TO-92 vs sot-23 transistors for power

On Wednesday, January 27, 2016 at 11:19:32 AM UTC-5, Winfield Hill wrote:
> George Herold wrote...
> >
> >>  One of my design interests is high-frequency,
> >>  high-voltage power amplifiers.  For example,
> >>  I'm doing a 45V, 5A peak, DC-10MHz amplifier.
> >>  My design uses 20 TO-220F power transistors,
> >>  mounted on a 200W fan-cooled heat-sink plate.
> >>  They're mounted under the PCB in this photo.
> >>  https://www.dropbox.com/s/k4k14zkaq0ewxa9/AMP-70A-2_proto_01.jpg
> >>  https://www.dropbox.com/s/u6di0gb4xgvj3xv/AMP-70A_3D.jpg
> >
> > I spy a bunch of carbon comp resistors.  (Can you say why?)
> 
>  Low inductance and transient power-handling.
> 
> > I assume the amp is class A?  
> 
>  Yes.  Here's a schematic of a slowish version.
>  https://www.dropbox.com/s/xwy4l5audcozylw/AMP-70A-2_sch_.pdf
> 

I sorta lost the signal path in the middle, OK I found the feedback.
But the output!
I was thinking the 3.3 ohm balancing R's were big,
and then I saw 50 ohms on the output. 
Doesn't that suck up all your voltage?  
What's the load?  Driving a coil? 
(is there a big capacitance path I'm missing?)

Re thermal stuff, I was going to suggest little
brass or Al standoffs from the pcb "thermal" plane 
to the heat sink below.  You always end up 
needing electrical isolation... how much pcb
overlap do you need between power and ground 
to have one thermal ohm.  (a thermal ohm is 
is one degree K/W.)  It's ok I can work out the 
answer.    

George H.    
> 
> -- 
>  Thanks,
>     - Win

On Wednesday, January 27, 2016 at 7:05:36 AM UTC-8, Winfield Hill wrote:
> Q.  Relative power dissipation capability
>  of TO-92 transistor packages (viz, Zetex
>  E-line parts) vs. sot-23 packages.

Usual TO-92 construction has a bent C tab, to
which the die mounts; that's lots worse than a  power package,
but it does mean the C lead wire carries significant heat, and can be sinked.
It also  means that the flat face of the TO-92 is less effective at
removing heat than the curved back (the C tab bends away from the flat face).
I've seen wrap-around TO-92 heatsinks, as well as epoxy-into-an-aluminum-block
(but that was for a log amp matching issue)

>  ... I'm worried about the viability of
>  crowding a bunch of power-dissipating SOT-23
>  and 1210 parts together.  They transfer heat
>  to the PCB, unlike through-hole parts, which
>  are better at getting their heat into the air.

One could line up SOT transistors, put a strip of silicone pad
atop 'em, and clamp an aluminum L-member, one leg to a heatsink
and the other leg atop the transistors.   Heat path through the epoxy is 
probably better (shorter path) than TO-92.

John Larkin wrote..
> Winfield Hill wrote:
>> John Larkin wrote...
>>> Tim Wescott wrote:
>>>
>>>> Hmm.  A SOT-23 with a big collector tab would be nice here.
>>>>
>>>> Look at their SOT-89 parts.  ...
>>>
>>> I decided that I could dissipate 3 watts from
>>> one of those Supertex SOT89 depletion fets.
>>
>> FETs, yes.  Some sot-23 MOSFETs do well,
>> rated over 1 watt.
>
> Bipolar chips are too damned small. 
> Which is why C is low.

 FETs are so damned large, C is high.


-- 
 Thanks,
    - Win

On Wednesday, January 27, 2016 at 4:44:32 PM UTC-5, Winfield Hill wrote:
> John Larkin wrote...
> > Robert Latest wrote:
> >> Winfield Hill wrote:
> >>
> >>>  I mentioned before my worry about replacing
> >>>  a set of TO-92 transistors with sot-23 parts.
> >>
> >> What you didn't mention, and what so far
> >> nobody asked, is why you want to go SMD
> >> at all. I don't see potential for a
> >> significant reduction of PCB footprint
> >> if that's what you're after.
> 
>  SMD resistors take far less space than
>  1/4-watt axial-lead parts.  But mixing
>  0805 in among TO-92 makes hand assembly
>  a pain.  Solder the 0805s first, yes.
>  But when coming back to change the value,
>  it's hard to get to with the iron tip.
> 
> 
> -- 
>  Thanks,
>     - Win

I bend the 1/4 W TH R's into a hair pins and stick
'em on a 0.1" grid.  That gets it tight... .
Yeah your low inductance R's want to sit flat and straight. 

