• These commercial threads are for private transactions. diyAudio.com provides these forums for the convenience of our members, but makes no warranty nor assumes any responsibility. We do not vet any members, use of this facility is at your own risk. Customers can post any issues in those threads as long as it is done in a civil manner. All diyAudio rules about conduct apply and will be enforced.

Universal filament regulator

The filament obeys Ohm's law regardless of what voltage/current source is connected. So if you want lower current in the filament, just lower the output voltage of the regulator. If you want the current to be 10 % lower, make the output voltage 10 % lower.
~Tom

It is not so simple.

Ohm's law is always obeyed, but that does not mean that the filament resistance has to be constant.... and indeed it is not. This is because of the temperature-dependence of resistance found in any common metallic conductor.

Typical DHT filaments show an increase in resistance from cold to working temperature of 400% to 1000%, and even when at working-temperature it shows a temperature-driven rise.

Over a 10% change in voltage, some filaments show only a small change in resistance, others change substantially.

.
 
I forgot to add that I'm running the valve with filament bias, therefore I will probably need 7V.

Nifty. With the low filament current of the 4P1L and relatively low Vgk, that's even realistic. Unlike the #26 amps that require 50 W power resistors for filament bias. :)

For 2.1 V, I suggest using the "low voltage" version of the regulator. That uses the LM22673-ADJ IC and is optimized for that. If you are committed to 7 V (-ish), then I suggest building the "high voltage" version based on the LM22673-5.0 IC. You'll get the best performance that way. If you want to be able to try both, then I suggest going with the -ADJ version of the IC and see if you can find a good set of components in WebBench that'll give you decent performance at both output voltages. The circuit board is the same for the two regulators, so you could get more boards and build both versions to perform an A/B comparison.

I like your Mecano "chassis" by the way.

I will do the initial tests with the filament boards I have and will send you a PM to get yours and do a comparison, ok?

I'm not going anywhere... When you're ready, just jump on my website and place an order. Easy as pie... Just keep in mind that I close down for the holidays starting Dec 13th. Any order placed after that date will ship in early January.

~Tom
 
Ohm's law is always obeyed, but that does not mean that the filament resistance has to be constant.... and indeed it is not. This is because of the temperature-dependence of resistance found in any common metallic conductor.

That's true. Some experimentation (or knowledge of the resistance curve for the filament) is needed to obtain a given current by changing the voltage. Easily done with a lab power supply and transferred to the filament board.

~Tom
 
I finished two of the boards and tested them with load resistors and a lab supply. Using the values in your spreadsheet for the 801A I got 7.4V out which I think is perfect. I'd rather be a bit low than a bit high.

Did you use 68 uH + 100 uF (designed for use with one tube) or 22 uH + 47 uF (two tubes)?

7.4 V is that measured at the tube or on the output connector of the regulator. I think WebBench rounds down, so you could be up to 2 % low (7.35 V). But the feedback is taken at the output connector - not at the tube socket, so you will have some resistive drop across the wires going to the tube socket. Not enough to matter in practical terms, but enough to make you go, "but... I thought the output voltage was supposed to be 7.5 V"). The voltages I'm seeing are all slightly below the target value, but within +0/-2 % measured at the output connector pins. Component tolerances...

Is there any danger to the parts on the board if you power this thing up without a load, say if the tube is unplugged?

Nope. It's fine without a load. In that case, the resistive divider R1, R2 provides enough load for the regulator to remain happy.

Attached are Transient Response (1140 mA DC + 100 mA step); Line Regulation (ripple rejection) in dB; Output Impedance (1.24 A DC + 50 mA AC) in Ohm for the 7.5 V, 22 uH, 47 uF, LM22673-5.0 option.

~Tom
 

Attachments

  • FILREG_3P1_7V5_22uH_47uF_Transient.jpg
    FILREG_3P1_7V5_22uH_47uF_Transient.jpg
    208.6 KB · Views: 534
  • FILREG_3P1_7V5_1150mA_22uH_47uF_LineReg.gif
    FILREG_3P1_7V5_1150mA_22uH_47uF_LineReg.gif
    12.3 KB · Views: 523
  • FILREG_3P1_7V5_1240mA_22uH_47uF_Zout.gif
    FILREG_3P1_7V5_1240mA_22uH_47uF_Zout.gif
    11.7 KB · Views: 486
Last edited:
I was going to build one for two tubes and test it and see if there was an improvement in building a separate regulator for each, but I just got lazy and built four boards with the 68uH+100uF.

I'm using these tubes well below their max DC cathode current so I would be comfortable running them down to 7V or so.

Soldering the DAP went well. At first, I was worried because we had no large tips but those Metcal stations handle it fine even with a smaller tip. I used like a 3mm wide chisel tip. Actually, I think I have used one of the fine point hooked tips for surface mount soldering to solder an N-connecter onto a piece of semi-rigid coax before. Those are good solder stations.

Thanks for making this board. It is just so cool to be able to provide 5A of filament current and only draw half that from the raw supply.
 
Thanks for making this board. It is just so cool to be able to provide 5A of filament current and only draw half that from the raw supply.

