HobbyKing.com New Products Flash Sale
Reply
Thread Tools
Old Feb 17, 2013, 04:00 PM
JohnM
Joined Mar 2011
23 Posts
Awww...John, ya' beat me to it.
CamLight is offline Find More Posts by CamLight
Reply With Quote
Sign up now
to remove ads between posts
Old Feb 17, 2013, 04:02 PM
JohnM
Joined Mar 2011
23 Posts
Mike,
A couple of other things...

- I don't remember, but doesn't the 7809 need at least 1uF on both its input and output?

- Is the op-amp power input decoupled very closely to the power pin? Use both 1uF and 0.01uF ceramic caps with the 0.01 very close to the pin.

- It might not make a difference in this circuit, but, electrolytic caps are useless for effective decoupling. Use ceramic caps or, if you have to, tantalum caps for the 10uF ones if you can't find ceramics with the right voltage rating at that size.
John
CamLight is offline Find More Posts by CamLight
Reply With Quote
Old Feb 17, 2013, 10:06 PM
Registered User
USA
Joined Jan 2002
4,372 Posts
Quote:
Originally Posted by jj604 View Post
Mike as someone who had a go at building the Biskit load and am now working on an alternative, the one thing I learned is to check the MOSFETs first if it won't work!
Do you mind sharing your ideas on an alternative? I've been thinking about designing/building a discharger myself and keeps circling between using MOSFETS, globes, and power resistors for the load. Sooner or later I'll need to get on the bus. How powerful is the alternative you're working on?
Ohmic is offline Find More Posts by Ohmic
Reply With Quote
Old Feb 18, 2013, 12:19 AM
ancora imparo
jj604's Avatar
Melbourne, Australia
Joined Jul 2005
5,628 Posts
It's a joint project with a couple of others on the forums and there is still a way to go to make sure it is all working reliably. Not trying to be secretive but it's not ready for public discussion just yet.

Basically an aircooled load using FETs as we wanted something simple and hassle free for testing LiPo packs - water cooling is the best for dumping lots of energy but it is just a pain to set up each time you want to use it.

Resistors are too inflexible and globes too inflexible and uncontrollable for serious load testing. Regenerative charging is fine if you have the big Pb batteries set up but I don't. A combo of switching FETs and resistors was an option but is more complicated than using straight FETs and the hybrid regen method discharging into a resistive load is also too much hassle for me. The shorted LM317 solution had no advantages over power FETs as a controlled load - it is fine just as a simple discharger IMO.

Using reasonably priced CPU coolers I think we are now confident of 700Watts per cooler, possibly 800Watts, so 3 coolers will give a fairly compact 2000 Watt load, possibly 2400 Watt.

Intention is to make it a current amplifier so you can add it on to a CBA or quality charger that does discharging with computer logging like a Hyperion or FMA.

Essentially the same functionality as West Mountain's CBA 2000Watt amplifier but a lot cheaper.

In real life it's trickier than it looks from the very simple circuit though (as John Muchow keeps reminding us).

John
Quote:
Originally Posted by Ohmic View Post
Do you mind sharing your ideas on an alternative? I've been thinking about designing/building a discharger myself and keeps circling between using MOSFETS, globes, and power resistors for the load. Sooner or later I'll need to get on the bus. How powerful is the alternative you're working on?
jj604 is offline Find More Posts by jj604
Reply With Quote
Old Feb 18, 2013, 02:41 AM
Registered User
flyboy1001e's Avatar
Nelson B.C. Canada
Joined Feb 2003
207 Posts
jj604

John I disconnected the mosfets and tested them and they are ok.The interesting thing is once you connect the Current Sensing Resistors into the circuit you have continuity (short) between drain and source.


btw: thanks for: Improved simple auto disconnector for battery charging/testing

----------

camlight

John I was using a 3s 1800mah pack, so I think it could have a huge rush of current flow.
Checked the mosfets are good
What's the MOSFET gate voltage range as you change the pot's value? Will have to solder it back together and measure
Do you make this measurement from gate to ground with supply pack connected?
My pack voltage was 12v regulated to 9v by 7809
I checked the pdf for lm7809 and it says typical input CI = 0.33μF, output CO = 0.1μF
Is the op-amp power input decoupled very closely to the power pin? Not quite sure what you mean?
Will use ceramic caps

