1. Home
  2. Electronics Design
  3. BJT Pair for Push-Pull, 1W, 33V, 200mA, 1MHz (& maybe no heatsink)?

BJT Pair for Push-Pull, 1W, 33V, 200mA, 1MHz (& maybe no heatsink)?

Hello, BJT experts.

I would like some suggestions for a good through-hole NPN/PNP pair to use for a push-pull amplifier, with max Vce of about 33V, max Ic of about 200mA, and max Pd of about 1W (AVG Pd is about 600mW). Bandwidth of the amplifier is about 1 MHz. (But also see "possible gotcha", farther below).

I would LIKE to not have to use heatsinks, if possible, if TO-220 devices are used.

Board space will be limited. TO-220 or TO-39/TO-5 or smaller should fit. However, there probably wouldn't be enough space left for a heatsink that extends horizontally, much, in any direction.

At lower power, I was using TO-92 devices (2N4401/4403), with neat little Aavid push-on heatsinks with a slotted vertical fin, Mouser.com # 532-575200B00. For TO-39/TO-5, they also have two different vertical-mount-type heatsinks that look like they would fit on the board, with thermal resistances of 35degC/W and 40degC/W. If I could use TO-220-cased (or similar) devices, I was hoping to be able to avoid using a heatsink, altogether (although some of the smaller "hat section" ones that sit up pretty high on the device might fit).

And, even though the Avg Pd is only about 600mW, the TO-92

2N4401/4403-type devices "apparently" can't be used, even WITH a heatsink, since, on the breadboard, they almost always immediately departed their cases at power-up, even when the avg Pd was only about 500mW and heatsinks were on (And that was before I added boost caps from output to split bias resistor pairs' junctions, which raised Pdavg to about 600mW).

In case anyone needs more details about the amp:

It has +/-17.5v supplies, NPN and PNP with emitters tied together through 1 Ohm resistors to the output, collectors to rails, two 806 Ohm resistors in series from rail to base for each transistor, with 47uF (could be larger) from each 806 Ohm resistor pair's junction to the output, 1N4148 diodes from input to each base (cathodes toward PNP base, mounted in thermal contact with their respective transistors), and two 1N4148 diodes in (anti)parallel from input to output.

The amp is fed by a high-speed opamp (1/2 LT1364) and is inside two of its feedback loops, and can act as either a current source or a voltage source. The load is fed through a 100 Ohm 2.25W resistor (and a 3 Ohm resistor that is part of the opamp feedback loop), when in current-source mode. And there is another 100Ohm/2.25W resistor that can be switched in, from amp output (actually, from between the 3 Ohm and 100 Ohm 2.25W resistors) to GND, for voltage-source mode.

The opamp and feedback parts of the circuit are configured thusly:

Same supplies, two 4.99k R's in series from opamp's + input to gnd, two

4.99k R's in series from opamp's - input to source, 300R and 3R in series from push-pull's output to opamp's - input, 300R and 3R in series from push-pull's output to opamp's + input (Overall output is taken from between THOSE two 300R and 3R resistors.). There is also a 3.3pF cap from opamp's output to opamp's - input. And there is a large resistance (about 3.85Meg) from the negative rail to the junction of the two 4.99k R's that go from opamp's + input to gnd, to try to zero the overall output's offset.

The overall output is taken from between the 300R and 3R that go from push-pull's output to opamp's + input, with 100 Ohms/2.25W always connected in series (from the overall output) with the load. And there's another 100 Ohms/2.25W to GND, with a switch that can optionally connect it to just before the first 100 Ohms/2.25W R, for selecting voltage-source mode instead of current-source mode.

