Connection from power transistors to printed-circuit board

hi group,

i'm planning to build my own power amps and i've narrowed down my choices to the LM3886 and one of the LMEs (49810 or 49811 or 49830), but i'm very much tempted with the LMEs (main two reasons: higher audio quality and configurable power -- that is, i'm not stuck with 70W; i can choose the right transistors and the right supply voltages and i can build a 50W amp, or 100W, 300W, etc.)

however, i don't have any experience building amps with discrete power transistors, in particular TO-3 transistors needing a heat sink that could make it difficult to solder the transistor pins directly to the printed-circuit board.

my question: what is the recommended practice? send thick wires (say, up to 4 or 5 inches long) from the transistors to the printed-circuit board and nothing else? place small capacitors (say, 1nF or 100pF --- or even less??) in the board or at the transistor pins (or both?) to prevent oscillation? if so, capacitors connected how? base to gnd and emitter to gnd? collector to base and base to emitter? (and collector to base?)

if right at the transistor pins, then how? three capacitors, one between each pair of pins? then send the wires to the printed-circuit board?

anything else in addition to (or instead of?) the capacitors?


or is all this hopeless and the only option is soldering the transistors directly to the printed-circuit board?

any advice or pointers (for example books that cover these practical issues) will be much appreciated!

thanks,
-Zico

 

[Phil Allison]

<zico@mailinator.com>

i'm planning to build my own power amps and i've narrowed down my choices to
the LM3886

** Long established and well proven power amp IC with full SOA protection -
not to mention very low cost.

and one of the LMEs (49810 or 49811 or 49830),


** Fraid that POS has no SOA or VI limiting circuitry.

Means the output devices and probably the fancy IC will be destroyed by the
briefest short on the speaker line.

Really dumb choice.


..... Phil

 

On Thursday, February 20, 2014 11:42:06 PM UTC-5, Phil Allison wrote:

zico

i'm planning to build my own power amps and i've narrowed down my choices to
the LM3886

** Long established and well proven power amp IC with full SOA protection -
not to mention very low cost.

good points.

since i had already almost completely decided in favor of the LMEs, i didn't want to spend much text in the reasons; but one additional disadvantage in my mind for the 3886 is the heat sink. i haven't been able to find any suitable heat sink in my local electronics store or online, other than hacking up a CPU heat skink (as in, a *computer CPU* heat sink), but it sounds all messy. with the LMEs, i can choose among a ton of available TO-3 heat sinks.

and one of the LMEs (49810 or 49811 or 49830),

** Fraid that POS has no SOA or VI limiting circuitry.

a little help here with the acronyms, please? POS i guess stands for "piece of s***t"? SOA i guess "safe operating area"? what about VI?? (just voltage-current?). so i guess in general you're pointing out the lack of built-in circuitry to protect against short and transient overloads?

anyway, three things:

(1) wouldn't a fast-action fuse at the very output stage do just fine? i mean, of course, it is very inconvenient, but still, a short on the output is not part of what one expects to occur frequently, mainly given my second point:

(2) the amps are going to be embedded in the speakers, with its outputs fixed (most likely soldered) to the speakers. in other words, the chances of a short while the amp is operating are virtually null.

(3) what about the shutdown circuitry? as i understand from the specs, it has an internal temperature sensor that shuts down, but also an external connection with which you can shut it down --- would it be a matter of placing a voltage comparator at each of the output feedback resistors (the 0.22 ohms in the suggested diagram in the datasheet) and have that one trigger a, say, 1-second pulse going to the shutdown input pin?

any other reasonably simple ways to add short / overload protection for the LME and for the output transistors?

tnaks,
-Zico

 

[George Herold]

On Thursday, February 20, 2014 8:37:46 PM UTC-5, zi...@mailinator.com wrote:

hi group,



i'm planning to build my own power amps and i've narrowed down my choices to the LM3886 and one of the LMEs (49810 or 49811 or 49830), but i'm very much tempted with the LMEs (main two reasons: higher audio quality and configurable power -- that is, i'm not stuck with 70W; i can choose the right transistors and the right supply voltages and i can build a 50W amp, or 100W, 300W, etc.)



however, i don't have any experience building amps with discrete power transistors, in particular TO-3 transistors needing a heat sink that could make it difficult to solder the transistor pins directly to the printed-circuit board.

