New Onsemi Devices

[sam9]

Has anyone seen these?

http://www.onsemi.com/pub/Collateral/NJL3281D.PDF

http://www.onsemi.com/pub/Collateral/AND8196-D.PDF

Output devices with built-in bias diodes. The application note leaves more questions than it answers. Such as "What if you use an any topology other than an EF tripple with three parrallel output devices?"

Interesting idea, but it seems to me it limits the workable configurations greatly.

[mikeks]

Actually, these devices are quite useful....

But greater flexibility would have been obtained if a high-beta BJT were used instead of a diode....

The user would then have the option of using said BJT in a Vbe-multiplier, or a temp sensor as part of SOA protection, or indeed, merely as a diode.

[mikeks]

Note that for purist application, onsemi's recommended application is sub-optimal....

A Vbe multiplier is still required.....

[tmblack]

Looks interesting.
Does anyone know ON's sample policy?
I know they allow up to 25 devices of 1 type, can someone sample multiple devices at one go?

Tom

[sam9]

I thought about this over lunch. Random thoughts reported---

In the app notes it says the bias is 3.4V so I presume that the voltage drop per diode is 3.4/6=0.56. Assuming for the sake of argument that this is about right for an EF-tripple (tripple meaning pre-driver > driver > output-device), then if you need more than three parallel ODs you just don't connect the additional diode pins.

If you only need a single pair or two pairs you can "trim" the bias by adding external diodes or a Vbe multiplier this looses the advantage of component count reduction but you still get some benefit in tracking speed. You probably have to mount the "trim" devices on the heatsink so there really isn't much simplification.

However, in the case of four output devices and using a mirror image IS + VAS while droping the pre-drivers, 0.56V x 4= 2.24V will be pretty close. (I recently built something like this but with a conventional Vbe-multiplier; bias voltage was 2.3V after setting the RE current).

A close but non-optimal bias that has nearly perfect thermal tracking probably performs better most of the time than optimal steady state bias that tracks with a significant lag. Overall, it makes some combinations of topology and output device count more "do-able" than others. I would have like to have seen a couple of alternative examples in the app-note.

A CFB topology will get no benefit since the output devices are not tracked.

Cheers and applause in advance to the first person reporting back here having used these!!

[ojg]

This is great news! I remember my college professor said in class when teaching us about power stages that he couldn't understand why noone made output devices with built in diodes! I guess his wish was finally granted, it only took nearly 10 years...

It will be interesting to see what advances can be made in class AB output stage distortion. I am sure Mr. D. Self will be looking at these!

[CraigBuckingham]

So is the general consensus that one can't get stable output bias from a Vbe multiplier mounted on the same heatsink as the o/p trannys?

[janneman]

Originally posted by CraigBuckingham
So is the general consensus that one can't get stable output bias from a Vbe multiplier mounted on the same heatsink as the o/p trannys?

No. Doug Self, mentiobed above, has done extensive studies and recommends specific heatsink locations/components to get an accurately tracking temp compensation.

Jan Didden

[ThorstenL]

Konnichiwa,

Originally posted by CraigBuckingham
So is the general consensus that one can't get stable output bias from a Vbe multiplier mounted on the same heatsink as the o/p trannys?

It is not consensus, it is fact. The thermal inertia of the system is well too large to allow the bias to remain accuratly set. You can only prevent extreme runaway, but the bias current will fluctuate.

As Douglas Self has shown, doing that if you use an Amplifier optimally biased into Class B is not a good idea.

Sayonara

[Mr Evil]

These look very much like Sanken SAP15N/P, except they are Darlingtons. I was considering using the SAPs, but the datasheets are a bit light on detail.