This whole thing might be used to provide base/gate staircase drive signals to a device-under-test, for my curve tracer, with maximum total step voltage and current of 15v and 150 mA (at 1V or 10mA per step), and minimums of 3mV and 30uA (at 200uV or 2uA per step). The input attenuator's output voltage range would be the same (200uV to 1V per step, in 12 selectable "1-2-5" ranges). The overall output-amp voltage gain is 1. And there are buffer opamps between the attenuator and the output-amp's input, one inverting and one non-inverting, with a switch to select either positive or negative staircase output. The steps only transistion to another step at a maximum rate of 22kHz. But the rise-(or fall-)time between steps might be as low as about 1 uS.

POSSIBLE "GOTCHA":

One thing that I am worried about is trying to get the very-low-level (200uV steps) signals through the push-pull stage. It works great in Spice. But, on the breadboard, it sometimes looks like it can't handle anything that's too close to zero. i.e. I think it might sometimes have no output instead of very small output. (I just started the breadboard testing and have to admit that I haven't tested it much with low-level inputs, yet.)

I do realize that there is a "deadzone" problem with push-pull-type topologies. But I guess I DON'T really know how MUCH of a problem that is, or if it's possible (and practical) to "balance" a push-pull stage well-enough to pass a 200uV signal.

So I guess I'm also wondering if I should just "chuck it" and go with some sort of power opamp type of device, instead of the push-pull BJT stage. Essentially, all I really need is to have more current available than a single opamp can provide. I could probably use even more than 150mA, too, for testing devices that need more base current. Maybe 500mA max would be a better goal?

Thanks for any and all suggestions, comments, etc, positive or negative.

Regards,

Tom Gootee

Reply to
tomg
Loading thread data ...

BC440 and 460 is one possibility in TO39. I used these back in 1980 in headphone amplifiers. Ft is 50MHz.

There'll be loads in TO-220 though. The Japanese devices tend to have higher Ft btw. Generally medium power devices from Europe and USA aren't that quick as a rule.

Look at Rohms 2SB1186 and 2SD1763 for example. I've used these in audio driver stages. They have an 80MHz Ft but may be hard to find now. Toshiba will have so many options I'd want to consult the data book.

I expect Zetex might have something suitable too.

Graham

Reply to
Eeyore

Hello Tom,

I used to do that with BD139/BD140. Very inexpensive and in a TO-126 package. You'd have to check how these would fare without heatsinks here, I always bolted them down.

--
Regards, Joerg

http://www.analogconsultants.com
Reply to
Joerg

Seems like the power output stages of every Function generator ever made, use the 2N2219-2N2905 pair. john

Reply to
John Jardine.

John,

Good call!

I had looked at those, first, after no TO-92-cased models seemed to be quite up to the task. And just this morning I tested 2N2219A and

2N2905A (metal can TO-39 cases) in the circuit and.... it looks like no heat sink will be necessary (well, maybe for the 2905A; but only a small one. I'd probably stick one on both, anyway, if there's room.).

I'm guessing that maybe they will also operate down to lower currents than larger devices. Or is that way off?

Thanks!

Tom Gootee

Reply to
tomg

Hi Joerg,

Those look good. And BD139 datasheet says max Pd @ 25c ambient is

1.25W, which should mean no problem without a heatsink. I have ordered some BD139 and BD140 to test them.

Do you know if they would be able to be utilize base currents around 8 to 9 uA, well-enough? Or would smaller devices be better at handling such small currents?

Thanks!

Tom Gootee

Reply to
tomg
[...]

made,

You should have little problem controlling signals down to the prevailing noise. Courtesy of available loop gain and offset voltage of that input opamp. Reading the words I get a gut feeling that the circuit has somehow run out of gain. It's awkward (I'm idle!) to mentally reconstruct your circuit description. Any chance you could put the relevant part of the drawing up somewhere, then maybe all the lads can have a perusal?. john

Reply to
John Jardine.

John,

Well, now that you mention it, there is one small problem that I noticed in Spice simulations: For the maximum step size, the largest step doesn't QUITE make it to its intended value. When it should reach

15v or 150 mA, it only gets to about 14.9V or 149mA.

