Muliple Peltier elements: parallel or series?

Hello,

I'm designing a 'thermo table', consisting of a 150 x 150 x 4 mm aluminium
surface (6" x 6" x 1/6"), which has to be kept at a precise temperature
throughout (+/- 0.1 degree centigrade). Since both heating and cooling are
required, the whole design is based on one or more Peltier elements.

Ideally, one would use a Peltier element of comparable size as the aluminium
surface, to minimize temperature differences due to small local heat
losses. However, for reasons of cost and available electrical power (100W
max), I'm limited to commonly available Peltier elements with a 40 x 40 or
50 x 50 mm surface area.

So I thought I'd use four identical 25W Peltier elements (40 x 40 mm)
instead of one 100W Peltier element (50 x 50 mm). My question: is it better
to connect these Peltier elements in parallel or in series? In other words:
is a Peltier element's heat transport mainly determined by voltage
difference or by current? Any ideas are appreciated.

And oh, I thought about using much thicker aluminium to minimize local
temperature differences, but with the above dimensions, I already have a
heat capacity of > 200 J/C, so with 100W power it already takes at least 2
seconds to heat it up by one degree.
Doubling the thickness also doubles this time, and that quickly becomes a
problem: users will want to change the temperature often, and don't want to
wait several minutes every time until the temperature settles to its final
value.

Thanks in advance for any ideas.

Richard Rasker
-- 
http://www.linetec.nl

On 19/07/2012 12:57, Richard Rasker wrote:
> Hello,
>
> I'm designing a 'thermo table', consisting of a 150 x 150 x 4 mm aluminium
> surface (6" x 6" x 1/6"), which has to be kept at a precise temperature
> throughout (+/- 0.1 degree centigrade). Since both heating and cooling are
> required, the whole design is based on one or more Peltier elements.
>
> Ideally, one would use a Peltier element of comparable size as the aluminium
> surface, to minimize temperature differences due to small local heat
> losses. However, for reasons of cost and available electrical power (100W
> max), I'm limited to commonly available Peltier elements with a 40 x 40 or
> 50 x 50 mm surface area.
>
> So I thought I'd use four identical 25W Peltier elements (40 x 40 mm)
> instead of one 100W Peltier element (50 x 50 mm). My question: is it better
> to connect these Peltier elements in parallel or in series? In other words:
> is a Peltier element's heat transport mainly determined by voltage
> difference or by current? Any ideas are appreciated.

Current. They are in essence a bunch of high power diode PN junctions.

Their efficiency as a cooler isn't all that great so insulation is 
important if you are to obtain good thermal stability as is a suitable 
heatsink on the dump side. Good for having no moving parts thought.

-- 
Regards,
Martin Brown

Martin Brown wrote:
> 
> On 19/07/2012 12:57, Richard Rasker wrote:
> > Hello,
> >
> > I'm designing a 'thermo table', consisting of a 150 x 150 x 4 mm aluminium
> > surface (6" x 6" x 1/6"), which has to be kept at a precise temperature
> > throughout (+/- 0.1 degree centigrade). Since both heating and cooling are
> > required, the whole design is based on one or more Peltier elements.
> >
> > Ideally, one would use a Peltier element of comparable size as the aluminium
> > surface, to minimize temperature differences due to small local heat
> > losses. However, for reasons of cost and available electrical power (100W
> > max), I'm limited to commonly available Peltier elements with a 40 x 40 or
> > 50 x 50 mm surface area.
> >
> > So I thought I'd use four identical 25W Peltier elements (40 x 40 mm)
> > instead of one 100W Peltier element (50 x 50 mm). My question: is it better
> > to connect these Peltier elements in parallel or in series? In other words:
> > is a Peltier element's heat transport mainly determined by voltage
> > difference or by current? Any ideas are appreciated.
> 
> Current. They are in essence a bunch of high power diode PN junctions.
> 
> Their efficiency as a cooler isn't all that great so insulation is
> important if you are to obtain good thermal stability as is a suitable
> heatsink on the dump side. Good for having no moving parts thought.
> 
> --
> Regards,
> Martin Brown

I'd control them individually, or wire them in series.  Compound
Peltiers have a nasty thermal instability, where they start getting
hotter as you go to higher current instead of colder.  

If you put them in series, they all see the same current, which along
with a spreader plate top and bottom) helps prevent runaway.  If you put
them in parallel, the ones with higher delta-T will draw less current
than the lower delta-T ones, which gives you a nice stabilizing action
at low drive current.  

