SMPS for tube HT

[trobbins]

If I was to use an external power supply (ie. plug pack or laptop style) with a 12VDC output, then I would want the protective earth (PE) from the mains supply to connect to the chassis of my equipment (via the DC terminals).  But that would tie the heater to 0V of the amp.

An open-frame mains AC power supply with floating 12VDC would allow the 12VDC to be elevated, and mains protective earth to connect to chassis.  A 12V to B+ isolated output dc/dc converter (like the module I linked to earlier) would then provide B+, where the heater supply could be connected to an elevated DC point off the B+ level.

 

[ruffrecords]

If I was to use an external power supply (ie. plug pack or laptop style) with a 12VDC output, then I would want the protective earth (PE) from the mains supply to connect to the chassis of my equipment (via the DC terminals).  But that would tie the heater to 0V of the amp.

An open-frame mains AC power supply with floating 12VDC would allow the 12VDC to be elevated, and mains protective earth to connect to chassis.  A 12V to B+ isolated output dc/dc converter (like the module I linked to earlier) would then provide B+, where the heater supply could be connected to an elevated DC point off the B+ level.

OK, got it. If you had said mains earth or safety earth I would have twigged. This is what I do at present for 12V heaters. I use a Mean Well 100W 12V enclosed supply; I think the part number is LRS-100-12. At present I have a linear HT supply which also generates the elevation voltage to which the heaters are connected.

So it boils down to finding a suitable isolated dc-dc converter for the HT.

Cheers

Ian

 

[Newmarket]

Yes, but it seems bit overkill to convert the mains to a lowish dc voltage at high current only to convert it again to a voltage similar to the mains at a lower current. I guess you could start with 48V which at least can be used elsewhere.

Yes - it does seem odd to go down to, say, 12V just to swith it back up to 300V + but I guess it might make sense in terms of what is available.

The current (no pun intended ) solution I outlined above doesn't do that. The PFC (Boost) converter has an output of around 370Vdc (accounts for high line on the ac @ 230Vac + 10%) and that goes into the DC/DC stage and gets converted to 330 to 360 Vdc chassis / mains earth referenced (for different instruments - programed by the feedback ratio).

FWIW the PFC stage is designed around UC3855 ic and the DC/DC uses UC3845 together with a FOD2741 error amp.

Unfortunately I can't post schematics for commercial / employment reasons 

 

[ruffrecords]

I see from the datasheet that the UC3855 is to be discontinued.

Cheers

Ian

 

[trobbins]

Sorry about the acronyms, and nomenclature differences.  For example I'd call that Meanwell unit an open-frame mains power supply - as it requires protection by another box from users.

I must admit I found it hard to resist the <$10 price of the module I just tested.

 

[Newmarket]

I see from the datasheet that the UC3855 is to be discontinued.

Cheers

Ian

I've just had a look on ti.com and can't see any obsolescence info on it...

 

[ruffrecords]

I've just had a look on ti.com and can't see any obsolescence info on it...

You are right. I misread the list of options. The device is listed as ACTIVE so it is suitable for new design. Sorry for the misinformation.

Cheers

Ian

 

[Newmarket]

You are right. I misread the list of options. The device is listed as ACTIVE so it is suitable for new design. Sorry for the misinformation.

Cheers

Ian

No Problem.

 

[ruffrecords]

One thing I have discovered is that the pp ripple of most SMPSs is about 1% of the dc output voltage. For a 12V output this means about 120mV which is not a problem for heaters. However, a 340VDC output could have 3.4V pp at its output - definitely not suitable for a tube mic pre. Looks like we will need a version of the HT250 board specifically tailored for a rather high frequency ripple.

Cheers

Ian

 

[Tubetec]

Hmm brings up a few questions in my mind ,
What kind of strategies do you think are appropriate for making a really low noise HTsmps, Is the noise spectrum of an smps entirely high frequency mush  or is there a low frequency component also . Would old fashioned Lf chokes be a good or bad idea after a switcher
I rescued a small three rail switcher from a clapped out dvd player recently +12,+5 ,-5 volts , its looks to be of reasonable quality ,but to cut costs they left out a final LC filter on each of the rails , what kind of values might be appropriate here.