I must admit I feel pretty silly telling you 
how I do circuit layout.  

For the transistors in the air, one thing to think 
about is where they are in the thermal path, if they 
are in contact w/ cool moving air.. good. 
other wise maybe it's best to get them in touch
 with something in contact with the case.   

It's cool what legg said about the to-93 cases!  
I've got a to-93 that gets a little hot and 
noisier under maximum load*, and I've often 
thought about one of those heat sink top hats.  

Oh and thanks for the circuit diagram!
 
George H. 
*it's in a cap multiplier and max load is also max thermal load.

George Herold wrote...
> Winfield Hill wrote:
>> George Herold wrote...
>>>
>>> I spy a bunch of carbon comp resistors.
>>> (Can you say why?)
>>
>> Low inductance and transient power-handling.
>> 
>>> I assume the amp is class A?  
>> 
>> Yes.  Here's a schematic of a slowish version.
>> https://www.dropbox.com/s/xwy4l5audcozylw/AMP-70A-2_sch_.pdf
>
> I sorta lost the signal path in the middle,
> OK I found the feedback.  But the output!
> I was thinking the 3.3 ohm balancing R's
> were big, and then I saw 50 ohms on the
> output.  Doesn't that suck up all your
> voltage?  What's the load?  Driving a coil? 
> (is there a big capacitance path I'm missing?)

 Traditionally signal generators have 50-ohm
 outputs to match the coax they'll be driving.
 They have a 50-ohm source resistor to insure
 proper coax-transmission-cable driving. 

 {Please see AoE III, 12.10.1, pages 858-874,
 and Appendix H, pages 1116 to 1130.

 Most signal generators are limited to 10 or
 20V to drive their 50-ohm output resistors.
 This voltage is cut in half at the load, if
 there's the expected nominal 50-ohm load.

 In addition to the traditional 50-ohm output
 the AMP-70 has a low-Z output, 1-ohm, meant
 to drive some RF 4x step-up transformers that
 create higher voltages, RIS-

 For the 1-ohm output, with up to 5A load, the
 drop across each resistor could be 3.3V, so
 the resistor value seems a bit high.  It was
 picked to avoid excessive quiescent current,
 whatever the output transistor's temp might be.

> Re thermal stuff, I was going to suggest little
> brass or Al standoffs from the pcb "thermal"
> plane to the heat sink below.  You always end
> up needing electrical isolation... how much pcb
> overlap do you need between power and ground 
> to have one thermal ohm.  (A thermal ohm is 
> is one degree K/W.)  It's OK I can work out
> the answer.    

 Very interesting, I'll think about that.  One
 point: wrt this amplifier, the only candidate
 transistors I've been able to find are TO-220F
 types, with thick plastic insulation.


-- 
 Thanks,
    - Win

legg wrote...
> Winfield Hill wrote:
>> Winfield Hill wrote...
>>>
>>> Q.  Relative power dissipation capability
>>> of TO-92 transistor packages (viz, Zetex
>>> E-line parts) vs. sot-23 packages.  More on
>>> this in my next post, first some background.
>>    [ snip ]
>>
>> I mentioned before my worry about replacing
>> a set of TO-92 transistors with sot-23 parts.
>> Not only do the sot-23 parts usually have a
>> lower Pdiss rating, they dissipate their heat
>> into the PCB (unlike TO-92 into air**), so
>> a set of them in close proximity would raise
>> the PCB ambient, further reducing their real
>> power-dissipation handling capability.
>>
>> Specifically I've been using Zetex' ZTX450
>> and '550 E-line complementary transistors,
>> which are rated at 1-watt for Tamb=25�C. 
>> Their corresponding sot-23 types might be
>> FMMT619, '624, or '722, all rated at 625mW.
>>
>> The datasheets have a note: For a device
>> surface mounted on 25mm X 25mm FR4 PCB with
>> high coverage of single sided 1 oz copper.
>
> The Zetex E-Line parts did not use conventional epoxy for their
> encapsulant, and derated based on 200degC junction temperature to
> achieve the 1W rating in free air. Rthjc 175degC/W is achieved through
> actual package material superior thermal conductivity. These parts can
> be profitably heatsunk.
>
> This is one reason why the generic substitutes in conventional epoxy
> have a lower free air listing, as derated based on 150degC Tj max at
> 200degC/W or ~625mW.
>
> The SOT23 or SOT223 bodies are similarly restricted to the same
> 150degC Tj max by the epoxy encapsulant.
>
> In an unimproved redesign, your first questions should be what
> junction temperatures are actually present in the current product, and
> what power dissipation and ambient temperatures are currently handled.
>
> A reduced Tj max means more area will be required for the same
> reliability margins, regardless of package size changes.
>
> RL

 Legg, I really appreciate your remarks in this case.