You can get connectors for TO-3 packs.. (I entered TO-3 connector into digikey.)

http://www.digikey.com/product-search/en?FV=fff40016%2Cfff8049c&k=TO-3+connector&mnonly=0&newproducts=0&ColumnSort=0&page=1&stock=1&quantity=0&ptm=0&fid=0&pageSize=25

my question: what is the recommended practice? send thick wires (say, up to 4 or 5 inches long) from the transistors to the printed-circuit board and nothing else? place small capacitors (say, 1nF or 100pF --- or even less??) in the board or at the transistor pins (or both?) to prevent oscillation? if so, capacitors connected how? base to gnd and emitter to gnd? collector to base and base to emitter? (and collector to base?)

At audio frequencies a few inches of hook-up wire is fine.

if right at the transistor pins, then how? three capacitors, one between each pair of pins? then send the wires to the printed-circuit board?

I don't know what the capacitors are for, just forget that idea. (IMHO)

anything else in addition to (or instead of?) the capacitors?



or is all this hopeless and the only option is soldering the transistors directly to the printed-circuit board?

any advice or pointers (for example books that cover these practical issues) will be much appreciated!

Art of Electronics is a good book. It has a chapter on various consutrction techniques.

George H.

thanks,

-Zico

 

my mind for the 3886 is the heat sink. i haven't been able to find any
suitable heat sink in my local electronics store or online,

** FFS - you can use just about any flat surface piece of aluminium
extrusion you like.

?? well, i have to confess that the mechanical parts of these projects are the the big nightmare for me. but still, i don't think it is as easy as you suggest. well, or not inexpensive in any case. simple calculations (as per LM3886 datasheet) indicate a 0.5 deg per W heat sink (and that is for just 50W). a quick lookup through digikey.com shows just two suitable units: one at $130 and the other one at $250. the rest are all intel CPU heat skinks between $30 and $40 (which i guess having a perfectly flat side, i could figure out a way to attach the 3886 to them)

the thing is: if i was thinking mass-production, then yes, i would find extrusions (even custom-made ones) from whatever manufacturer at a reasonable price. but as a hobbyist project, i'm finding heat sinks to be the big nightmare of the whole project.

btw, the thermal conductivity of the TO-3s is much better (though i haven't looked up the thermal resistance of the mica film that i need to place to electrically isolate the TO-3 case from the heat sink), plus the max. junction temperature is also better. on the other hand, it's true that for these transistors i'm looking at, each of them is more expensive than the LM3886, and i need two of them in addition to the LME!! so that's a good point to consider.

> ** Huh?

well, for example, through digikey:

selecting all items that contain TO-220 I get 501 results -- the lowest thermal conductivity (wihtout forced convection) is 2.6 degrees per W. if i include every item that contains some TO-2XX, i get two that report 1.1 and 0.9 but that clearly are mistakes (they are very small pieces).

if i select TO-3 i get, well, ok, not a ton, but i do get six items below 1 degree per W. ok, three of them are above $100, but one of them, with .96 deg per W, is just $22.

anyway, i'm probably digressing much more than planned. not sure whether i should just start with the LM3886. still, the advantages of the LMEs seem quite solid.

> TO3 pack transistors are as obsolete as DOS.

?? ok, now you confused me. i see plenty of them available, and i see plenty of heat sinks for TO-3. what makes them obsolete? are there other alternatives that offer better performance or better convenience? (i guess i could partly answer my own question, in that things with the three pins going parallel and thus easy to solder to a printed-circuit board is a lot more convenient than the TO-3 -- precisely the reason that prompted me to post this question here!)