If you push them, or you lose cooling water to your heat exchanger, the
sign of the gain can invert and the whole thing turns to lava very
fast.  (That can happen with individual control or series-connection as
well, but the melty ones can't hog current from the cooler ones, so it
isn't so unstable.)

The other thing is that you have to allow the Peltiers to slide around a
bit as the cold plate cools down, because otherwise they'll crack.  So
use Arctic Silver or a very small amount of very good thermal grease on
one side, and solder on the other.  IIRC you can solder the 30 mm ones,
but not the 50 mm ones--it'll be in the datasheet. (Solder is about 100x
better than ordinary thermal grease.)

I talk a lot about Peltiers in my free thermal chapter,
http://electrooptical.net/www/book/thermal.pdf .  There's also an
incomplete draft that was intended to accompany the second edition,
which has some more stuff in it, at
http://electrooptical.net/www/book/draftthermal.pdf .

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
845-480-2058

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

Phil Hobbs wrote:
> 
> Martin Brown wrote:
> >
> > On 19/07/2012 12:57, Richard Rasker wrote:
> > > Hello,
> > >
> > > I'm designing a 'thermo table', consisting of a 150 x 150 x 4 mm aluminium
> > > surface (6" x 6" x 1/6"), which has to be kept at a precise temperature
> > > throughout (+/- 0.1 degree centigrade). Since both heating and cooling are
> > > required, the whole design is based on one or more Peltier elements.
> > >
> > > Ideally, one would use a Peltier element of comparable size as the aluminium
> > > surface, to minimize temperature differences due to small local heat
> > > losses. However, for reasons of cost and available electrical power (100W
> > > max), I'm limited to commonly available Peltier elements with a 40 x 40 or
> > > 50 x 50 mm surface area.
> > >
> > > So I thought I'd use four identical 25W Peltier elements (40 x 40 mm)
> > > instead of one 100W Peltier element (50 x 50 mm). My question: is it better
> > > to connect these Peltier elements in parallel or in series? In other words:
> > > is a Peltier element's heat transport mainly determined by voltage
> > > difference or by current? Any ideas are appreciated.
> >
> > Current. They are in essence a bunch of high power diode PN junctions.
> >
> > Their efficiency as a cooler isn't all that great so insulation is
> > important if you are to obtain good thermal stability as is a suitable
> > heatsink on the dump side. Good for having no moving parts thought.
> >
> > --
> > Regards,
> > Martin Brown
> 
> I'd control them individually, or wire them in series.  Compound
> Peltiers have a nasty thermal instability, where they start getting
> hotter as you go to higher current instead of colder.
> 
> If you put them in series, they all see the same current, which along
> with a spreader plate top and bottom) helps prevent runaway.  If you put
> them in parallel, the ones with higher delta-T will draw less current
> than the lower delta-T ones, which gives you a nice stabilizing action
> at low drive current.
> 
> If you push them, or you lose cooling water to your heat exchanger, the
> sign of the gain can invert and the whole thing turns to lava very
> fast.  (That can happen with individual control or series-connection as
> well, but the melty ones can't hog current from the cooler ones, so it
> isn't so unstable.)
> 
> The other thing is that you have to allow the Peltiers to slide around a
> bit as the cold plate cools down, because otherwise they'll crack.  So
> use Arctic Silver or a very small amount of very good thermal grease on
> one side, and solder on the other.  IIRC you can solder the 30 mm ones,
> but not the 50 mm ones--it'll be in the datasheet. (Solder is about 100x
> better than ordinary thermal grease.)
> 
> I talk a lot about Peltiers in my free thermal chapter,
> http://electrooptical.net/www/book/thermal.pdf .  There's also an
> incomplete draft that was intended to accompany the second edition,
> which has some more stuff in it, at
> http://electrooptical.net/www/book/draftthermal.pdf .
> 

One other thing about Peltiers that's often overlooked: compared with
heaters, they can make the system far less vulnerable to thermal
forcing.  With just a heater, you have to have a big heat leak in order
for the loop to work at any reasonable speed, so the bandwidth and the
thermal forcing go up together.  With a Peltier you don't have that
constraint, because it can pull as well as push.