Mike
flyboy1001e is offline Find More Posts by flyboy1001e
Reply With Quote
Old Feb 18, 2013, 07:03 AM
Registered User
United States, AL
Joined Apr 2007
821 Posts
Jj: I think you will ultimately find CPU coolers very inadequate for those power levels. The heat conduction path is too long and there is not nearly enough surface area to dump so much heat into the air. Your FETs will get too hot too fast for any significant power level.

For 2kw you will need at least 8 FETs at around $5 each and they will have to be individually controlled so add $10 each. That's $120 for the electronics alone. You will need more FETs because those CPU coolers will not do 500w each. You're going to be spending more that the cost of a PL6. Not worth the money. Do it if you want to learn but don't do it to save $ because you won't.
biskit is offline Find More Posts by biskit
Reply With Quote
Old Feb 18, 2013, 12:33 PM
JohnM
Joined Mar 2011
23 Posts
Quote:
Originally Posted by flyboy1001e View Post
camlight

John I was using a 3s 1800mah pack, so I think it could have a huge rush of current flow.
Checked the mosfets are good
What's the MOSFET gate voltage range as you change the pot's value? Will have to solder it back together and measure
Do you make this measurement from gate to ground with supply pack connected?
My pack voltage was 12v regulated to 9v by 7809
I checked the pdf for lm7809 and it says typical input CI = 0.33μF, output CO = 0.1μF
Is the op-amp power input decoupled very closely to the power pin? Not quite sure what you mean?
Will use ceramic caps

Mike
Hi Mike,
Yes, measure the gate voltage to ground when then the supply pack is connected and the pot is being rotated. But, you'll need to connect a pack or power supply to be "discharged" too. Otherwise, the op-amp will just drive the FETs hard on, at max. possible voltage, trying to get any current to pass through the current sense resistors. It's best to use a constant-current supply here as you can set it to a value low enough to not burn out the FETs during testing if there are any problems.

Ahhh, my bad regarding the 7809. Use the cap. the datasheet recommends.

I'd add a 0.01uF or 0.1uF ceramic cap directly across pins 4 and 8 (GND and VDD) of the op-amp. This "decouples", or separates/isolates, the op-amp from any noise being brought in via the power pin. This noise can cause all sorts of problems. Keep the cap's leads as short as possible.
CamLight is offline Find More Posts by CamLight
Reply With Quote
Old Feb 18, 2013, 12:54 PM
JohnM
Joined Mar 2011
23 Posts
Quote:
Originally Posted by jj604 View Post
Using reasonably priced CPU coolers I think we are now confident of 700Watts per cooler, possibly 800Watts, so 3 coolers will give a fairly compact 2000 Watt load, possibly 2400 Watt.

Intention is to make it a current amplifier so you can add it on to a CBA or quality charger that does discharging with computer logging like a Hyperion or FMA.

John
I love the idea and am looking forward to hearing more about it!

I have the same worries as biskit though. How are you measuring your temperatures?

I haven't seen any CPU coolers that are rated for less than 0.35C/W thermal resistance (sink-to-air) and that's going to limit your power ratings even if you have a lot of FETs (to lower their equivalent thermal resistance). Unless you've found some incredible coolers that I'd love the hear about!

Just as an example...
Using an efficient heat sink (<0.2C/W), four huge TO-264 FETs with very low thermal resistance (<0.083C/W junction-sink, equiv.), and a very high-speed/high-flow fan, I was only able to get to about the 500W level before exceeding the max. junction temperature rating of the FETs in my CC-400 loads. Running the FETs at a resonable temp., to ensure reliability, limited the power level to 400W.

