So it seems that most of us agree on my finding that SMPS are the way to go for hungry filaments like 304TLs. So, the question is which one specifically? Anyone tried - tested something good? Any proposal?
I would say go for it. I have with GM70 as said.
You may have trouble with the SMPS not wanting to start into the almost short circuit of a cold filament on big valves like 304TL. That's where a current source or Rods regulators in parallel will help. You can minimise dissipation in the linear part by tweaking the output voltage of the SMPS.
Cheers
Matt.
You may have trouble with the SMPS not wanting to start into the almost short circuit of a cold filament on big valves like 304TL. That's where a current source or Rods regulators in parallel will help. You can minimise dissipation in the linear part by tweaking the output voltage of the SMPS.
Cheers
Matt.
Panos, modern Dell server power supplies are generally around 12V at multiple tens of Amps. Could you not run the 304TL filaments at 10V and lose the rest in a CCS?
I have some very modern ones here that are 12V 97A in a case about a sixth of the size of a shoe box. They run up around 970Khz-1Mhz in the switching stages. Also use synchronous rectifiers. Very efficient. There is a fair bit of information on these things on the radio control forums as they like to use them for charging their batteries.
Those guys use them flat out but as you won't be I don't think there will be a need for the fan. There are also ways via the (variably) standard Dell PSU socket to vary fan speed if fitted at all.
Cheers
Matt
I have some very modern ones here that are 12V 97A in a case about a sixth of the size of a shoe box. They run up around 970Khz-1Mhz in the switching stages. Also use synchronous rectifiers. Very efficient. There is a fair bit of information on these things on the radio control forums as they like to use them for charging their batteries.
Those guys use them flat out but as you won't be I don't think there will be a need for the fan. There are also ways via the (variably) standard Dell PSU socket to vary fan speed if fitted at all.
Cheers
Matt
Last edited:
I use industrial power supplies made for convection cooling. Computer ones need forced air, and can not be cooled without certain unacceptable noise.
Of course you are correct Wavebourn but these guys generally have two boards. The input rectifier/PFC board and the main board. By opening them up and allowing good convection whilst also running way under flat out they are fine.
Also forgot to say that server PSU's are designed to run 24/7 without failure. Out of possibly (as it varies in our DC) 15000 Dell power supplies I have seen maybe 20 failures in the last 4 years. All of those have been over ten years old and all failures have been the input smoothing cap after the PFC.
Of course I take these things to the bench and do an autopsy.
Cheers
Matt
Also forgot to say that server PSU's are designed to run 24/7 without failure. Out of possibly (as it varies in our DC) 15000 Dell power supplies I have seen maybe 20 failures in the last 4 years. All of those have been over ten years old and all failures have been the input smoothing cap after the PFC.
Of course I take these things to the bench and do an autopsy.
Cheers
Matt
Last edited:
I tried Compaq (HP Proliant server) PS for a pair of Gu-13, can't live without forced air. Too small transformer and other parts that need cooling.
Here is amp standing on PS-es that I rejected in favour of the one that is currently under the chassis. 🙂
https://www.facebook.com/Wavebourn/videos/1157283967663236/
Last edited:
We have some HP kit. Yeah it's ok but that's it.
I do agree on using an industrial supply as I came from that background for nearly 20 years.
Modern high reliability switchers are out there and they need hardly any cooling.
Cheers
Matt.
I do agree on using an industrial supply as I came from that background for nearly 20 years.
Modern high reliability switchers are out there and they need hardly any cooling.
Cheers
Matt.
That random hf noise is most likely the smps going into burst mode. Smps do that if they are lightly loaded. For this reason and for best efficiency, one must select a smps that is running 70-99% of its rated power. Only then are u guarantied that they switch at their fixed higher frequency.
So if u need 5v and 25amp select a smps that can give 30amps not much more.
You have to be sure the chosen supply will start up into a filament - or use external current limiting.
Big Transmitter filaments are nearly a short-circuit when they are cold, and will try to take 4 - 10x the hot running current. I doubt that many 30A converters will start-up into a 25A rated filament, without the current-limit protection circuit locking-out.
Big Transmitter filaments are nearly a short-circuit when they are cold, and will try to take 4 - 10x the hot running current. I doubt that many 30A converters will start-up into a 25A rated filament, without the current-limit protection circuit locking-out.
Good point Rod, but many smps have over current protection in the form of that they turn into current sources at rated current, so a 30A supply will stop at 30A and probably source that long enough till the filaments are up. Most supplies handle several 1000 uF capacitance and therefore can handle the initial short on power up.
We need real data, to see if that makes sense.