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
845-480-2058

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

On 19/07/2012 14:29, Phil Hobbs wrote:
> Martin Brown wrote:

>> Current. They are in essence a bunch of high power diode PN junctions.
>
> I'd control them individually, or wire them in series.  Compound
> Peltiers have a nasty thermal instability, where they start getting
> hotter as you go to higher current instead of colder.
>
> If you put them in series, they all see the same current, which along
> with a spreader plate top and bottom) helps prevent runaway.  If you put
> them in parallel, the ones with higher delta-T will draw less current
> than the lower delta-T ones, which gives you a nice stabilizing action
> at low drive current.
>
> If you push them, or you lose cooling water to your heat exchanger, the
> sign of the gain can invert and the whole thing turns to lava very
> fast.  (That can happen with individual control or series-connection as
> well, but the melty ones can't hog current from the cooler ones, so it
> isn't so unstable.)
>
> The other thing is that you have to allow the Peltiers to slide around a
> bit as the cold plate cools down, because otherwise they'll crack.  So
> use Arctic Silver or a very small amount of very good thermal grease on
> one side, and solder on the other.  IIRC you can solder the 30 mm ones,
> but not the 50 mm ones--it'll be in the datasheet. (Solder is about 100x
> better than ordinary thermal grease.)
>
> I talk a lot about Peltiers in my free thermal chapter,
> http://electrooptical.net/www/book/thermal.pdf .  There's also an
> incomplete draft that was intended to accompany the second edition,
> which has some more stuff in it, at
> http://electrooptical.net/www/book/draftthermal.pdf .

Handy info thanks. Although the TECs are available their datasheets are 
often inadequate and miss out details needed for real world designs.

Can I pick your brains for a slightly odd requirement? Ideally one that 
can be done with at most a pair of TECs one high temp and one normal. On 
paper they were rated for 3A and 14v or so each but in practice I was 
struggling in reverse to get 2v and a few mA out with a candle flame 
heated plate at 250C on one side and a slab of aluminium at -18C on the 
other. I gave up in the end as it was taking too long and safety 
concerns of very hot metal plates and children scuppered it in the end.

The aim would be for an Xmas science demo to harvest some of the ~100W 
waste heat of a candle flame and drive a ~1W LED. This seemed a modest 
aim when I started out but in practice it proved impossible. To be any 
good the LED must be a *lot* brighter than the candle flame!

I am guessing that to stand any chance I also need a fan assisted heat 
sink and to be authentic it must all be powered by the TECs. It is for a 
physics demo so hiding a button cell somewhere is just not on...

-- 
Regards,
Martin Brown

On Jul 19, 7:57=A0am, Richard Rasker <spamt...@linetec.nl> wrote:
> Hello,
>
> I'm designing a 'thermo table', consisting of a 150 x 150 x 4 mm aluminiu=
m
> surface (6" x 6" x 1/6"), which has to be kept at a precise temperature
> throughout (+/- 0.1 degree centigrade). Since both heating and cooling ar=
e
> required, the whole design is based on one or more Peltier elements.
>
> Ideally, one would use a Peltier element of comparable size as the alumin=
ium
> surface, to minimize temperature differences due to small local heat
> losses. However, for reasons of cost and available electrical power (100W
> max), I'm limited to commonly available Peltier elements with a 40 x 40 o=
r
> 50 x 50 mm surface area.
>
> So I thought I'd use four identical 25W Peltier elements (40 x 40 mm)
> instead of one 100W Peltier element (50 x 50 mm). My question: is it bett=
er
> to connect these Peltier elements in parallel or in series? In other word=
s:
> is a Peltier element's heat transport mainly determined by voltage
> difference or by current? Any ideas are appreciated.
>
> And oh, I thought about using much thicker aluminium to minimize local
> temperature differences, but with the above dimensions, I already have a
> heat capacity of > 200 J/C, so with 100W power it already takes at least =
2
> seconds to heat it up by one degree.
> Doubling the thickness also doubles this time, and that quickly becomes a
> problem: users will want to change the temperature often, and don't want =
to
> wait several minutes every time until the temperature settles to its fina=
l
> value.
>
> Thanks in advance for any ideas.
>
> Richard Rasker
> --http://www.linetec.nl

TEC's in series as other have suggested.  What's the temperature
range?  And most important what's on the other side of the TEC's?  In
my limited experience you need to spend at least as much time thinking
about how you keep the temperature of the 'other side' of the TEC
roughly constant.

George H.