 

[trobbins]

Ian, the output B+ ripple for a "smps" could be anything, and would vary with output power level.  My little module  was well below 100mVrms on a my Kiethley 197 at 41W loading on the selected secondary tap I used that gave isolated 430Vdc.

 

[ruffrecords]

Hmm brings up a few questions in my mind ,
What kind of strategies do you think are appropriate for making a really low noise HTsmps, Is the noise spectrum of an smps entirely high frequency mush  or is there a low frequency component also . Would old fashioned Lf chokes be a good or bad idea after a switcher
I rescued a small three rail switcher from a clapped out dvd player recently +12,+5 ,-5 volts , its looks to be of reasonable quality ,but to cut costs they left out a final LC filter on each of the rails , what kind of values might be appropriate here.

AFAIK thereshould only be the switching frequency and its harmonics. Inductor values would depend on the frequency I guess.

My normal HT smoothing strategy is to use several RC stages and aim for 90dB ro 120dB of ripple reduction. At a 100KHz  (or higher) switching frequency this should not be too hard.

Cheers

Ian

 

[ruffrecords]

Ian, the output B+ ripple for a "smps" could be anything, and would vary with output power level.  My little module  was well below 100mVrms on a my Kiethley 197 at 41W loading on the selected secondary tap I used that gave isolated 430Vdc.

Good to know. I found this handy boost calculator which actually allows you to spcifiy output ripple. Seems to simply adjust the value of the output cap:

https://learn.adafruit.com/diy-boost-calc/the-calculator

Cheers

Ian

 

[gyraf]

1% ripple is probably simply a design parameter - output rectification capacitors sized to fit that.

Which means that the 1% is at rated load.

Anyway easily modifiable (because of high frequency) by adding R/C or L/C filtering.

 

[ruffrecords]

I think I will buy a MAX1771 (because it comes in a DIP package) a few inductors and a couple of meaty N channel FETs and have play just to get a feel for what goes on in these things.

Cheers

Ian

 

[mrclunk]

Nice write up on a nixie clock HT supply with the MAX 1771 here. http://desmith.net/NMdS/Electronics/NixiePSU.html
Might be of some use?

 

[Matador]

Nice write up on a nixie clock HT supply with the MAX 1771 here. http://desmith.net/NMdS/Electronics/NixiePSU.html
Might be of some use?

I already posted that earlier.

I built my own version of his design, and it's actually quite a challenge because of the high step-up's involved.

EVERYTHING matters in this design: milli-ohms of trace impedance make a big difference, so the layout is critical (especially in how vias tie traces to planes) and component selection is critical as well.  You need a fast recovery diode, and the lowest channel resistance FET you can find, and an inductor sized to the worst case output current you need to sustain. Tiny ESR differences in the output cap make a large difference in how much voltage you can squeeze out.

Ian, I would forego the DIP packages: in order to follow the recommendations you really need to stay SMD to keep the loop areas around the feedback path short and low impedance, or MAX1771 chip can oscillate due to parasitic capacitance.  Unlike a linear regulator, you can't clobber the feedback path with capacitance to fix things either. The circuit doesn't work well (if at all) on a breadboard. The only real chore with getting above 250V is that SMD components rated at those voltages can be hard to find, so if memory serves some of the output caps on my boards were PTH (or you can stack SMD caps in series as well).

Keep in mind that with 60 kHz+ ripple voltages that even small CLC filtering works quite well for even small SMD inductor sizes.  I haven't had any ringing or noise issues with vanilla common-cathode gain stages used in microphones or microphone preamps, and you can get a large amount of power out of a very tiny size with very high efficiencies.