I had to play with the 806 Ohm split bias resistors' values, to get the max output as high as possible (something like 14.5v out for 15v in), and then had to add the "boost" capacitors from the output to the bias resistors' junctions, to get it to go to 14.9v for 15v in.

If I lowered the emitter resistors' values some more, from, say, 1 Ohm to 0.1 Ohm, then the output would reach 15 v, for 15v in. I'm not sure that I really need them, actually, but didn't want to take the chance of not having them. And 1 Ohm already seemed a little low.

I could also change the opamp's input and feedback resistor ratios. I forget why I didn't want to do that, though.

I will post the schematic at:

formatting link
(95 kbytes)

Thanks again!

Regards,

Tom Gootee

Reply to
tomg

Hello Tom,

I don't know what you mean by that but such large transistors have a very limited beta. Don't expect much at high currents. If you don't have much drive current consider using FETs instead.

--
Regards, Joerg

http://www.analogconsultants.com
Reply to
Joerg

Hello Tom,

Lads, let's not forget the lasses :-)

Ouch, that is low. Make sure D1 and D2 are thermally coupled to their respective transistor or your bias current might run off and turn the whole thing into a puff of smoke.

--
Regards, Joerg

http://www.analogconsultants.com
Reply to
Joerg
[snip]

Tom,

I think I'd avoid the Howland configuration, due to potential instabilities, and make some kind of direct current source with OpAmps and MOSFET's; or even BJT's.

...Jim Thompson

-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona Voice:(480)460-2350 | | | E-mail Address at Website Fax:(480)460-2142 | Brass Rat | |

formatting link
| 1962 | I love to cook with wine. Sometimes I even put it in the food.

Reply to
Jim Thompson

[...]

Tom. Unless I've missed summat, the sim is screaming in my ear that the LT1363 is the problem. That opamp is built for speed and not for precision. Datasheet gain seems a miserly X2000. With low mV staircase input signals, most of the gain will be lost just biasing the output to some arbitrary point. I'd also be worried about it's input bias current of upto 3.6uA and the .6uA offset current. All in all, if it was me I'd just replace the opamp with something like a TL084. Pretty good slew rate and no bias current problems. john

Reply to
John Jardine.

I agree. That design looks very thermally unstable to me.

I'd increase the emitter R value for 'bias current' stability straight away. Quite likely the cause of the problem with exploding 2N4401/3s. In fact I'd make a number of changes.

Graham

Reply to
Eeyore

What's the design load impedance ? 50 ohms ?

There's a number of comments I could make. Are you able to make any serious changes at this time ?

By the way, I remember Motorola's old 1 watt 'extended TO-92' devices that would be a fit for this application too aand probably quite inexpensive.

It's the MPSW variants of the TO-92 MPSA range.

formatting link

Graham

Reply to
Eeyore

then

would

formatting link

Graham,

Thanks, much, for the replies.

YES, I CAN still make serious changes. All ideas are *most* welcome.

As far as the load impedance, it could be virtually anything; whatever the user of the curve tracer decides to connect to the output terminal. But, typically, this (staircase) output would be used to drive the base or gate of a BJT or FET, the other two leads of which would be connected to the curve tracer's sweep-signal terminals (with anywhere from 1v p-p to 30v p-p sawtooth, triangle, or sine driving them). So I've been simulating this with everything from a short circuit for the load (typically in current-source mode) to an open-circuit load (typically in voltage-source mode), as well as various transistors (with the additional sweep circuitry then also used).

Tom

Reply to
tomg

John,

Thanks for the reply.

I agree. I don't know why I left the LT1363 in there. I knew of its "problems". But I had stuck it in there, at some point, while playing with the circuit in LTSpice, and just hadn't yet gotten around to really thinking about it again, yet. I'm sorry about having wasted people's time on that part. TL084 looks pretty good and I have some on hand.

