pooge said:
pooge
That is one of the matters that Douglas and I have discussed by email. Sorry that some of the discussion is done by email and not online.
If as you say that they only have a few hours to “search” for prior art, and don’t find it in the local newspaper then its ok to claim a new patent?
Such a thing would newer occur in Europe.
Stinius
stinius said:
Don't be so sure. European examiners can't be ALL knowing either, and they have a limited time, too. You can't spend your entire career on one application. It's a mental sweat shop. You can also blame applicants for not giving the examiners the best known art of which they are aware. The examiner then spends precious time finding the art already known to the applicant, instead of being able to use that time more effectively.
pooge said:
I go with Stinius on this one. I've found the British patent search to be very, very thorough. On one of my patents they came up with something that Marconi had patented in 1908! Had to do a bit of redrafting to get round it , too.
But I fear we drift off topic. I hope to get back on the TT quest soon. That thermal model is a pain. What are we doing wrong??
DouglasSelf said:
Hi Doug,
Yes, under the right conditions, one can patent almost anything, and that is not just in the US patent system. The real test is whether the patent has any value and can be defended in court.
Witness how Halcro patented essentially my version of the Hawksford EC scheme, about 10 years after it was published right in the JAES. In that patent, he does not even cite that relevant prior art.
I am not familiar with a patent by Sanken on their approach, if indeed they filed one. However, it is an important distinction that the Sanken approach is, I believe, truly monolithic, with its advantages and disadvantages, while the Onsemi patent is explicitly two die. The lack of electrical isolation in the Sanken scheme is a disadvantage in that you cannot use their diodes in a Vbe multiplier-based bias spreader, as far as I can tell.
Had Sanken patented their idea, and at the same time written a broad enough claim to cover non-monolithic implementations, then Onsemi might have been foreclosed from patenting their approach if the examiner was properly made aware of the Sanken prior art, which in principle is the duty of the applicant.
I see no evidence of a reference to the Sanken approach in the Onsemi patent, be it a Sanken patent or any other Sanken publication of the part (including a spec sheet), so in that regard I think Onsemi may have fallen short in their attempt to know or disclose prior art. It's possible they did not even know about the Sanken part. I believe that had they disclosed and discussed the Sanken prior art, they might have still been able to patent their approach, although some of us may have argued that the Onsemi approach would have been obvious in light of the Sanken approach.
Cheers,
Bob
DouglasSelf said:
IMO, two things:
1. Understimating the thermal resistance of the polymer compound under the diode.
2. Overestimating the equivalent thermal mass of the plate. Unfortunately, I don't think a lumped and linear 1D model really fits here.
I still have to dig in my 10 years old files for the details, I'll try to do this as soon as I'm back home.
OT: I remember the Sterophile review of the Halcro DM38 by John Atkinson -- Candy was quite cagey about describing the their patents -- peculiar in that the material is in the public domain:
Low Distortion Amplifier:
6,600,367
Amplifier Improvements: (Both Mr. Self and Mr. Cordell are referenced in this one!)
6,798,285
Wideband operational amplifier having a plurality of feedback loops 6,052,027
Amplifier Having Ultra Low Distortion:
5,892,398
Power Supply Regulator:
5,563,498
Just search on the assignee: BHC Consulting Pty Ltd.
Low Distortion Amplifier:
6,600,367
Amplifier Improvements: (Both Mr. Self and Mr. Cordell are referenced in this one!)
6,798,285
Wideband operational amplifier having a plurality of feedback loops 6,052,027
Amplifier Having Ultra Low Distortion:
5,892,398
Power Supply Regulator:
5,563,498
Just search on the assignee: BHC Consulting Pty Ltd.
By taking the data given in the patent and estimating the surface of the epoxy layer (number 28 in patent) as twice the size of the diode semiconductor we get:
thickness of epoxy 25 micron R between 10°C/W and 20°C/W ( K= 0.5 and 0.25 W/°C m
thickness 150 microns then R between 60 and 120
JPV
thickness of epoxy 25 micron R between 10°C/W and 20°C/W ( K= 0.5 and 0.25 W/°C m
thickness 150 microns then R between 60 and 120
JPV
DouglasSelf said:
Thanks for the info John.