John
CamLight is offline Find More Posts by CamLight
Last edited by CamLight; Feb 18, 2013 at 03:00 PM. Reason: mispelling
Reply With Quote
Old Feb 18, 2013, 06:32 PM
ancora imparo
jj604's Avatar
Melbourne, Australia
Joined Jul 2005
5,628 Posts
Very possibly but we will wait and see. I will put up some data when I think it is reliable and repeatable. I certainly had a twin FET load running for an extended period on a single cooler at just under 700 Watts but then another didn't make that load. There is also the whole question of how genuine the FETs are as well. Some of the cheaper on-line suppliers are very suss. The FETs function but the original thermal specs get a bit bent.

"Not worth the money" is entirely a value judgment based on your own priorities.

I never said anywhere I was attempting to make a cheap solution. I know that was the original driver behind this thread but it's not mine. I have plenty of cheap solutions that will provide 2000Watts + load but none are what I want. And I have a Camlight 400 which with a CBA will do 500Watts.

However as far as I know the only currently available device that is both accurate, controllable and can act as a current amplifier at this sort of load (2000Watts) is the West Mountain CBA amplifier - which is a special order item costing $3378.60 plus shipping.

Enough speculative discussion. Promise I will come back with a progress report - good or bad.
Quote:
Originally Posted by biskit View Post
Jj: I think you will ultimately find CPU coolers very inadequate for those power levels. The heat conduction path is too long and there is not nearly enough surface area to dump so much heat into the air. Your FETs will get too hot too fast for any significant power level.

For 2kw you will need at least 8 FETs at around $5 each and they will have to be individually controlled so add $10 each. That's $120 for the electronics alone. You will need more FETs because those CPU coolers will not do 500w each. You're going to be spending more that the cost of a PL6. Not worth the money. Do it if you want to learn but don't do it to save $ because you won't.
jj604 is offline Find More Posts by jj604
Reply With Quote
Old Feb 18, 2013, 06:51 PM
ancora imparo
jj604's Avatar
Melbourne, Australia
Joined Jul 2005
5,628 Posts
John, I will have to do a bit more work and will share the numbers then. Most vendors seem a bit coy about their actual Rsa and the computer test guys are really only interested in the max. chip temp under particular load tests. Measured Rsa on my cooler is 0.065 C/Watt (unless I have made some silly mistake) and the manufacturer claimed 0.7. WG got slightly lower numbers for his similar cooler. These are fairly common multiple heat pipe CPU coolers with a decent solid Cu base.

Based on (Measured HS temp-Ambient temp)/Measured Power in.

Agree measuring the cooler heat sink temp is the tricky part. Pity FETs don't come with inbuilt temp sensors like CPU chips! Will need to talk about how accurate the value measured is but will do that later when I am more confident. Two of us have done it using different methods and got similar results. Will very much appreciate your expert input.

John

Quote:
Originally Posted by CamLight View Post
I love the idea and am looking forward to hearing more about it!

I have the same worries as biskit though. How are you measuring your temperatures?

I haven't seen any CPU coolers that are rated for less than 0.35C/W thermal resistance (sink-to-air) and that's going to limit your power ratings even if you have a lot of FETs (to lower their equivalent thermal resistance). Unless you've found some incredible coolers that I'd love the hear about!

Just as an example...
Using an efficient heat sink (<0.2C/W), four huge TO-264 FETs with very low thermal resistance (<0.083C/W junction-sink, equiv.), and a very high-speed/high-flow fan, I was only able to get to about the 500W level before exceeding the max. junction temperature rating of the FETs in my CC-400 loads. Running the FETs at a resonable temp., to ensure reliability, limited the power level to 400W.

John
jj604 is offline Find More Posts by jj604
Reply With Quote
Old Feb 18, 2013, 08:32 PM
Registered User
United States, AL
Joined Apr 2007
821 Posts
So what is your goal? If you just want a few kW then you will probably have to build it yourself as you obviously know!

The heat path for at air cooled heat sink is so long it requires a bigger temperature gradient and thus hotter FETs. A water cooled sink can have a very short heat path and thus run cooler.

I will tell you with the fda FET earlier in this thread running at 300w with turbulent cold water less than 1cm away through aluminum(ie an infinite heat sink) the top of the FET will burn you. Other than using salt water below freezing or a better conducting material(copper) there isn't much you can do to get more cooling.

