Bipolar Cap Mult DC Heaters

[atavacron]

Which 3A ST part?

Sorry! STF3LN80K5 hits all the marks, cost/performance wise, for the HT pass element. Guarded tab too. 33mA x 205V = 6.8W. 20W part.

Would be nice to know a scaled down version for the phantom pass element, 672mW. Prob pushing it in T0-92 but there are some packages that size with a tab…

 

[Cqwet Dbdfte]

In Thor's circuit I'd lose the C4, move it to R1/R2, make C4 a larger value, make R1 a larger value (like 1Ox, as the gate needs no current) and is a AC leak term between input and output.
This make the FET basically grounded gate, the LT 3080 controls the FET like in a cascode thru its source.
A big RC constant can be made with R1 for a desirable slow ramp, like a GZ34.
I have used 1.5mm aluminum sheets inside steel boxes as heatsinks, invisible to outside, and easy to apply.
Using fully insulated HV FETs makes life easier too.

 

[thor.zmt]

In Thor's circuit

I noted it's not mine and that there is need for adjustment... Just to show how to use a low voltage 3-Pin regulator at high voltage, IN PRINCIPLE.

Sorry! STF3LN80K5 hits all the marks, cost/performance wise, for the HT pass element. Guarded tab too. 33mA x 205V = 6.8W. 20W part.

Thermal resistance junction to heatsink, 6.25K/W. With 6.8W this is a temperature rise of 42.5 Kelvin with a 0K/W heatsink. A 8K/W heatsink is 37.5 x 32 x 20mm in size.

This would give 100 degrees temperature rise at 6.8W and likely approach > 130 degrees temperature of the actual semiconductor. I think a bigger heatsink is needed. Around 80 degrees junction temperature is the max I'd like.

When you operate FET's designed for switching in linear mode they are prone to "hotspotting" which can lead to catastrophic failure when the "hot spot" exceeds permissible junction temperature. And no, this is not in the datasheet. In fact, the datasheet does not cover linear operation at all.

With the PSU insides at elevated temperature, let's guess 40 degrees, we need a 0K/W heatsink, so this design is thermally not possible.

Thor

 

[Cqwet Dbdfte]

That STFL3 is a vanilla enhancement mode N-MOSFET.
I would use something bigger, never get close to mfg. limits. Say min 60W, not too big as the drive capacitance can get difficult, but like under 1000pF should be OK.

 

[moamps]

Sorry! STF3LN80K5 hits all the marks, cost/performance wise, for the HT pass element. Guarded tab too. 33mA x 205V = 6.8W. 20W part.

That's load's power. The power dissipated on the FET is 230*33-205*33 (mW), IMO.

 

[thor.zmt]

That's load's power. The power dissipated on the FET is 230*33-205*33 (mW), IMO.

My bad, I did not read carefully enough.

Less than 1W, that should be fine with no heatsink (62.5K/W).

I am not used to people using load power as part of rated max dissipation.

Thor

 

[Cqwet Dbdfte]

With a decent size FET and big heatsink shorted output will be handled gracefully.

 

[atavacron]

I am not used to people using load power as part of rated max dissipation.

That is a very nice way of saying I erred

 

[atavacron]

That's load's power. The power dissipated on the FET is 230*33-205*33 (mW), IMO.

 

[atavacron]

With a decent size FET and big heatsink shorted output will be handled gracefully.

Right, now I gotta calc that in the events of A) a mangled PSU-to-rack cable and B) a short on a single channel card. HT limits would be ~130mA and ~33mA, around their respective mosfets (right?). PSU and card heater limits are determined by their respective regulator specs I assume. The PSU’s HT pass mosfet shouldn’t even get warm in typical use, as it’s only dropping the ripple erased by the RCRC.

The PSU case is the heatsink, I could choose one with fins on the side. The channel card I’m not too sure about. I was gonna have each card on a sled that goes into the main chassis. Maybe the pass elements and regulator on the card couple to the sled through PCB cutouts. The chassis is like any of the 1U 3-slot 500 series chassis you see out there, but 100mm x 200mm cards and no backplane — just a shell with a single male 4-pin XLR and jumpers to get power to the cards. Front panel of the sled holds controls, back panel of the sled holds XLRs.

Oh and @thor.zmt I will digest the schemo you reposted and the modifications to it that you and others are suggesting, keep ‘em coming. Thanks!

 

[Cqwet Dbdfte]

HT pass mosfet shouldn’t even get warm in typical use, as it’s only dropping the ripple erased by the RCRC.

The FET + linear reg will eat more ripple than the RC, 40 - 60dB more, but a small 20uF + 100R helps a lot. A new "C4" of like 10uF mylar on FET gate does the heavy lifting, besides slow ramping

 

[thor.zmt]

Right, now I gotta calc that in the events of A) a mangled PSU-to-rack cable and B) a short on a single channel card. HT limits would be ~130mA and ~33mA, around their respective mosfets (right?). PSU and card heater limits are determined by their respective regulator specs I assume.

You could add polyfuses (they come in 230V AC) on each PCB, rated for the hold current that you want as maximum continuous DC current.