Tom

Reply to
tomg

Jim,

Noted! Thanks.

I have read some negative things about the Howland configuration. It's just that it seemed to work pretty well, in this case, in simulations and on the breadboard. But, I am looking seriously at alternatives. And rest assured that a negative vote from you will carry a lot of weight.

Regards,

Tom Gootee

Reply to
tomg

I'm pondering how to do it best. If I come up with something I'll post.

What sort of range-switching are you doing/want-to-do?

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.
Reply to
Jim Thompson

We are just now designing a gadget that will have a vaguely similar class-AB output stage, with TO-223 bipolars heatsunk to pcb foil pours. One of my guys had the clever idea to replace the current-limit diodes (your D3,D4) with a tiny AC-input optoisolator, so we get about

+-1 volt drop and a free, isolated overload flag.

Our circuit is a power absorber, as opposed to a power source, so things are a bit different. It can swing +-30 on the output with +-7 volt final stage power rails... the customer pulls us up!

John

Reply to
John Larkin

Jim,

Right now, I have a 12-step resistive tapped attenuator to feed the input buffer(s) of the voltage-controlled current/voltage source that we're discussing. [It's built around a small (1/2" diam) NKK pcb-mount rotary switch (1P12T).]

The attenuator's input is 1-volt steps, from 0v to +15v, repeating endlessly to form a "staircase" waveform (with step transitions occurring at selectable frequencies from 60 Hz to 22 kHz).

I currently have the attenuator set up for selectable output voltage steps of 200uV, 500uV, 1mV, 2mV, 5mV, 10mV, 20mV, 50mV, 0.1V, 0.2V,

0.5V, and 1V.

The attenuator's output is then unity-gain-opamp-buffered, inverting and non-inverting, with the output of the desired buffer selected via a switch.

The selected buffer's output voltage is the input to the voltage-controlled current source that we have been discussing.

Currently, I have the v-controlled current/voltage source set up for unity voltage gain, so that it will produce a voltage-step output that's identical to the attenuator output voltages mentioned above.

Between the output of the current/voltage source and the device-under-test (DUT), there has to be a resistor, so I can sense the current into the base or gate of the DUT. I have been using 100 Ohms, which gave fairly-reasonable current ranges for up to 15V, and fairly-reasonable differential voltage levels across it for current sensing (and display on a scope).

[Note that with the Howland current-source setup that I showed, at least, I could switch between current-source and voltage-source modes by switching another 100-Ohm resistor in or out, from the output (before the 100-Ohm current-sense resistor) to ground. Then, WITH the resistor to ground, it became a voltage source.]

Now that I have these "new" lower-voltage calibrated step sizes available (e.g. 200 uV/step), I do have to amplify (x5, IIRC) the

100-Ohm resistor's current-sense voltage, to make it usable with a standard oscilloscope's vertical amplifier and CRT, while still keeping the 15v-max (i.e. 150 mA sensed) range displayable on the scope (since I currently only have a fixed amplfication factor for the current-sense voltage).

I guess that 200 uV (or 2uA) per step may be "pushing it"; possibly too small and/or noisy to be very usable, if amplified x5, displayed as 1mV per step on the scope, using, say, the 2 mV/div vertical scale on the scope, to fit the 15 steps into 8 vertical graticule divisions, while a

15v max current-sense voltage would then still be viewable with 10 V/div vertical scale of the scope (i.e. 75 V max viewed within an 80V "full-scale" 8-div vertical graticule).

It may be better to have the small step-sizes available, and relatively difficult to use, rather than not have them at all. Or maybe not. I could change it. I could also probably have the amplfication factor change for different step sizes. (And with careful ground-return design, etc, etc, noise was actually not too bad, the last time I checked an actual pcb version of something *very* similar.)

Thanks, Jim!

Right now, unfortunately, I have to go drain my car's gas tank (contaminated, somehow). Back later!

Regards,

Tom

Reply to
tomg

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.