Switching polarity without relays

[hermit]

It's funny how Q1 and Q6 are "soldered".

Aye.. the 3D model for it is rotated (bug in that footprint which is from KiCad's standard lib), but I checked the footprints and they look right

E for edit footprint in pcbnew and go to 3d properties.  You can rotate it from there if you want.

[Muxr]

It's funny how Q1 and Q6 are "soldered".

Aye.. the 3D model for it is rotated (bug in that footprint which is from KiCad's standard lib), but I checked the footprints and they look right

E for edit footprint in pcbnew and go to 3d properties.  You can rotate it from there if you want.

Thanks for that tip. Didn't really bother me but now that I know it can be easily fixed might as well fix it. Cheers!

[Andreas]

Hello,

there are some design considerations with peltiers:

On heating cooling boxes there is usually the hint that when switching
from heating to cooling you will need a cool down (off) phase of 5 minutes.

The reason is that a peltier also acts as generator (battery) when there is a temperature difference.
So when immediately reversing polarity there is a large current flow (both voltage sources in series).

So I use dedicated heater foils (similar to car mirror heaters) together with the peltier to control the temperature.
https://www.reichelt.de/Heating-Foils/THF-77110/3/index.html?ACTION=3&GROUPID=7777&ARTICLE=108462&OFFSET=16&
(unfortunately pricing has gone up by a factor 3-4 recently).

With best regards

Andreas

[Zero999]

Hello,

there are some design considerations with peltiers:

On heating cooling boxes there is usually the hint that when switching
from heating to cooling you will need a cool down (off) phase of 5 minutes.

The reason is that a peltier also acts as generator (battery) when there is a temperature difference.
So when immediately reversing polarity there is a large current flow (both voltage sources in series).

So I use dedicated heater foils (similar to car mirror heaters) together with the peltier to control the temperature.
https://www.reichelt.de/Heating-Foils/THF-77110/3/index.html?ACTION=3&GROUPID=7777&ARTICLE=108462&OFFSET=16&
(unfortunately pricing has gone up by a factor 3-4 recently).

With best regards

Andreas

That's interesting. I wonder if short circuiting the Peltier module, rather than leaving it open circuit, will improve the thermal transfer?

[raptor1956]

An H-bridge is certainly the conventional solution.  It doesn't have to be that expensive.  For instance, here is a $4 h-bridge controller with integrated switches and a 5 amp current limit:

https://www.digikey.com/product-detail/en/infineon-technologies/TLE5205-2G/TLE5205-2GINCT-ND/1557940?WT.z_cid=ref_findchips_standard

The Rds On is fairly high at 200 mOhm, so it will dissipate quite a bit of heat.  Discrete MOSFETs would do better, but not as simple, and possibly more expensive.

That said: you said you want to use this to drive a peltier.  It is not really recommended to drive a peltier with either unfiltered PWM or thermostat style on-off regulation. 

The problem with PWM (especially in cooling mode) is that the cooling power is proportional to I, but the parasitic heating is proportional to I^2.  So with PWM, you get the worst case resistive heating which you then have to cool.

The problem with slow on-off control is that the frequent thermal cycling causes stress from thermal expansion.

You really want to drive a PWM with DC if you want good performance and longevity.  Fortunately, this is relatively easy to do with a H bridge PWM driver and a simple LC filter.  This basically turns it into a buck converter.

Thanks, appreciate your input.  I think I'm going to proto with H-bridge but select units with lower RdsON.  I also planned to filter the PWM output to reduce the peak-to-peak down to the average using caps or LC if needed. 

In a past life (IBM) we used peltier coolers to cool chemicals used in the semiconductor industry to dope the silicon and getter electrical contaminants.  Back in the day we used BBr3 and POCl as well as oxide cleaner TCA and these needed to be maintained at between 5C and 20C.  The PS for these chillers were, as I remember, 6A.


Brian

[raptor1956]

Hello,

there are some design considerations with peltiers:

On heating cooling boxes there is usually the hint that when switching
from heating to cooling you will need a cool down (off) phase of 5 minutes.

The reason is that a peltier also acts as generator (battery) when there is a temperature difference.
So when immediately reversing polarity there is a large current flow (both voltage sources in series).