Lastly, I powered my board with a commercial wall-wart that supplied 12V at 5A to provide over 20W of power on B+, coupled with a buck SMPS to provide 40W of heater supply (although it's easier to just stack heater coils in series and drive them directly from the raw 12V supply).  It's nice to be able to leave the mains voltage as someone else's problem in tube designs.

 

[ruffrecords]

Nice write up on a nixie clock HT supply with the MAX 1771 here. http://desmith.net/NMdS/Electronics/NixiePSU.html
Might be of some use?

Thanks for that. I have seen something similar but nothing that goes into as much detail or is as well presented.

One area I am still getting to grips with is selecting the FET. I have been looking a 500V types for a  350V supply but finding ones with a low Rdson is not easy. I know there are some exotic (=expensive) FETs out there but I would rather avoid them if possible. Suggestions would be welcome.

Cheers

Ian

 

[ruffrecords]

I already posted that earlier.

I thought it was familiar!

Ian, I would forego the DIP packages: in order to follow the recommendations you really need to stay SMD to keep the loop areas around the feedback path short and low impedance, or MAX1771 chip can oscillate due to parasitic capacitance.  Unlike a linear regulator, you can't clobber the feedback path with capacitance to fix things either. The circuit doesn't work well (if at all) on a breadboard. The only real chore with getting above 250V is that SMD components rated at those voltages can be hard to find, so if memory serves some of the output caps on my boards were PTH (or you can stack SMD caps in series as well).

I am very reluctant to abandon through hole. Back in the 1980s I was instrumental in introducing SMT into new designs but these 67 year old eyes and shaky hands are not really built to handle parts that small. Back in the 80s we were using Schottky TTL at tens of MHz so I am sure 300KHz in a DIP can be workable.

Keep in mind that with 60 kHz+ ripple voltages that even small CLC filtering works quite well for even small SMD inductor sizes.  I haven't had any ringing or noise issues with vanilla common-cathode gain stages used in microphones or microphone preamps, and you can get a large amount of power out of a very tiny size with very high efficiencies.

Good to hear

Lastly, I powered my board with a commercial wall-wart that supplied 12V at 5A to provide over 20W of power on B+, coupled with a buck SMPS to provide 40W of heater supply (although it's easier to just stack heater coils in series and drive them directly from the raw 12V supply).  It's nice to be able to leave the mains voltage as someone else's problem in tube designs.

Very interesting. Was the wall wart itself an SMPS?

Most of my designs use 12V heaters anyway so for small projects, an external 12V wall wart plus an internal SMPS for the HT is an attractive idea. The only problem with SMPS and heaters is inrush current. Many SMPS will not provide enough initial current to warm the heaters enough for their resistance to rise and get the supply out of short circuit limiting. The best supplies I have found are those with hiccup mode but even they need to be significantly over rated to be certain of starting the heaters.

Cheers

Ian

 

[Matador]

Was the wall wart itself an SMPS?

Indeed it is.  There are a plethora of inexpensive 12V high current wall adapters on the market because of the popularity of LED strips.  I purchased mine on Amazon for about $15, and it even came with European plugs.

Most of my designs use 12V heaters anyway so for small projects, an external 12V wall wart plus an internal SMPS for the HT is an attractive idea. The only problem with SMPS and heaters is inrush current. Many SMPS will not provide enough initial current to warm the heaters enough for their resistance to rise and get the supply out of short circuit limiting. The best supplies I have found are those with hiccup mode but even they need to be significantly over rated to be certain of starting the heaters.

This isn't so much an issue with PWM buck topologies, because the FET and inductor are both in series with the load.  Essentially the controller turns on the FET until the output voltage rises to the set point at which point it starts to PWM average the output.  I could see what you describe happening with a boost module, since it relies on charging the output cap through a diode after the FET switches off.  However I've never had any problems with using the raw LED supply for the heaters on any of the tubes I've tested so far.

I've probably logged 200+ hours on the scope with my design and would be happy to help if you need it.  I've tested half a dozen FET's, and since the current requirements for tube gain stages are modest, you don't need to use the $15 ones either.