Indeed.
I have searched for ready-made alternatives, but unsuccessfully: some delayed fuses might work, but it would very type dependent, and very low voltage tungsten filament bulbs could do, but their size would be unpractical, and nowadays incandescence lamps are on the way out.
.
Yes, they would work, but they would somewhat water-down the very sharp nature of the CFP. Since the C does not require any resistors, it takes full advantage of the CFP.
Darlington versions have been built, and they have a much higher intrinsic degeneration than CFP's, thus they are more or less comparable to a degenerated CFP, and the builders were satisfied with the result, but to faithfully translate the C into something bigger, the unadulterated CFP remains preferable.
Anyway, a PCB made for home-made PTC's will also accept normal 0.33 or 0.47 ohm degeneration resistors.
I'm looking at
https://www.mouser.com/datasheet/2/281/murata_10012018_PTC_Positor_PRG21BC-1485368.pdf
and up to 100 degrees C, it's temp characteristics seem to be usable for this case.
But I guess, ugly or not, will try to make more home-made resistors, and then try to match them...
The characteristic remains quite abrupt (the left-side scale is log), but with a correct sizing they could probably be usable, except the 25°C resistance is much too high.
You would need to parallel ~5 of them to arrive at a suitable value
You would need to parallel ~5 of them to arrive at a suitable value
That's what I thought. However I see now they are very small, only 2mm long, 1mm wide. How the heck I'm supposed to solder this thing..
Since they need to be paralled, I would need to make micro pcbs for each group of them, or for a bigger group for 4 transistors (say 16 resistors on one tiny pcb)..
You would also need to test the actual thermal behavior of the assembly you create, because it is an essential parameter.
That's what I did with the CR25 form factor, and once it is done, it remains valid for any build, perhaps with corrections depending on the mounting method, etc., but basically it is a good, standardized working base.
Another idea: since the wire lengths for steel wire seem to be <1 inch, you could just use PCB pads separated by the right distance and solder a straight wire between them. It would need to be covered with a silicone coating, but that would not be difficult.
The wires should not be thermally coupled, as this would impair their balance function, but a separation of >0.5*L would probably be sufficient.
Edit:
I am going to make some tests on that base, to determine the optimum lengths vs. resistance for the dissipation applicable to this case
That's what I did with the CR25 form factor, and once it is done, it remains valid for any build, perhaps with corrections depending on the mounting method, etc., but basically it is a good, standardized working base.
Another idea: since the wire lengths for steel wire seem to be <1 inch, you could just use PCB pads separated by the right distance and solder a straight wire between them. It would need to be covered with a silicone coating, but that would not be difficult.
The wires should not be thermally coupled, as this would impair their balance function, but a separation of >0.5*L would probably be sufficient.
Edit:
I am going to make some tests on that base, to determine the optimum lengths vs. resistance for the dissipation applicable to this case
Last edited:
No: no need for mechanical support, and with <1 inch length of straight wire, self inductance is not going to be an issue
I have put the idea to a quick test (maybe too quick): the results are disappointing.
I measured a ~20mm length of wire taut between two FR4 PCB pads and covered with one-component RTV silicone.
When I attempted to measure the resistance at 3A, the wire simply fused, which was unexpected.
This would leave little safety margin: the nominal rms current should be ~1.8A, but the wire already acts as a fuse at 3A..
It could be that I went too fast: I made the test immediately after having applied the silicone, meaning it was completely uncured.
I suspect that the heating of the wire caused the apparition of microbubbles around the wire, thermally insulating the wire and allowing its meltdown at a relatively low current.
I will need to remake the test, properly this time
I measured a ~20mm length of wire taut between two FR4 PCB pads and covered with one-component RTV silicone.
Code:
I(A) R(mΩ)
0 89
0.5 89.5
1 95
1.5 99
2 116
2.5 144This would leave little safety margin: the nominal rms current should be ~1.8A, but the wire already acts as a fuse at 3A..
It could be that I went too fast: I made the test immediately after having applied the silicone, meaning it was completely uncured.
I suspect that the heating of the wire caused the apparition of microbubbles around the wire, thermally insulating the wire and allowing its meltdown at a relatively low current.
I will need to remake the test, properly this time
I'll try the same.
I'm planning to have the wire soldered 2-3mm above pcb surface,
and then put the cement under and above it, so it's completely
covered.
This will take me a while, since I'm leaving for 1 week vacations..
I'm planning to have the wire soldered 2-3mm above pcb surface,
and then put the cement under and above it, so it's completely
covered.
This will take me a while, since I'm leaving for 1 week vacations..
I have redone the test, properly this time.
Everything is the same except the coating is now 2 component, high temperature RTV silicone properly cured.
The results were strictly identical, and when I attempted to make the 3A measurement, the wire blew, exactly like the previous one.
This means that the fusing limit is linked to the substrate, not the coating.
The coating is certainly necessary, but it is not critical.
The thermal conductivitity of the FR4 is insufficient compared to the alumina.
The conductivity of the PCB could be increased by a copper area, but this sounds haphazard.
The cylindrical construction seems to be safer
Everything is the same except the coating is now 2 component, high temperature RTV silicone properly cured.