The PSU’s HT pass mosfet shouldn’t even get warm in typical use, as it’s only dropping the ripple erased by the RCRC.

Remember to account for high/low line in all calculations. if you sell in the EU the ranges are wider than in the USA.

I tend to use a Excel Spreadsheet to track all data for the various components and then use a formula and traffic lighting to see how close to max or derated max I am.

Oh and @thor.zmt I will digest the schemo you reposted and the modifications to it that you and others are suggesting, keep ‘em coming. Thanks!

If you want to use this kind of HT regulator (there are also options to use [say] NE5532 and series Pass FET's or BJT's), or my suggestion (CCS + Zenner string followed by RC) I would suggest to open a dedicated thread.

While they can be reliable and short circuit proof, they have a lot of potential to become things that go bang in the night, despite being far from the frontlines. Of course, dedicated HV components are as bad. A slipped probe and you get a face full of plastic shards and silicone smoke and the part has a small crater in the front.

The combination of high voltages and (relatively high) currents is something that silicone is just not that good for, hence most people stick to Tube regulators for the HV.

Something seriously leftfield is to use choke input with CLC filtering after and a (hybrid) shunt regulator on the output. No CCS, the Choke input PSU is in effect the CCS.

Thor

 

[Cqwet Dbdfte]

In what way?
Already posted answer.
IINM, the OP wants to use a depletion MOSFET.

Which one? And already answered

 

[abbey road d enfer]

Already posted answer.

And not satisfactory. Although secondary currents are different between bridge and CT, the actual efficiency is the same, so there's no "waste". Not 30% difference.

 

[Cqwet Dbdfte]

...a waste of time.
Thanks for your opinion.

 

[abbey road d enfer]

...a waste of time.
Thanks for your opinion.

Thank you, Your Highness...

 

[moamps]

In what way?

IINM, the OP wants to use a depletion MOSFET.

Two-diode rectification is less efficienct than bridge rectifier. The ripple current in each half winding is 1/sqrt2 times the total RMS. For the same windings (two in parallel with a bridge, versus the same two in series with 2 diodes), you can only pull about 70% as much DC out compared with a bridge.

Which one? And already answered

It seems to me that Transformer Utilization Factor is being discussed here.

https://electricalbaba.com/transformer-utilization-factor/

It is 0.672 for CT full rectifier and 0.81 for bridge full rectifier. The ratio is 0.672/0.81=0.83. This relationship is not so significant in this case because the mains transformer (100VA) has enough power for both versions and high redundancy. The peak voltage at the output of the rectifier is far more critical here, because the voltage redundancy is too small for the good operation of the two LDO regulators. Here it should be kept in mind that the peak charging current of the capacitor is high, and I would always assume that the voltage on the schottky diodes will be around 0.5V. The situation will be worsened by the fact that the mains voltage can be below the nominal, and that it can have significant THD of odd harmonics, and that the coefficient 1.41 is no longer valid.
In this sense, I would adhere to Abbey's advice to use a CT full wave rectifier if such a design is already used, although nowadays I would definitely consider the solutions represented by Thor.

 

[abbey road d enfer]

It seems to me that Transformer Utilization Factor is being discussed here.

https://electricalbaba.com/transformer-utilization-factor/

I have serious doubts about this document.
For example: "Primary winding current of transformer is pure sinusoidal as current flows for both positive and negative half cycle irrespective of the fact that which diode conducts. Thus the rms value of transformer primary current is (Im/√2)."
This is pure nonsense, since the primary curent is constituted of pulses occuring at the top of the incoming sinewave.
The notion of "individually find the VA rating of transformer primary and secondary and then take average of their values." is simply ludicrous.
And the main limiting factor in all cases is the magnetic core, not the windings.

 

[Emmathom]

My bench version of this 4-channel preamp is probably gonna be a Meanwell for heaters, a Meanwell for phantom, and a prebuilt Maida for the HT off of a basic 25-30VA trafo, so I can focus on the circuit itself. But ultimately I’d like to see if it’s convenient to do a full linear supply with just that Antek hexfilar and everything on one PCB.

I just finished a quad mic pre using 12AY7 into 12AU7 as AF. 2 boards of 2 channels.
I used 2 Meanwell RS 15-12 (12v • 1,3A) for heating 2 tubes per board so 300mA current consomption per board (when hot cause the heaters deserve more current when cold > a simple 1,3A couldn't feed 4 tubes at starting and it went on standy)
And 1 Meanwell RS 35-48 (48v • 800mA) for phantom power and everything's doin' fine without any "noise"

 

[moamps]

I have serious doubts about this document.
..........
And the main limiting factor in all cases is the magnetic core, not the windings.

That article is obviously made for resistive loading only, and is correct in that context IMO.
If we decompose the impulse charging current when the load is a capacitor and a resistor into a certain number of sinusoidal harmonic components, we can find power ratios for each of them using the same principle and add all these individual ratios and we will probably get the same result as if the current were sinusoidal.
As for the limitation in the transmission power of the transformer core, it is not so critical here IMO because there is a capacitor bank as an energy reservoir, so only the conduction time of the rectifier diodes will be longer, and the impulse charging current will be lower in peak value and wider in duration.