Firstly, let's fire 30A into 1000µF:
C*V=i*t
(1e-3 *10)/30 = t = 330µs
i.e.: this level of current will ramp the voltage to 10V in about 330µs! This time-constant is of-the-order of 10K times (x 10^4) too short to be relevant to a Transmitter Filament, so it will not give us confidence that the startup is feasible.
In my industrial electronics-design life, I turn to XP Power for so-called 'off-line' (mains→dc) converters, so I looked at their line-up for 200-400W converters.
In the high-efficiency class, the CCB200 is the 200W module:
http://www.xppower.com/Portals/0/pdfs/LF_CCB200.pdf?ver=2015-12-11-110327-590
Reading through this data, on page 2 we find:
- Overload current trips the supply at 110 -150% of rated current;
- That the short-circuit protection mode is Auto Trip and Restart (hiccup mode).
Sadly it does not state the restart-time, but given that the trip current is substantially above the rated current, it is fairly safe to assume that a cool-down time is needed to protect the power semiconductors. It seems unlikely that the filament will heat under such circumstances, and would certainly need careful testing before we say that it can.
Now, I am not declaring that you can't find a supply that would solve this problem, but I am saying that you need to check with real data carefully before buying one.
Rather than oversize the converter to handle the start-up current which will increase the absolute value of losses -- and/or cause the converter to drop into burst mode, I would prefer to add external current-limiting. This will also allow the desirable high-impedance feed to the filaments, to prevent the music-signal across the filament from being crushed. And reduce the differential noise, of course.
Firstly, let's fire 30A into 1000µF:
C*V=i*t
(1e-3 *10)/30 = t = 330µs
i.e.: this level of current will ramp the voltage to 10V in about 330µs! This time-constant is of-the-order of 10K times (x 10^4) too short to be relevant to a Transmitter Filament, so it will not give us confidence that the startup is feasible.
In my industrial electronics-design life, I turn to XP Power for so-called 'off-line' (mains→dc) converters, so I looked at their line-up for 200-400W converters.
In the high-efficiency class, the CCB200 is the 200W module:
http://www.xppower.com/Portals/0/pdfs/LF_CCB200.pdf?ver=2015-12-11-110327-590
Reading through this data, on page 2 we find:
- Overload current trips the supply at 110 -150% of rated current;
- That the short-circuit protection mode is Auto Trip and Restart (hiccup mode).
Sadly it does not state the restart-time, but given that the trip current is substantially above the rated current, it is fairly safe to assume that a cool-down time is needed to protect the power semiconductors. It seems unlikely that the filament will heat under such circumstances, and would certainly need careful testing before we say that it can.
Now, I am not declaring that you can't find a supply that would solve this problem, but I am saying that you need to check with real data carefully before buying one.
Rather than oversize the converter to handle the start-up current which will increase the absolute value of losses -- and/or cause the converter to drop into burst mode, I would prefer to add external current-limiting. This will also allow the desirable high-impedance feed to the filaments, to prevent the music-signal across the filament from being crushed. And reduce the differential noise, of course.
Last edited:
Beyond my paygrade to design an off-line switcher, although I have played with DC-DC converters. Some of the newer DC-DC switches from Linear Tech have "soft-start" capacitor.
Agree Rod and a current limiter is a good idea both saving the filament from the strain of an angry current surge, but also allowing the smps to operate within specs.
jackinnj, the soft-start is often only set for some tens of us, but if you have the schematic and a microscope so you can see the tiny cap, it may possibly be replaced with a bigger value for longer soft start period. Unlikely info available to most buyers tho.
jackinnj, the soft-start is often only set for some tens of us, but if you have the schematic and a microscope so you can see the tiny cap, it may possibly be replaced with a bigger value for longer soft start period. Unlikely info available to most buyers tho.
These trully briliant Dell SMPSs outputing 12v can be adjustedto 10V easily.Question is: if I run filaments in series I am not going to trouble biasing the internaly paralled triodes in 304TL with different bias ?
Fortunately I had power transformers with 12.6V windings, it was enough to help SMPS to start when tubes are cold (a Shottky bridge) for GK-71 tubes.
I would gamble that the hick-up mode on start will slowly but surely get the filaments up to a level where the resistance comes up to nominal. 'Slow' in smps world is probably just some tens of ms. The smps I've played with into much too large capacitances worked that way but maybe I have been lucky.
Without 12.6V help for a couple of GK-71 a 150W SMPS hicks up forever; with help of 12.6V and Shottky bridge it ticks 3 times then tubes glow.
I've done it in the past. My supplies used an optic isolator as feedback on the 5 volt section. I added a trim pot to tweak the voltage. Just keep in mind that watts are still watts. You'll have to derate the current accordingly.
Put another way, with any DC filaments all parts of the filament are operated at different bias.
All good fortune,
Chris