John has already posted how he could only get 100w per FET with air cooling. We all would love to see your contraption that can get 350w.
biskit is offline Find More Posts by biskit
Reply With Quote
Old Feb 18, 2013, 09:08 PM
JohnM
Joined Mar 2011
23 Posts
Quote:
Originally Posted by jj604 View Post
John, I will have to do a bit more work and will share the numbers then. Most vendors seem a bit coy about their actual Rsa and the computer test guys are really only interested in the max. chip temp under particular load tests. Measured Rsa on my cooler is 0.065 C/Watt (unless I have made some silly mistake) and the manufacturer claimed 0.7. WG got slightly lower numbers for his similar cooler. These are fairly common multiple heat pipe CPU coolers with a decent solid Cu base.

Based on (Measured HS temp-Ambient temp)/Measured Power in.

Agree measuring the cooler heat sink temp is the tricky part. Pity FETs don't come with inbuilt temp sensors like CPU chips! Will need to talk about how accurate the value measured is but will do that later when I am more confident. Two of us have done it using different methods and got similar results. Will very much appreciate your expert input.

John
Wow...0.065C/W is an unbelievably low number for a good multiple heat pipe cooler. But, the technology is always improving! If you're measuring the temp (no IR gun) from inside the heat sink, directly next to the MOSFET base, then you've got a good reading on the heat sink temp. Any further than, let's say, 1/8" away from center of the rear mounting plate of the MOSFET (from inside the sink) and the temp reading will be way too low.

It can often be easier to just drill through the heat sink and measure the MOSFET temp directly (via type-K thermocouple against the rear plate) to get the most accurate temperature. Especially since you can only estimate the FET-case-to-sink thermal resistance. With that reading, you have the actual case temp of the FET and it's easy to derive the junction temp. But there are other ways to do all this measurement stuff and I'm looking forward to hearing how you guys did it.

I just noticed your mention of dual FETs running at 700W on a single heat sink. 350W/FET on a standard heat sink is incredible. Actually, it's incredible no matter what heat sink is used! I am verrrrry anxious to hear more about your setup...hint...hint.

Anyway...just a few thoughts of mine as we wait for your development & testing to be completed.
CamLight is offline Find More Posts by CamLight
Reply With Quote
Old Feb 18, 2013, 09:17 PM
Registered User
USA
Joined Jan 2002
4,372 Posts
Quote:
Originally Posted by jj604 View Post

However as far as I know the only currently available device that is both accurate, controllable and can act as a current amplifier at this sort of load (2000Watts) is the West Mountain CBA amplifier - which is a special order item costing $3378.60 plus shipping.

Enough speculative discussion. Promise I will come back with a progress report - good or bad.
I look forward to your progress report.

When you get into the price range for the West Mountain Radio CBA amplifier there are several commercial manufacturers who make load test products for the industry. These are primarily used for load testing for power supplies. HP/Agilent come to mind.
Ohmic is offline Find More Posts by Ohmic
Reply With Quote
Old Feb 18, 2013, 11:23 PM
ancora imparo
jj604's Avatar
Melbourne, Australia
Joined Jul 2005
5,628 Posts
John, we did have an argument about how accurate the temp reading was. Wayne and I took two different approaches but interestingly got similar Rsa numbers. Obviously if the heat-sink temperature is not accurate and fairly represents the temp at the FET embedded heat-sink surface then any calculations of FET die temperature are wrong. His cooler has a solid Cu base with embedded heat pipes held in mechanically and he used a surface thermocouple, mine is a hybrid Al and Cu pipe version which is probably not as good and may well have substantial temp gradient.

I drilled the base and fitted a K type about 0.5mm below the surface with epoxy then lapped the surface flat. It is very close to the FET tab but not touching it so hopefully represents the temperature of the cooler. The thermal mass of the epoxy is very small as the hole is just big enough to clear the weld bead. It was all a bit tricky as the spacing between the heat pipes is tiny.

Didn't seem to have damaged anything. Attached diagram gives you the idea.