-- 
 Thanks,
    - Win

On Wednesday, January 27, 2016 at 8:19:30 PM UTC-5, Winfield Hill wrote:
> George Herold wrote...
> > Winfield Hill wrote:
> >> George Herold wrote...
> >>>
> >>> I spy a bunch of carbon comp resistors.
> >>> (Can you say why?)
> >>
> >> Low inductance and transient power-handling.
> >> 
> >>> I assume the amp is class A?  
> >> 
> >> Yes.  Here's a schematic of a slowish version.
> >> https://www.dropbox.com/s/xwy4l5audcozylw/AMP-70A-2_sch_.pdf
> >
> > I sorta lost the signal path in the middle,
> > OK I found the feedback.  But the output!
> > I was thinking the 3.3 ohm balancing R's
> > were big, and then I saw 50 ohms on the
> > output.  Doesn't that suck up all your
> > voltage?  What's the load?  Driving a coil? 
> > (is there a big capacitance path I'm missing?)
> 
>  Traditionally signal generators have 50-ohm
>  outputs to match the coax they'll be driving.
>  They have a 50-ohm source resistor to insure
>  proper coax-transmission-cable driving. 
> 
>  {Please see AoE III, 12.10.1, pages 858-874,
>  and Appendix H, pages 1116 to 1130.
Hah, I've been reading that section,  I know 
about sticking 50 ohms in the source line 
for matching coax... I didn't know about the 
"echo" in the source end shown in fig 12.115.  

I guess I was confused by the 
5A/ 45 V and 50 ohms.  

George H. 
> 
>  Most signal generators are limited to 10 or
>  20V to drive their 50-ohm output resistors.
>  This voltage is cut in half at the load, if
>  there's the expected nominal 50-ohm load.
> 
>  In addition to the traditional 50-ohm output
>  the AMP-70 has a low-Z output, 1-ohm, meant
>  to drive some RF 4x step-up transformers that
>  create higher voltages, RIS-
> 
>  For the 1-ohm output, with up to 5A load, the
>  drop across each resistor could be 3.3V, so
>  the resistor value seems a bit high.  It was
>  picked to avoid excessive quiescent current,
>  whatever the output transistor's temp might be.
> 
> > Re thermal stuff, I was going to suggest little
> > brass or Al standoffs from the pcb "thermal"
> > plane to the heat sink below.  You always end
> > up needing electrical isolation... how much pcb
> > overlap do you need between power and ground 
> > to have one thermal ohm.  (A thermal ohm is 
> > is one degree K/W.)  It's OK I can work out
> > the answer.    
> 
>  Very interesting, I'll think about that.  One
>  point: wrt this amplifier, the only candidate
>  transistors I've been able to find are TO-220F
>  types, with thick plastic insulation.
> 
> 
> -- 
>  Thanks,
>     - Win

In article <n8ambh015ao@drn.newsguy.com>,
 Winfield Hill <hill@rowland.harvard.edu> wrote:

>  Q.  Relative power dissipation capability
>  of TO-92 transistor packages (viz, Zetex
>  E-line parts) vs. sot-23 packages.  More on
>  this in my next post, first some background. 
> 
>  One of my design interests is high-frequency,
>  high-voltage power amplifiers.  For example,
>  I'm doing a 45V, 5A peak, DC-10MHz amplifier.
>  My design uses 20 TO-220F power transistors,
>  mounted on a 200W fan-cooled heat-sink plate.
>  They're mounted under the PCB in this photo.
>  https://www.dropbox.com/s/k4k14zkaq0ewxa9/AMP-70A-2_proto_01.jpg
>  https://www.dropbox.com/s/u6di0gb4xgvj3xv/AMP-70A_3D.jpg
> 
>  The input-stage driving circuitry uses 10W,
>  a significant part of which is dissipated by
>  ten TO-92 E-line transistors, lower left.
> 
>  Many of the parts are large: TO-220 trannies,
>  low-inductance power resistors, electrolytics,
>  converter modules, etc., and not well suited
>  to surface-mount versions.  So I decided to
>  forego most SMT parts on the layout, but I'm
>  rethinking that decision for my next pass.
> 
>  However I'm worried about the viability of
>  crowding a bunch of power-dissipating SOT-23
>  and 1210 parts together.  They transfer heat
>  to the PCB, unlike through-hole parts, which
>  are better at getting their heat into the air.
>  So I'm worried about a hotspot and a reduced
>  dissipation capability from datasheet values.
> 
>  Please reply here if you want to discuss the
>  amplifier, otherwise advance to my next post
>  to discuss the TO-92 vs SOT-23 issues.

