don't know why you stated such.
The Krell KSA50 is a ClassA amplifier that achieves 50W of ClassA output into 8r0.

The maximum output from a single ended ClassA amplifier is equal to the output stage bias current. The JLH is not a ClassA SE output stage. It has some form of modulated bias.

The maximum output from a Push-Pull ClassA amplifier is equal to twice the output stage bias current. Some of the later Krells were not ClassA, they had some form of modulated bias.

If the KSA50 has a bias of 1.9A, then the maximum ClassA output current is <3.8Apk. The maximum ClassA power output is <3.8^2 * 8r0 / 2 < 57W, but not as low as 28.8W
 
Kenpeter, give and example of your suggestion it may be interesting to try.
These four parts grafted onto your existing JLH...
Turn it into square law vs square law Class A, and a lot less heat.

Afterthoughts of this hurried drawing:

Q1 ought to be on or near output heatsink, so it can feel the heat.
BD139 has a mounting hole for such purposes , 2n2222 does not.
2n2222 probably wasn't the best choice I could have illustrated.

The Schottky diodes should never be bolted to the output sink, but
bent away in a slightly cooler location. This (and Q1 at full heat)
assures the shunting will shape a safe temperature coefficient.
 

Attachments

  • fourparts.gif
    fourparts.gif
    8.5 KB · Views: 287
I don't know why you stated such.
The Krell KSA50 is a ClassA amplifier that achieves 50W of ClassA output into 8r0.

The maximum output from a single ended ClassA amplifier is equal to the output stage bias current. The JLH is not a ClassA SE output stage. It has some form of modulated bias.

The maximum output from a Push-Pull ClassA amplifier is equal to twice the output stage bias current. Some of the later Krells were not ClassA, they had some form of modulated bias.

If the KSA50 has a bias of 1.9A, then the maximum ClassA output current is <3.8Apk. The maximum ClassA power output is <3.8^2 * 8r0 / 2 < 57W, but not as low as 28.8W

You are right 57 watt peak, but KSA remains just an over biased class AB, you can bias a KSA50 as low as a few mA.
 
Thanks Kenpeter, I will try your idea tomorrow and report on the practical finding -
You might throw a switch across one or both of those Schottkys.
Any less than 0.65V total drop across here will shut off Q1, and
return the circuit back to normal (extremely warm) JLH operation.

A DPST switch might also provide convenient mounting location
for Schottkys? Since there are tempco benefits for these to be
away from the heat that Q1 is sensing.

Even if you got tempco completely wrong, and Schottkys run away:
Our worst case is merely a return to normal JLH, just as-if shorted
on purpose by the switch. Only of concern if cheating with smaller
heatsink than JLH would normally require.

I don't forsee any such problem. As these diodes are TO-220, and
not dropping enough voltage to make significant heat of their own..
Wire loss to such a switch only makes the current sensing more
stable (slightly resistive) , so a remote diode mounting is fine.

Make sure your Schottkys are not rated in excess of the maximum
current you need them to carry. Else you may not be riding on the
best part of the square law curve.
 
Last edited:
No the peak power output or instantaneous maximum is ~116W into 8r0, from 3.8Apk and 30.4Vpk

It is not a Krell KSA50, if you choose to ignore the manufacturer's set up information and miss-set the bias to less than 1.9A.....
If the KSA50 has a bias of 1.9A, then the maximum ClassA output current is <3.8Apk. The maximum ClassA power output is <3.8^2 * 8r0 / 2 < 57W, but not as low as 28.8W.....
The JLH is not a ClassA SE output stage. It has some form of modulated bias.

Whatever...

BTW. I never said JLH is SE, I said the only real class A is SE. Kindly explain the modulated bias you propose for JLH?
You cannot run this JLH in class B, but you can run a KSA50 in class B so KSA remains an over biased class B
 
Last edited:
Hi Kenpeter, tried it and your idea works. It runs cold, only thing is the distortion is very much audible now. I tried different values of R which changes both the power dissipation and distortion. The higher the dissipation the less the distortion. When I say audible it is quite obvious, probably 5% or more (subjectively).
 
Last edited:
I don't understand the active bias part of the JLH.
What I can confirm is that the output bias varies with through put signal level.
But I don't know how to prove that.

As to your other point.
Your SE to PP comparison is in effect claiming that only SE can be ClassA and that PP must always be ClassB or ClassAB.
In my opinion that claim is completely wrong.
I state and firmly believe that both SE and PP can be ClassA.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.