0.08in = 2.03mm, so it looks like my guess of 2mm was pretty good.
That gives a thermal capacity of 2.25 J/degC, while both my experiments & Bob's indicate about 8 J/degC. Until we work out what is wrong here, I don't think we'll get any further.
Hi Doug,
The only thing I can speculate on this discrepency (assuming we have really got it right as far as all three dimensions of the copper header) is that the thermal capacity of the rest of the package is in the mix as well. Although we believe that the thermal resistance of the packaging material is much higher than that of copper, it still would seem to have considerable thermal mass (I'm just waiving my hands here).
Note that when I estimated the thermal mass of the header, I did it with two different experiments. One was with the device mounted to a large heatsink with a mica washer. The other was with the device essentially in free air. The thermal mass estimates for the two experiments were something like 9 J/C and 8 J/C, which made me feel good at the time.
It would seem that the latter experiment, with the device in free air, would have tended to include some contribution from the thermal mass of the plastic package.
Cheers,
Bob
Bob Cordell said:
The reasoning sounds right but with the dimensions in the patent we shoud have a capacitor of 2.2 J/°C.
I wonder if the convection coef is not higher and should not be neglected. If yes , then your 'slew rate' is in fact a RC time constant and you have an overestimate if you think of a pure capacitor.
A good experiment would be to isolate the transistor with polystyrene and redo the slew experiment with thin feeding wires Perhaps this time the capacitor will be closer to the calculated value of 2 J/°C.
I do not think that the thermal mass of the the plastic will influence because with an estimated volume of twice the copper, I estimate Cplastic = 0.18 J/°C
JPV
JPV said:
Thanks, JPV. Yes, this is still a bit of a mystery. If I get a chance I'll go back and do the thermal slew rate experiment as you suggested, with the ThermalTrak device as "thermally floating" as possible.
Some more experiments where the BJT junction was used as the temperature sensor of the header (i.e., when it is biased by only about 10 mA) would also make sense, since it is much more intimately in thermal contact with the header than the tracking diode.
Cheers,
Bob
Bob Cordell said:
Of course the bjt junction should be the best monitor of the header temperature.
What about the following:
Let T1 be the header temp and T2 the diode temperature.
If we consider that the thermal resistance from transistor junction to header is very low, and the total package very well isolated, the model is 2 capacitor C1 and C2 // with a resistor ( epoxy) in between ( A pi network)., C1 thermal mass of header + chip, C2 thermal mass of diode.
By applying constant power to C1 and measuring slew rate as you suggested you measure C1 and T1 and T2
Then cutting the power, if everything is well isolated, T1 should decay and T2 should rise until they are equal let say T' at equilibrium.
By equating the energies at shut off of the power in the two capacitors and at equilibtrium we have:
(T1². C1/2) + T2².C2/2 = T'².(C1+C2)/2
where T1 and T2 are the temp at shut off time
then (T1²-T'²)/(T'²-T2²) =C2/C1. this gives C2
The time constant towards equilibrium will give the R of epoxy.
If the equilibrium temperature stays stable for a while, we know that our experiment is sound and the estimates are correct.
So in one experiment we can estimate C1 C2 and R epoxy.
Another remark.
In the patent, the author has a strange comment that I do not understand. He is saying that the distance of the diode to the transistor is 3 mm and creates a beneficial hysteresis ???
Perhaps he is insisting on the discrete approach to avoid infrigement of a patent related to an integrated device.
What do you think?
JPV
he may very well have separated the diode from the die to get a patent, but the comment regarding hysteresis is probably there to support his claim that it is different. I presume (without reading the patent) that the hysteresis is related to a time or amplitude delay that he finds beneficial.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.