So I use dedicated heater foils (similar to car mirror heaters) together with the peltier to control the temperature.
https://www.reichelt.de/Heating-Foils/THF-77110/3/index.html?ACTION=3&GROUPID=7777&ARTICLE=108462&OFFSET=16&
(unfortunately pricing has gone up by a factor 3-4 recently).

With best regards

Andreas

The deep space probes that use Pu as the energy source (RTG) generate electricity using thermocouples using the Seebeck effect -- quite similar to the way peltier modules work. 


Brian

[Andreas]

I also planned to filter the PWM output to reduce the peak-to-peak down to the average using caps or LC if needed. 

Caps alone will be not sufficient to filter the current ripple to below 10% of the average current (requirement for peltiers).
So you will have to use a relative high PWM frequency (and fast gate drivers for the FETs) if you want to avoid very bulky inductors. (6A).
And for the fans you will need the (nearly) full supply voltage.


with best regards

Andreas

[raptor1956]

I also planned to filter the PWM output to reduce the peak-to-peak down to the average using caps or LC if needed. 

Caps alone will be not sufficient to filter the current ripple to below 10% of the average current (requirement for peltiers).
So you will have to use a relative high PWM frequency (and fast gate drivers for the FETs) if you want to avoid very bulky inductors. (6A).
And for the fans you will need the (nearly) full supply voltage.


with best regards

Andreas

The other thing is that since the polarity switches the caps would need to be tolerant of either polarity.   That won't be an issue for inductive components, but electrolytic caps could be a problem. 


Brian

[Zero999]

I also planned to filter the PWM output to reduce the peak-to-peak down to the average using caps or LC if needed. 

Caps alone will be not sufficient to filter the current ripple to below 10% of the average current (requirement for peltiers).
So you will have to use a relative high PWM frequency (and fast gate drivers for the FETs) if you want to avoid very bulky inductors. (6A).
And for the fans you will need the (nearly) full supply voltage.


with best regards

Andreas

The other thing is that since the polarity switches the caps would need to be tolerant of either polarity.   That won't be an issue for inductive components, but electrolytic caps could be a problem. 


Brian

If you want to filter PWM, you need inductors. A capacitor alone will increase the power dissipation in the switch, as there will be nothing to limit the charging current.

[Muxr]

I put my little h-bridge together.. will be testing it this weekend.



That polyfuse is huge so I mounted it in the back.. (note to self, always check the dimensions).

[Zero999]

That polyfuse is huge so I mounted it in the back.. (note to self, always check the dimensions).

I doubt it will do anything to protect the transistors though.

[jbb]

If you're doing PWM already, you might as well go the whole hog and do H-bridge PWM.  You can get PWM control and polarity switching, plus smooth control all the way from +ve output to -ve output (and vice versa).

One thing to be aware of: when you are transitioning the system from warming to cooling (and vice versa), the Peltier can actually operate as a thermopile generator, and feed energy (via the power converter) back into the DC link.

[raptor1956]

OK, so I'm going to prototype the H-bridge and have ordered some parts from Digikey and Amazon.  I first ordered some SMD MOSFET's but the first batch I ordered were too small, I mean like 3.3mm on a side.  Amazingly though, the the constant current rating is -11A to -15A with pulse to -46A.  The RdsON for the P-channel NTTFS3A08PZTWG is 6.7mOhm at -4.5Vgs.  I ordered a similar N-channel Si7108DN which has RdsOn values of 6.1mOhm at 4.5Vgs and 4.9mOhm at 10Vgs.  Again, the power and current handling capabilities of MOSFET's barely more than an 1/8" on a side is amazing. 

But, I realized my mistake in ordering parts that small and then ordered some SOP8 sized parts that should be more workable to prototype.  The P-channel version is a Vishay Si7141DP that has RdsON values of 3.0mOhm at -4.5Vgs and 1.9mOhm at -10Vgs with constant current well above the 5A I need.  The N-channel one, also SOP8 sized, actually has a bit higher RdsON values of 3.2mOhm at 4.5Vgs and 2.6mOhm at 10Vgs.  These two are 1.27mm pitch versus the 0.65mm pitch of the first two and the SOP8 proto boards (DFRobot FIT0290) will work with the latter two parts.

I am just amazed at where we've come with power devices and with such low RdsON values and current capacities of more than 20A continuous in devices that are just about a quarter the size of a finger nail -- wow...


Brian