(1) wouldn't a fast-action fuse at the very output stage do just fine?

** No, a dead short would destroy ALL the output devices in 5 mS.

ok (i actually thought so -- i was double checking)

> ** When voice coils burn and fail, they end up going to a near short.

i guess i haven't had many of those; but the ones i've seen blowing up they all end up an open circuit (essentially, the coil acts as a fuse). that's why i wasn't too worried about it.

placing a voltage comparator at each of the output feedback resistors (the
0.22 ohms in the suggested diagram in the datasheet) and have that one
trigger a, say, 1-second pulse going to the shutdown input pin?

** A dead short would destroy ALL the output devices in 5 mS.

so? the comparator would produce a rising edge within a microsecond or maybe even less (no?). is there an additional delay from the shutdown input pin to actual shutdown? or something else that i'm missing?

> You sure YOU can engineer a schemes to stop that ?

well, i'm sure i can try certainly the comparator circuit that produces a pulse in the shutdown input is trivial to design and implement. though granted, originally i wasn't even considering the need for such aspects in the design (i was assuming i'd go with the design as suggested in the LMEs datasheets -- which btw, i haven't entirely discarded this option yet; you're certainly giving some good arguments against it)

YOU cannot even engineer a simple heatsink or see how easy it is to mount
TO3s through a PCB.

as Bob posted, one issue is certainly the lots of empty board space that would be needed. plus the lack of flexibility in terms of locating the heat sinks vs. the location of the printed-circuit board -- i'm worried about this because of the flow of hot air from the heat sink. i would like to have flexibility to locate those so that the heat goes away. i guess i'll continue to think about this.

thanks,
-Zico

 

[Bob Masta]

On Thu, 20 Feb 2014 17:37:46 -0800 (PST),
zico@mailinator.com wrote:

hi group,

i'm planning to build my own power amps and i've narrowed down my choices t=
o the LM3886 and one of the LMEs (49810 or 49811 or 49830), but i'm very mu=
ch tempted with the LMEs (main two reasons: higher audio quality and config=
urable power -- that is, i'm not stuck with 70W; i can choose the right tra=
nsistors and the right supply voltages and i can build a 50W amp, or 100W, =
300W, etc.)

however, i don't have any experience building amps with discrete power tran=
sistors, in particular TO-3 transistors needing a heat sink that could make=
it difficult to solder the transistor pins directly to the printed-circuit=
board.

my question: what is the recommended practice? send thick wires (say, up =
to 4 or 5 inches long) from the transistors to the printed-circuit board an=
d nothing else? place small capacitors (say, 1nF or 100pF --- or even les=
s??) in the board or at the transistor pins (or both?) to prevent oscillati=
on? if so, capacitors connected how? base to gnd and emitter to gnd? col=
lector to base and base to emitter? (and collector to base?)

if right at the transistor pins, then how? three capacitors, one between e=
ach pair of pins? then send the wires to the printed-circuit board?

anything else in addition to (or instead of?) the capacitors?


or is all this hopeless and the only option is soldering the transistors di=
rectly to the printed-circuit board?

any advice or pointers (for example books that cover these practical issues=
) will be much appreciated!

thanks,
-Zico

You can get TO-3 sockets, or just the individual-pin
variants (whatever they're called), but I've usually just
soldered directly to the pins. For the case, a
lockwasher-type solder tab works fine. Don't worry about
short wire runs to the PC board. You shouldn't need any
extra caps.

I've mostly built power amps using only discretes, with the
exception of one design that used a high-voltage Apex TO-3
that I ran on +300 and -15 V for a special job. In that
case, I did include the TO-3 right on the PC layout. The
TO-3 pins were long enough to go through the heat sink and
board and solder normally. Likewise, the TO-3 mounting
screws passed through the heat sink and PC. This type of
mounting means you need a lot of "empty" board space around
the TO-3... no problem in this single-device application,
but in a more-normal amp with multiple devices I'd certainly
have put them off-board.