I don't do professional work but...

It looks I see light coming through your PCB.  SMD devices will be 
severely de-rated if there's no copper plane behind them.

Power SMDs would be a good place to get real numbers for various board 
designs.  TI refers to document SPRA953 for using their flip chip 
regulators.  Lumileds and CREE have thermal calculations in their 
application design notes.

-- 
I will not see posts from astraweb, theremailer, dizum, or google
because they host Usenet flooders.

On a sunny day (27 Jan 2016 11:46:59 -0800) it happened Winfield Hill
<hill@rowland.harvard.edu> wrote in <n8b6rj020qo@drn.newsguy.com>:

>>> Yes.  Here's a schematic of a slowish version.
>>> https://www.dropbox.com/s/xwy4l5audcozylw/AMP-70A-2_sch_.pdf

>> Could that not be done like the old audio amp a
>> lot simpler?
>> Or is there a reason for all the
>> complicatiatitiated stuff?
>
> Some is to keep the TO-92 dissipation under
> control, other stuff is for thermal matching.
> In the next pass, add protective circuitry.

I understand your quest for fast slew rate,
but take for example the bias generator Q20 and Q24 in series.
If is just mirroring right?
ONE transistor with different resistors as zener would do just as well?
And if you are so worried about slew rate why not decouple it?
The rest is the same carbon copy mirror joke excuse my opinion.

FWIW, I've got a cutesy test amp here -- 2N7002 cascode. Built just to see 
how well it works.

http://seventransistorlabs.com/Images/Wideband2N7002.jpg

I won't bother with a schematic as you can read the PCB pretty well, I'd 
guess... anyway, you can easily see the pad size for Q1 and Q2 (they are 
as you see, no vias or buried stuff).  At 100mA (about 4Vds per 
transistor, so 400mW), they get rather toasty, enough that I don't think I 
would approve of putting this amplifier inside a hot box with other stuff.

So for safe purposes, with minimal footprints, I think I'd want to limit 
SOT-23s to 200mW or so, with the option for higher, but only with serious 
consideration of the thermal environment (i.e., how much trace/pour area 
can you *really* spare; how about vias?).

As others have mentioned, SOT-89s are pretty darn good for small power 
SMT.

Tim

-- 
Seven Transistor Labs, LLC
Electrical Engineering Consultation and Contract Design
Website: http://seventransistorlabs.com

"Winfield Hill"  wrote in message news:n8andt017fk@drn.newsguy.com...

Winfield Hill wrote...
>
> Q.  Relative power dissipation capability
> of TO-92 transistor packages (viz, Zetex
> E-line parts) vs. sot-23 packages.  More on
> this in my next post, first some background.
    [ snip ]

I mentioned before my worry about replacing
a set of TO-92 transistors with sot-23 parts.
Not only do the sot-23 parts usually have a
lower Pdiss rating, they dissipate their heat
into the PCB (unlike TO-92 into air**), so
a set of them in close proximity would raise
the PCB ambient, further reducing their real
power-dissipation handling capability.

Specifically I've been using Zetex' ZTX450
and '550 E-line complementary transistors,
which are rated at 1-watt for Tamb=25�C.
Their corresponding sot-23 types might be
FMMT619, '624, or '722, all rated at 625mW.

The datasheets have a note: For a device
surface mounted on 25mm X 25mm FR4 PCB with
high coverage of single sided 1 oz copper.

Can anyone give me some guidance on further
derating for closely-spaced sot-23 parts?

BTW, going to packages like sot-223 could
mean an even larger PCB footprint.

** I know that TO-92 transistors dissipate
some heat into the PCB, so I'm limiting
their dissipation to under 500mW.


-- 
Thanks,
    - Win