On Jul 19, 4:57=A0am, Richard Rasker <spamt...@linetec.nl> wrote:
> Hello,
>
> I'm designing a 'thermo table', consisting of a 150 x 150 x 4 mm aluminiu=
m
> surface (6" x 6" x 1/6"), which has to be kept at a precise temperature
> throughout (+/- 0.1 degree centigrade). Since both heating and cooling ar=
e
> required, the whole design is based on one or more Peltier elements.
>
> Ideally, one would use a Peltier element of comparable size as the alumin=
ium
> surface, to minimize temperature differences due to small local heat
> losses. However, for reasons of cost and available electrical power (100W
> max), I'm limited to commonly available Peltier elements with a 40 x 40 o=
r
> 50 x 50 mm surface area.
>
> So I thought I'd use four identical 25W Peltier elements (40 x 40 mm)
> instead of one 100W Peltier element (50 x 50 mm). My question: is it bett=
er
> to connect these Peltier elements in parallel or in series? In other word=
s:
> is a Peltier element's heat transport mainly determined by voltage
> difference or by current? Any ideas are appreciated.
>
> And oh, I thought about using much thicker aluminium to minimize local
> temperature differences, but with the above dimensions, I already have a
> heat capacity of > 200 J/C, so with 100W power it already takes at least =
2
> seconds to heat it up by one degree.
> Doubling the thickness also doubles this time, and that quickly becomes a
> problem: users will want to change the temperature often, and don't want =
to
> wait several minutes every time until the temperature settles to its fina=
l
> value.
>
> Thanks in advance for any ideas.
>
> Richard Rasker
> --http://www.linetec.nl

If you NEED the surface to be at a uniform temperature, I'd look into
getting a heat pipe made up for it, or at least talk to those people.

Can't find their contact info. But, at trade shows they hand out
little 'hockey stick' coppercolored heat pipes as promos. The surface
temperature of those sticks is uniform, fast.

Probably sell to multicore processor cooling too. transfer heat out to
fan area.

Martin Brown wrote:
> 
> On 19/07/2012 14:29, Phil Hobbs wrote:
> > Martin Brown wrote:
> 
> >> Current. They are in essence a bunch of high power diode PN junctions.
> >
> > I'd control them individually, or wire them in series.  Compound
> > Peltiers have a nasty thermal instability, where they start getting
> > hotter as you go to higher current instead of colder.
> >
> > If you put them in series, they all see the same current, which along
> > with a spreader plate top and bottom) helps prevent runaway.  If you put
> > them in parallel, the ones with higher delta-T will draw less current
> > than the lower delta-T ones, which gives you a nice stabilizing action
> > at low drive current.
> >
> > If you push them, or you lose cooling water to your heat exchanger, the
> > sign of the gain can invert and the whole thing turns to lava very
> > fast.  (That can happen with individual control or series-connection as
> > well, but the melty ones can't hog current from the cooler ones, so it
> > isn't so unstable.)
> >
> > The other thing is that you have to allow the Peltiers to slide around a
> > bit as the cold plate cools down, because otherwise they'll crack.  So
> > use Arctic Silver or a very small amount of very good thermal grease on
> > one side, and solder on the other.  IIRC you can solder the 30 mm ones,
> > but not the 50 mm ones--it'll be in the datasheet. (Solder is about 100x
> > better than ordinary thermal grease.)
> >
> > I talk a lot about Peltiers in my free thermal chapter,
> > http://electrooptical.net/www/book/thermal.pdf .  There's also an
> > incomplete draft that was intended to accompany the second edition,
> > which has some more stuff in it, at
> > http://electrooptical.net/www/book/draftthermal.pdf .
> 
> Handy info thanks. Although the TECs are available their datasheets are
> often inadequate and miss out details needed for real world designs.
> 
> Can I pick your brains for a slightly odd requirement? Ideally one that
> can be done with at most a pair of TECs one high temp and one normal. On
> paper they were rated for 3A and 14v or so each but in practice I was
> struggling in reverse to get 2v and a few mA out with a candle flame
> heated plate at 250C on one side and a slab of aluminium at -18C on the
> other. I gave up in the end as it was taking too long and safety
> concerns of very hot metal plates and children scuppered it in the end.
> 
> The aim would be for an Xmas science demo to harvest some of the ~100W
> waste heat of a candle flame and drive a ~1W LED. This seemed a modest
> aim when I started out but in practice it proved impossible. To be any
> good the LED must be a *lot* brighter than the candle flame!
> 
> I am guessing that to stand any chance I also need a fan assisted heat
> sink and to be authentic it must all be powered by the TECs. It is for a
> physics demo so hiding a button cell somewhere is just not on...
> 
> --
> Regards,
> Martin Brown

People do this over camp stoves, between a pot of water and the burner. 
Peltiers are really stinky at power generation--their maximum efficiency
is only about 3% iirc.

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
845-480-2058

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

Phil Hobbs wrote:

> Martin Brown wrote:
>> 
>> On 19/07/2012 12:57, Richard Rasker wrote:

...