Best regards,




Bob Masta

DAQARTA v7.50
Data AcQuisition And Real-Time Analysis
www.daqarta.com
Scope, Spectrum, Spectrogram, Sound Level Meter
Frequency Counter, Pitch Track, Pitch-to-MIDI
FREE Signal Generator, DaqMusiq generator
Science with your sound card!

 

[Phil Allison]

<zico@mailinator.com>
Phil Allison wrote:

i'm planning to build my own power amps and i've narrowed down my choices
to
the LM3886

** Long established and well proven power amp IC with full SOA
rotection -
not to mention very low cost.

good points.

since i had already almost completely decided in favor of the LMEs, i didn't
want to spend much text in the reasons; but one additional disadvantage in
my mind for the 3886 is the heat sink. i haven't been able to find any
suitable heat sink in my local electronics store or online,

** FFS - you can use just about any flat surface piece of aluminium
extrusion you like.

with the LMEs, i can choose among a ton of available TO-3 heat sinks.

** Huh?

TO3 pack transistors are as obsolete as DOS.

and one of the LMEs (49810 or 49811 or 49830),

** Fraid that POS has no SOA or VI limiting circuitry.

a little help here with the acronyms, please? POS i guess stands for "piece
of s***t"? SOA i guess "safe operating area"? what about VI?? (just
voltage-current?). so i guess in general you're pointing out the lack of
built-in circuitry to protect against short and transient overloads?

** For all the expensive TO3 darlingtons YOU have to add.


(1) wouldn't a fast-action fuse at the very output stage do just fine?

** No, a dead short would destroy ALL the output devices in 5 mS.


(2) the amps are going to be embedded in the speakers, with its outputs
fixed (most likely soldered) to the speakers. in other words, the chances
of a short while the amp is operating are virtually null.

** When voice coils burn and fail, they end up going to a near short.

But having the amp permanently wired to the load is a big plus for an
unprotected output stage.


(3) what about the shutdown circuitry? as i understand from the specs, it
has an internal temperature sensor that shuts down, but also an external
connection with which you can shut it down --- would it be a matter of
placing a voltage comparator at each of the output feedback resistors (the
0.22 ohms in the suggested diagram in the datasheet) and have that one
trigger a, say, 1-second pulse going to the shutdown input pin?


** A dead short would destroy ALL the output devices in 5 mS.

You sure YOU can engineer a schemes to stop that ?

YOU cannot even engineer a simple heatsink or see how easy it is to mount
TO3s through a PCB.



.... Phil

 

[Phil Hobbs]

On 02/21/2014 10:22 AM, zico@mailinator.com wrote:

my mind for the 3886 is the heat sink. i haven't been able to find
any suitable heat sink in my local electronics store or online,

** FFS - you can use just about any flat surface piece of
aluminium extrusion you like.

?? well, i have to confess that the mechanical parts of these
projects are the the big nightmare for me. but still, i don't think
it is as easy as you suggest. well, or not inexpensive in any case.
simple calculations (as per LM3886 datasheet) indicate a 0.5 deg per
W heat sink (and that is for just 50W). a quick lookup through
digikey.com shows just two suitable units: one at $130 and the other
one at $250. the rest are all intel CPU heat skinks between $30 and
$40 (which i guess having a perfectly flat side, i could figure out a
way to attach the 3886 to them)

the thing is: if i was thinking mass-production, then yes, i would
find extrusions (even custom-made ones) from whatever manufacturer at
a reasonable price. but as a hobbyist project, i'm finding heat
sinks to be the big nightmare of the whole project.

btw, the thermal conductivity of the TO-3s is much better (though i
haven't looked up the thermal resistance of the mica film that i need
to place to electrically isolate the TO-3 case from the heat sink),
plus the max. junction temperature is also better. on the other
hand, it's true that for these transistors i'm looking at, each of
them is more expensive than the LM3886, and i need two of them in
addition to the LME!! so that's a good point to consider.