You're quite welcome. Thanks for your feedback. I'm glad to hear the build went well.

Yeah.... The whole Power IN = Power OUT (almost anyway) of the switchers is what attracted me to the project in the first place. I've run over 2 A out of these boards and they may get lukewarm. In the 5 V, 1.4 A boards I use for the 300B's, the regulators reach 31 deg C with 20 deg C ambient temp... :) No need for huge heat sinks. No thermals to manage. Yay. I can focus on the tube part of the amplifier design.

~Tom
 
I don't see a problem per se, my personal preference is just to leave a little bit of margin in my designs. The regulator is spec'ed to handle 3 A continuously. Just make sure the ripple current in the inductor doesn't hit the saturation current for the inductor or the worst case current limit for the IC.

Whether the inductor current hits those limits depends on the input voltage to the regulator. It's pretty easy to check with WebBench.

If you're running a lot of indirectly heated tubes with the heaters in parallel, don't forget to check that you don't violate the heater-cathode voltage spec on any of the tubes. In many amps, the cathodes are at vastly different voltages, hence, if you put all tube heaters in parallel you'll violate that Vhk spec somewhere in the amp. I'm not saying it's a problem in your amp (as I have no knowledge of that) just a common pitfall to be aware of.

~Tom
 
Last edited:
The inductor specified on the 2.5A parts is rated to up to 4amps. So would the LM22679 be the only change? Or does that different part require a different inductor? (fyi, I think I'll be using a 6.3-0-6.3 trafo (radioshack) OTOH, this can be changed.. I just happen to have a few of those)

Whoa.. I just opened webbench.. It looks like it wants to design for me..

--Nevermind.. I figured out I start w/ the part you link to above, and it's simple!
 
Last edited:
My intent for the regulator was to power it with two 6.3 V windings. Your RatShack trafo will work just fine provided it can handle the power. You can leave the center tap disconnected and just use the end points of the secondary.

The input to the filament regulator board should be rectified DC so you'll need an external diode bridge and cap.

As long as the current in the output inductor doesn't reach the saturation current of the inductor, you should be fine. Webbench will tell you... The Coiltronics inductors I use are listed in Webbench. As are the electrolytic caps if I recall correctly.

~Tom
 
Ok, I've been struggling to get webbench to match your schematics, but I guess it's because my requirements are a bit different.

Looks like I will have to use a lower value inductor.. It doesn't give me the option for anything more than 15uH w/ the lm22673, and 10uH is the max w/ lm22679. There's some ripple on the output.. 33mv vs your 3.

actually, I used a 22uH inductor w/ a different footprint and ran the sim, and still get 30mv ripple. Is that simply a result of pulling 3A vs your 1.5?
 
Last edited:
The voltage ripple on the output is the ripple current through the inductor multiplied by the ESR of the output capacitor. The ripple current of the inductor is set by the inductance (lower L --> higher Iripple) and the input, output voltages.

What happens is that energy is stored in the inductor when the main switch inside the IC is on, then while the switch is off, this energy is dissipated in the load (OK, handwavy explanation, but you get the idea). As the amount of energy supplied to the load is constant, the amount of energy stored in the inductor per cycle is constant as well. The on-time, Ton, of the switch is Ton = Vout/(Vin*Fswi), where Fswi is the switching frequency. So a higher Vin means the switch has less time to dump the required energy into the inductor. So the ripple current goes up. As Vripple = Iripple*ESR, to get lower ripple voltage, you either need to pick a capacitor with very low ESR (watch for stability issues) or a higher inductance, thereby, reducing the ripple current. Whatever inductor you pick needs to be able to fit on the footprint and be able to handle the peak current without saturating.

WebBench seems to pick some optimum values. "Optimum" in this case is some combination of the total BOM cost, board area, etc. I find that I can trick it into using my values by either hand-picking them in the component list or by typing them in as custom components.

Notice the Phase Margin in the Operating Point spreadsheet? That needs to be MINIMUM 45 degrees - I tend to not go below 60 degrees. At 0 degrees PM the circuit is unstable, but with PM < 45 deg, expect some severe ringing on the transient response.

~Tom
 
I've updated the Known Good BOM spreadsheet to reflect the options tested thus far. Also, as I ran out of boards (I have one left!) I decided to add a footprint for a tantalum cap (case size 6032-28 and 7343-31) on top of the footprint for the SMD electrolytic cap (C11). This actually works quite well. Aside from this small modification, Rev. 3.1 and 3.2 are identical. And, actually, it is possible to solder a tantalum cap to the Rev. 3.1 board. It's just not as easy or pretty as with a footprint designed for it...

Universal Filament Regulator : Neurochrome.com : : Audio

Rev. 3.2 boards are available for pre-order. I expect to have boards in stock in about 10 days.

~Tom
 
Last edited:
Yeah, the internal compensation circuitry is probably different (optimized for the higher output currents) in the LM22679. This allows you to pick more optimal external components, resulting in better performance. The same is the case between the -ADJ version and the -5.0 and is why I use the -5.0 for output voltages 5 V or higher.

~Tom