A properly dimensioned resistor is more fail-safe.
And those bulk film ones for heatsink mounting are not so expensive.
Patrick
And those bulk film ones for heatsink mounting are not so expensive.
Patrick
I've been looking at different solutions for HV regulators and this the most interesting I've seen so far. The suggested series pass device is EOL, although still in stock. I was looking at alternatives that have higher DC SOA and found a couple.
I found a MOSFET from Onsemi FCH041N60E (data sheet https://www.mouser.com/datasheet/2/308/1/FCH041N60E_D-1806137.pdf) and an IGBT from Onsemi FGH60N60SMD (data sheet https://www.mouser.com/datasheet/2/308/FGH60N60SMD_D-1808867.pdf). Both of these are good for 600V 1A DC within the SOA curve (as specified). The MOSFET is a bit expensive, though.
Curiously enough for an IGBT, the FGH60N60SMD has a DC SOA specified. It has a higher max temperature (max Tj = 175°C vs 150°C for the MOSFET) and power derating (300W at Tc = 100°C vs 237W Tc = 100°C for the MOSFET). It's also almost half the price and because of positive temperature co-efficient it could be paralleled. I know that IGBTs don't have reputation for working well with DC in the past (die hot spots etc), but one would think they wouldn't specify DC SOA if they haven't tested it...
Am I totally in deep waters here, or could these transistors be alternatives for the pass device?
I found a MOSFET from Onsemi FCH041N60E (data sheet https://www.mouser.com/datasheet/2/308/1/FCH041N60E_D-1806137.pdf) and an IGBT from Onsemi FGH60N60SMD (data sheet https://www.mouser.com/datasheet/2/308/FGH60N60SMD_D-1808867.pdf). Both of these are good for 600V 1A DC within the SOA curve (as specified). The MOSFET is a bit expensive, though.
Curiously enough for an IGBT, the FGH60N60SMD has a DC SOA specified. It has a higher max temperature (max Tj = 175°C vs 150°C for the MOSFET) and power derating (300W at Tc = 100°C vs 237W Tc = 100°C for the MOSFET). It's also almost half the price and because of positive temperature co-efficient it could be paralleled. I know that IGBTs don't have reputation for working well with DC in the past (die hot spots etc), but one would think they wouldn't specify DC SOA if they haven't tested it...
Am I totally in deep waters here, or could these transistors be alternatives for the pass device?
Johan, very good finds! I have been looking for a rep-lacement, missed these devices.
I agree, if the DC SOA is specified you can reasonably assume that it works to that.
I have not tried an IGBT in the T-reg but I see no reason why it wouldn't work well.
But stability is something to keep an eye on.
Another thing is that the IGBT has a higher on threshold voltage, at the specified max of 6V you probably need some change to the circuit to accomodate that, a change of one of the zener diode's voltage.
Why not try it!
Jan
I agree, if the DC SOA is specified you can reasonably assume that it works to that.
I have not tried an IGBT in the T-reg but I see no reason why it wouldn't work well.
But stability is something to keep an eye on.
Another thing is that the IGBT has a higher on threshold voltage, at the specified max of 6V you probably need some change to the circuit to accomodate that, a change of one of the zener diode's voltage.
Why not try it!
Jan
Thanks for the reply. I might have a try at this eventually.
Just for curiosity, I was also looking at the new generation of SiC JFETs that with a MOSFET in a cascode combination could work well in such an application. There are integrated cascode packages available. However, these are still very expensive, so not an option when there are cheaper alternatives available. Here is one device for reference, though: UnitedSiC UJ3C065030T3S. The DC SOA is not as good as the previously mentioned MOSFET and IGBT (there are more powerful cascodes, but much more expensive ones from the same manufacturer).
Just for curiosity, I was also looking at the new generation of SiC JFETs that with a MOSFET in a cascode combination could work well in such an application. There are integrated cascode packages available. However, these are still very expensive, so not an option when there are cheaper alternatives available. Here is one device for reference, though: UnitedSiC UJ3C065030T3S. The DC SOA is not as good as the previously mentioned MOSFET and IGBT (there are more powerful cascodes, but much more expensive ones from the same manufacturer).
I've ordered my HT Regulator and I'm ordering parts that I don't have. 2 electrolytics are listed as an "elec-rad10" package? What is that? The BOM says "pitch 0.1in". Is that the pin spacing?
@jan.didden I could not found any information about this regulator Power-supply ripple rejection (PSRR). Do you know how much it is db? Any graph for this?
Hi @jan.didden
With semiconductors being hard to come buy at this time (e.g. your replacement recommendation NTP110N65S3HF is not available for the next months from Mouser or Digikey), I was wondering, which criteria the MOSFETs must fit.
My layman's guess would be:
If these parameters are enough, there would be plenty of options.
Is this correct understanding?