** Huh?

well, for example, through digikey:

selecting all items that contain TO-220 I get 501 results -- the
lowest thermal conductivity (wihtout forced convection) is 2.6
degrees per W. if i include every item that contains some TO-2XX, i
get two that report 1.1 and 0.9 but that clearly are mistakes (they
are very small pieces).

if i select TO-3 i get, well, ok, not a ton, but i do get six items
below 1 degree per W. ok, three of them are above $100, but one of
them, with .96 deg per W, is just $22.

anyway, i'm probably digressing much more than planned. not sure
whether i should just start with the LM3886. still, the advantages
of the LMEs seem quite solid.

TO3 pack transistors are as obsolete as DOS.

?? ok, now you confused me. i see plenty of them available, and i
see plenty of heat sinks for TO-3. what makes them obsolete? are
there other alternatives that offer better performance or better
convenience? (i guess i could partly answer my own question, in that
things with the three pins going parallel and thus easy to solder to
a printed-circuit board is a lot more convenient than the TO-3 --
precisely the reason that prompted me to post this question here!)

(1) wouldn't a fast-action fuse at the very output stage do just
fine?

** No, a dead short would destroy ALL the output devices in 5 mS.

ok (i actually thought so -- i was double checking)

** When voice coils burn and fail, they end up going to a near
short.

i guess i haven't had many of those; but the ones i've seen blowing
up they all end up an open circuit (essentially, the coil acts as a
fuse). that's why i wasn't too worried about it.

placing a voltage comparator at each of the output feedback
resistors (the 0.22 ohms in the suggested diagram in the datasheet)
and have that one trigger a, say, 1-second pulse going to the
shutdown input pin?

** A dead short would destroy ALL the output devices in 5 mS.

so? the comparator would produce a rising edge within a microsecond
or maybe even less (no?). is there an additional delay from the
shutdown input pin to actual shutdown? or something else that i'm
missing?

You sure YOU can engineer a schemes to stop that ?

well, i'm sure i can try certainly the comparator circuit that
produces a pulse in the shutdown input is trivial to design and
implement. though granted, originally i wasn't even considering the
need for such aspects in the design (i was assuming i'd go with the
design as suggested in the LMEs datasheets -- which btw, i haven't
entirely discarded this option yet; you're certainly giving some good
arguments against it)


YOU cannot even engineer a simple heatsink or see how easy it is to
mount TO3s through a PCB.

as Bob posted, one issue is certainly the lots of empty board space
that would be needed. plus the lack of flexibility in terms of
locating the heat sinks vs. the location of the printed-circuit board
-- i'm worried about this because of the flow of hot air from the
heat sink. i would like to have flexibility to locate those so that
the heat goes away. i guess i'll continue to think about this.

thanks, -Zico

TO3s aren't used nearly as much as they once were, for many of the same
reasons you're finding: they're hard to mount to both heatsink and board
at once. The obvious method is to put the HS on the back of the board,
and run the leads straight through plated-through holes. Unfortunately,
since there's very little give in those thick straight leads, this tends
to lead to cracking the copper or the solder due to temperature cycling.

TO220s aren't as good thermally, but (a) they're much cheaper, so you
can use more of them and spread the heat out, and (b) the leads have a
nice bend to them when you mount them, so they're much more forgiving of
temperature cycling. You do have to be careful about mounting them,
because the single screw tends to make the clamping force very uneven
across the copper plate. This is especially important with newer
TO-220s, which have little wimpy tabs. A spring clamp right on the
epoxy part is generally better than screws.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510

hobbs at electrooptical dot net
http://electrooptical.net

 

[Phil Allison]

<zico@mailinator.com>

my mind for the 3886 is the heat sink. i haven't been able to find any
suitable heat sink in my local electronics store or online,

** FFS - you can use just about any flat surface piece of aluminium
extrusion you like.

?? well, i have to confess that the mechanical parts of these projects are
the the big nightmare for me. but still, i don't think it is as easy as you
suggest. well, or not inexpensive in any case. simple calculations (as per
LM3886 datasheet) indicate a 0.5 deg per W heat sink (and that is for just
50W).