The results were strictly identical, and when I attempted to make the 3A measurement, the wire blew, exactly like the previous one.
This means that the fusing limit is linked to the substrate, not the coating.
The coating is certainly necessary, but it is not critical.
The thermal conductivitity of the FR4 is insufficient compared to the alumina.
The conductivity of the PCB could be increased by a copper area, but this sounds haphazard.
The cylindrical construction seems to be safer
So basically the heat generated within the wire, has been
transferred to the resistor (via metal/soldering), right?
I guess what counts here is the mass of the resistor.
I'm assuming the PCB pads you were using were substantial in size/area?
Since it's quick, I'm gonna try the same thing with cement; it should behave better than silicone. And I'm going to use huge pads, a few of cm2.
No, apparently it remained confined to the FR4 epoxy
The thermal mass of the resistor (strictly speaking, its thermal capacity) is negligible.
What counts is its thermal resistance, and also the thermal capacity of its surroundings (the ceramic core in the case of the CR25)
Yes they were: 3.2mm dia. but that's small compared to the length of the wire (~20mm)
Unless the volume thermal conductivity of the cement is hugely better than silicone, the result shouldn't change significantly: the problem is the substrate.
It can be improved by using copper under the wire, but I am not sure it is a good idea for a number of reasons
I'm currently planning/ordering mechanical parts for the chassis/heatsinking.
If there are no further changes to the Circlo itself, I could start on assembling boards.
Now that rails will be 75V-80V, I need to find out what's the beast way to
get 12V for 2 cooling fans/temp_controller (at least 200mA).
Lm2596hv only allows max 60V input (perhaps there is a way to cascode it?).
Any ideas? If there is no easy/cool way to get 75V==>12V, I guess I'll need
to use small external psu for fans...
I am in the process of simulating a 100V rails, 600W/8 ohm version
For a lower power (<~300W), you might get away with just device changes, but some structural mods would nevertheless be beneficial
Since your total supply is going to be 150V DC, you can perfectly use a universal input wall-wart or converter operating from 100 to 264VAC (it will work equally well with DC)
I am making some changes, to accommodate the higher currents/voltages.
For a lower power (<~300W), you might get away with just device changes, but some structural mods would nevertheless be beneficial
Either use dedicated supplies, or a converter module suited to that kind of voltage.
Since your total supply is going to be 150V DC, you can perfectly use a universal input wall-wart or converter operating from 100 to 264VAC (it will work equally well with DC)
I'll wait then.
But if I can get away with re-using my original PCBs (already made), I would prefer that.
On the other hand is always fun to design new PCB.. To me, that's always most satisfying part of the project..
Ha! That's good thing. Will try it. Even my smaller circlophone should work with wall-wart then (100V DC).
Lots of fake Lm2596hv out there. Most of them are not really 'HV'.
Last edited:
An interim progress report: things are not going well.
The C topology does wonders for its intended range (typically 10 to 100W), but it doesn't scale well for really high power.
Early effect and paralleling cause huge difficulties, and I am not sure I will arrive at something worth building.... but I'll keep trying
The C topology does wonders for its intended range (typically 10 to 100W), but it doesn't scale well for really high power.
Early effect and paralleling cause huge difficulties, and I am not sure I will arrive at something worth building.... but I'll keep trying
Thanks Elvee! Don't overwork yourself 🙂
Well, if it doesn't scale, I'll go for bridged Circlo..
Just a thought - Few years ago I built an amp with lateral fets. Worked like a champ, and it was easy to build.
It occurred to me that lateral fets might be much easier to parallel, and to manage thermally in this case of Hi-Power Circlophone.
It would be much more expensive to build, but perhaps it would work nicely here...
Some thoughts:
If you use a wall wart with universal mains input for the fan, put it on the mains and not on the amp DC rails. Peace of mind regarding switching noise.
RECOM, CUI, Myrra, MeanWell etc. make AC/DC 4-pin print modules of approx. 25x25mm with a few Watts, probably easier than dissecting a wall wart.
Stay away from inrush limiting PTCs if you can. They have a monster of a voltage coefficient, modulating the Vbe with third harmonics.
Instead of paralleling, Thorens used a single MOSFET in the TEM3200 Circlo. They have 800-1400 W dissipation and come in SUPER264 or TO227 housing. The latter has internal ceramic insulation, copper mounting base plate and screw connections. The TEM3200 is the top reference amp of one HIFI magazine as described here:
6moons audio reviews: Thorens TEM 3200
If you use a wall wart with universal mains input for the fan, put it on the mains and not on the amp DC rails. Peace of mind regarding switching noise.
RECOM, CUI, Myrra, MeanWell etc. make AC/DC 4-pin print modules of approx. 25x25mm with a few Watts, probably easier than dissecting a wall wart.
Stay away from inrush limiting PTCs if you can. They have a monster of a voltage coefficient, modulating the Vbe with third harmonics.
Instead of paralleling, Thorens used a single MOSFET in the TEM3200 Circlo. They have 800-1400 W dissipation and come in SUPER264 or TO227 housing. The latter has internal ceramic insulation, copper mounting base plate and screw connections. The TEM3200 is the top reference amp of one HIFI magazine as described here:
6moons audio reviews: Thorens TEM 3200