What I also do not get is why the chosen type is that powerful. Even with Vin-Vout in the area of 100-200V and a reasonable current, this seems like complete overkill.
I almost randomly picked one found with parameterized search in Mouser (IPA95R130PFD7XKSA1), which, for moderate Vin-Vout seems to be pretty OK. For higher current, the Pd is not enough, sure, but depending on the use case, I think there can be lots of alternatives, right?
BTW, using the IPA95R130PFD7XKSA1 in the sim, the circuit was not stable. I had to tame it using a zobel network between G and S. Is that because it is faster?
EDIT: Another random (much more powerful) find by the above parameters is SIHG15N80AE-GE3 by Vishay and that is rather inexpensive.
Thanks,
Mo
With semiconductors being hard to come buy at this time (e.g. your replacement recommendation NTP110N65S3HF is not available for the next months from Mouser or Digikey), I was wondering, which criteria the MOSFETs must fit.
My layman's guess would be:
- (very) low Rds On to minimize Pd
- high enough Id and Pd
- Vgs high enough for the circuit (should be easy, in my LTSpice sim this is 3-4V).
If these parameters are enough, there would be plenty of options.
Is this correct understanding?
What I also do not get is why the chosen type is that powerful. Even with Vin-Vout in the area of 100-200V and a reasonable current, this seems like complete overkill.
I almost randomly picked one found with parameterized search in Mouser (IPA95R130PFD7XKSA1), which, for moderate Vin-Vout seems to be pretty OK. For higher current, the Pd is not enough, sure, but depending on the use case, I think there can be lots of alternatives, right?
BTW, using the IPA95R130PFD7XKSA1 in the sim, the circuit was not stable. I had to tame it using a zobel network between G and S. Is that because it is faster?
EDIT: Another random (much more powerful) find by the above parameters is SIHG15N80AE-GE3 by Vishay and that is rather inexpensive.
Thanks,
Mo
Last edited:
Low Rds(on) is no requirement, the FET never is fully on.
Besides Vds(max) and Id(max) you also should look at the SOA in the data sheet.
If the supply is shorted and the FET has to support the full set max Iout at the full Vin, does it survive say half a second before it switches off?
In practise, you would look for a case that does not have the plastic isolator on the tab as that worsens thermal impedance to the heatsink and is also shown as a worsening of the SOA.
Jan
Besides Vds(max) and Id(max) you also should look at the SOA in the data sheet.
If the supply is shorted and the FET has to support the full set max Iout at the full Vin, does it survive say half a second before it switches off?
In practise, you would look for a case that does not have the plastic isolator on the tab as that worsens thermal impedance to the heatsink and is also shown as a worsening of the SOA.
Jan
Why allow such a long a period before switching off?
BTW... the below is the zobel network, which helped to keep some of the MOSFETs I tried in LTSpice from oscillating. I could not see any disadvantage when adding it.
Bad idea, you modulate the output voltage with the input ripple.
If you got oscillations, get the right MOSFET.
Jan
If you got oscillations, get the right MOSFET.
Jan
I compared two variants in LTSPice now: One with a type for which the ZN is not necessary in Spice, and one which needs it.
You are right, the difference for a 300V input signal with 30V noise modulated on top, is 100uV noise at the output without the ZN and 300uV with it. So it is higher, but still is not too bad if it allows a broader selection of MOSFETs IMO.
The advise to get a better fitting MOSFET still make sense of course. But the FDP12N60NZ is not available anymore and the recommended alternative NTP110N65S3HF is not available for the next time either.
So, how did you select the two above? Trial and error? Or are there some other parameters not mentioned yet?
Thanks,
Mo
A good way is to go to Mouser or Digikey and put in the parameters you want which quickly gives you a shortlist.
Then check the data sheets for things like SOA.
It can take some time but if you do it in a logical step by step way it's not that bad.
Jan
Then check the data sheets for things like SOA.
It can take some time but if you do it in a logical step by step way it's not that bad.
Jan
Yes, thanks, I did. I do not have working spice models for them yet. For FCH041N60E I have a lib from Fairchild but it is encrypted and I need to figure out how to use.
And for FGH60N60SMD I couldn't find any model.
I hear some people saying "it is only simulation", but because I tried to find replacements using the approach you suggested (parameterized Mouser search, which I also mentioned in #89), which do not work in the circuit in Spice at all, I am reluctant to use any that does not work at least in Spice.
Thanks
Design a version using a Philbrick tube opamp. http://www.philbrickarchive.org/k2-w_refurbished.pdf
The purists will remain happy.
Not sure it will be useful to go into that detail in LTspice - the models are almost sure not accurate enough.
In cases like this I use LTspice as a 'proof of concept' investigation, not to determine whether one device oscillates and the other not.
Modeling circuit and PCB parasitics is anyway hopeless.
Jan