** What utter crap !!

A heatsink with 5 deg C per watt is sufficient.

Look at what others are using you damn fool.

** Huh?

TO3 pack transistors are as obsolete as DOS.

?? ok, now you confused me. i see plenty of them available, and i see
plenty of heat sinks for TO-3. what makes them obsolete?

** Err - large flat pack devices.

Like TO3P and TO247.

(1) wouldn't a fast-action fuse at the very output stage do just fine?

** No, a dead short would destroy ALL the output devices in 5 mS.

** When voice coils burn and fail, they end up going to a near short.

i guess i haven't had many of those; but the ones i've seen blowing up they
all end up an open circuit (essentially, the coil acts as a fuse). that's
why i wasn't too worried about it.


** Then you are a complete fool.

Voice coils are double layer wound, so the first insulation failure is layer
to layer shorting making the impedance drop to half or less. Then the ( now
misshaped) voice coil starts rubbing on the magnet making complete shorts.

** A dead short would destroy ALL the output devices in 5 mS.

You sure YOU can engineer a schemes to stop that ?

well, i'm sure i can try


** Trying is all you are, very trying.

YOU cannot even engineer a simple heatsink or see how easy it is to mount
TO3s through a PCB.

as Bob posted,

** Fuck Bob.

You are a clueless, raving lunatic - piss off.


..... Phil

 

[Phil Allison]

"George Herold"


> At audio frequencies a few inches of hook-up wire is fine.

** Audio frequencies are not the problem - fuckwit.

Somebody should tatoo " Public Menace" on your forehead as a warning.




.... Phil

 

[John Larkin]

On Thu, 20 Feb 2014 17:37:46 -0800 (PST), zico@mailinator.com wrote:

hi group,

i'm planning to build my own power amps and i've narrowed down my choices to the LM3886 and one of the LMEs (49810 or 49811 or 49830), but i'm very much tempted with the LMEs (main two reasons: higher audio quality and configurable power -- that is, i'm not stuck with 70W; i can choose the right transistors and the right supply voltages and i can build a 50W amp, or 100W, 300W, etc.)

however, i don't have any experience building amps with discrete power transistors, in particular TO-3 transistors needing a heat sink that could make it difficult to solder the transistor pins directly to the printed-circuit board.

my question: what is the recommended practice? send thick wires (say, up to 4 or 5 inches long) from the transistors to the printed-circuit board and nothing else? place small capacitors (say, 1nF or 100pF --- or even less??) in the board or at the transistor pins (or both?) to prevent oscillation? if so, capacitors connected how? base to gnd and emitter to gnd? collector to base and base to emitter? (and collector to base?)

if right at the transistor pins, then how? three capacitors, one between each pair of pins? then send the wires to the printed-circuit board?

anything else in addition to (or instead of?) the capacitors?


or is all this hopeless and the only option is soldering the transistors directly to the printed-circuit board?

any advice or pointers (for example books that cover these practical issues) will be much appreciated!

thanks,
-Zico

Why use ancient TO3s? TO247 types are cheaper and better thermally.

You can do something like this:

https://dl.dropboxusercontent.com/u/53724080/Thermal/Amp.jpg

or

https://dl.dropboxusercontent.com/u/53724080/Thermal/L650_HS_proto.JPG

and avoid all those wires.




--

John Larkin Highland Technology, Inc

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

 

[Phil Allison]

"Bob Masta"

** FYI:

here are two examples of HOW to mount TO3s with no wires.

http://www.halfgaar.net/media/amplifier-module

http://www.jaycar.com.au/products_uploaded/productLarge_13915.jpg

If you mount the PCB on the flat side of a heasink ( see link) with copper
side down, pass the B and E pins through holes in the board and solder them
as the final step in assembly, you lose no PCB space at all.

http://www.voti.nl/common/heatsink-07.jpg



..... Phil