Still not convinced it is accurate but I can't think of any better way to measure the sink temperature accurately. You can't drill through the heat pipe!

I did try and measure the FET case temp directly. I was able to embed a thermocouple in a small excavation in the tab, but the soldering threw the readings out. See photo. Thermocouples can be funny things. I suspect in this case the solder was bridging the junction and the various couples don't add up properly. I have seen reference to soldered K types but it all seems to be a bit of a black art.

Be very grateful for your expert comment on the heat flows in that cross section and the likely error in the measured temperature. It is quite a complex thermal system. I suspect these phase change heat pipe coolers are significantly better than simple water cooling. I was amazed when I looked up the amount of heat that gets absorbed in changing from water to steam.

Really want to hold off on any claim this actually works until we have consistent reliable operation. It may well be an optimistic dream but the very first run I did on this cooler with a single FET and twin fans on the cooler I got several minutes of:

24.1Volts and 20.8 Amps which is 501.3 Watts.

The HS temp I measured stabilised very quickly at 54.7C and the calculated max from the published Rjc and Rcs for that FET was 75C. If there is any viability in the numbers I think from preliminary tests 350Watts/FET is feasible with two FETs on the one cooler.

It's all in bits at the moment awaiting final assembly.

Didn't mean to hijack biskit's thread but it just grew out of one of the comments I made. Will start a proper thread for this in due course. Promise I will keep you all up to date when we have some results we are confident in. Till then I'm not speculating any more.

John


Quote:
Originally Posted by CamLight View Post
Wow...0.065C/W is an unbelievably low number for a good multiple heat pipe cooler. But, the technology is always improving! If you're measuring the temp (no IR gun) from inside the heat sink, directly next to the MOSFET base, then you've got a good reading on the heat sink temp. Any further than, let's say, 1/8" away from center of the rear mounting plate of the MOSFET (from inside the sink) and the temp reading will be way too low.

It can often be easier to just drill through the heat sink and measure the MOSFET temp directly (via type-K thermocouple against the rear plate) to get the most accurate temperature. Especially since you can only estimate the FET-case-to-sink thermal resistance. With that reading, you have the actual case temp of the FET and it's easy to derive the junction temp. But there are other ways to do all this measurement stuff and I'm looking forward to hearing how you guys did it.

I just noticed your mention of dual FETs running at 700W on a single heat sink. 350W/FET on a standard heat sink is incredible. Actually, it's incredible no matter what heat sink is used! I am verrrrry anxious to hear more about your setup...hint...hint.

Anyway...just a few thoughts of mine as we wait for your development & testing to be completed.
jj604 is offline Find More Posts by jj604
Reply With Quote
Old Feb 18, 2013, 11:52 PM
JohnM
Joined Mar 2011
23 Posts
John,
Thanks for the info! Sent you a PM to not dilute biskit's thread anymore.

Your sensing sure seems to be set up for accurate heat sink temps. I also use thermal compound or thermal epoxy to embed the thermocouple. I never tried soldering but I can definitely see you having problems with that. No idea why it's recommended when there are great epoxies out there (as long as you can run the test long enough for the thermal lag of the epoxy to not be a problem).

Looking forward to your updates!

[Edit] Thanks for the additional info. Your numbers look good and I'm amazed by that heat sink.
CamLight is offline Find More Posts by CamLight
Last edited by CamLight; Feb 19, 2013 at 01:17 AM.
Reply With Quote
Reply


Thread Tools

Similar Threads
Category Thread Thread Starter Forum Replies Last Post
Discussion PIC PWM and FETs as constant-current discharger? meteor DIY Electronics 21 Jan 09, 2010 10:17 AM
Mini-HowTo How to build a constant current battery discharger biskit Batteries and Chargers 25 Sep 06, 2009 07:53 PM
how to build a vaccum forming box? rayc Scale Kit/Scratch Built 14 Feb 28, 2008 10:46 AM
Discussion DIY analog constant-current discharger? meteor DIY Electronics 7 Jan 06, 2006 10:09 AM
Anbody know how to build a prop? Pork Tornado Parkflyers 4 Aug 03, 2001 04:10 PM