>> > So I thought I'd use four identical 25W Peltier elements (40 x 40 mm)
>> > instead of one 100W Peltier element (50 x 50 mm). My question: is it
>> > better to connect these Peltier elements in parallel or in series? In
>> > other words: is a Peltier element's heat transport mainly determined by
>> > voltage difference or by current? Any ideas are appreciated.
>> 
>> Current. They are in essence a bunch of high power diode PN junctions.
>> 
>> Their efficiency as a cooler isn't all that great so insulation is
>> important if you are to obtain good thermal stability as is a suitable
>> heatsink on the dump side. Good for having no moving parts thought.

[snip]

> I talk a lot about Peltiers in my free thermal chapter,
> http://electrooptical.net/www/book/thermal.pdf .  There's also an
> incomplete draft that was intended to accompany the second edition,
> which has some more stuff in it, at
> http://electrooptical.net/www/book/draftthermal.pdf .

Wonderful information, thank you very much! I already have a small-sized (30
x 30 mm 10W) single-Peltier setup working properly, but this information
will be very useful when scaling things up.

Richard Rasker
-- 
http://www.linetec.nl

On Jul 19, 9:50=A0am, Martin Brown <|||newspam...@nezumi.demon.co.uk>
wrote:
> On 19/07/2012 14:29, Phil Hobbs wrote:
>
>
>
>
>
> > Martin Brown wrote:
> >> Current. They are in essence a bunch of high power diode PN junctions.
>
> > I'd control them individually, or wire them in series. =A0Compound
> > Peltiers have a nasty thermal instability, where they start getting
> > hotter as you go to higher current instead of colder.
>
> > If you put them in series, they all see the same current, which along
> > with a spreader plate top and bottom) helps prevent runaway. =A0If you =
put
> > them in parallel, the ones with higher delta-T will draw less current
> > than the lower delta-T ones, which gives you a nice stabilizing action
> > at low drive current.
>
> > If you push them, or you lose cooling water to your heat exchanger, the
> > sign of the gain can invert and the whole thing turns to lava very
> > fast. =A0(That can happen with individual control or series-connection =
as
> > well, but the melty ones can't hog current from the cooler ones, so it
> > isn't so unstable.)
>
> > The other thing is that you have to allow the Peltiers to slide around =
a
> > bit as the cold plate cools down, because otherwise they'll crack. =A0S=
o
> > use Arctic Silver or a very small amount of very good thermal grease on
> > one side, and solder on the other. =A0IIRC you can solder the 30 mm one=
s,
> > but not the 50 mm ones--it'll be in the datasheet. (Solder is about 100=
x
> > better than ordinary thermal grease.)
>
> > I talk a lot about Peltiers in my free thermal chapter,
> >http://electrooptical.net/www/book/thermal.pdf. =A0There's also an
> > incomplete draft that was intended to accompany the second edition,
> > which has some more stuff in it, at
> >http://electrooptical.net/www/book/draftthermal.pdf.
>
> Handy info thanks. Although the TECs are available their datasheets are
> often inadequate and miss out details needed for real world designs.
>
> Can I pick your brains for a slightly odd requirement? Ideally one that
> can be done with at most a pair of TECs one high temp and one normal. On
> paper they were rated for 3A and 14v or so each but in practice I was
> struggling in reverse to get 2v and a few mA out with a candle flame
> heated plate at 250C on one side and a slab of aluminium at -18C on the
> other. I gave up in the end as it was taking too long and safety
> concerns of very hot metal plates and children scuppered it in the end.
>
> The aim would be for an Xmas science demo to harvest some of the ~100W
> waste heat of a candle flame and drive a ~1W LED. This seemed a modest
> aim when I started out but in practice it proved impossible. To be any
> good the LED must be a *lot* brighter than the candle flame!
>
> I am guessing that to stand any chance I also need a fan assisted heat
> sink and to be authentic it must all be powered by the TECs. It is for a
> physics demo so hiding a button cell somewhere is just not on...
>
> --
> Regards,
> Martin Brown- Hide quoted text -
>
> - Show quoted text -

Interesting,  I take your question to be, "Where along the I/V curve
is a TEC most efficient when run as a generator?"  (heat to electric
power.)
That should be fairly easy to determine experimentally.

Perhaps you need more elements in series?

And how did you get the -18C?
If this is in the winter time perhaps you'd allow for the cold end to
be dunked into an ice bath... cooled with snow from the great
outdoors.

George H.