XL6009 based DC/DC Converter for sensitive audio voltages like e.g. phantom power?

[rock soderstrom]

Hi guys,

modern switched-mode power supplies are slowly arriving in my archaic tube audio world. Lately, I've been trying to question my cherished habits and take a different approach to realizing my projects.

Out of curiosity, I bought two of these DC/DC up converter boards on aliexpress.

https://de.aliexpress.com/item/1005...t_main.11.21ef5c5fshABJR&gatewayAdapt=glo2deu

As supplied, these boards can convert over 38V from 9V. The promised 52V can only be achieved after a feedback resistor hack.



52VDC minus some filtering makes good 48V?

What do you think, are such boards suitable for obtaining a clean phantom voltage from lets say the heater voltage (or a separate DC voltage)?

Could you heat an unusual single tube (P or U tubes) with these boards to operate them with other conventional E-tubes?

...or would it be better to stay away from such solutions, as they contaminate everything in their vicinity with HF noise?


20mV/5muS per division, Vout=24V

What do you think? Does anyone have experience with this?



PS: I'm a beginner with this stuff and have almost no experience with it. So, dear experts, bear with me. (Yes, you too Khron )

 

[rock soderstrom]

I played a bit more with the DCDC converter board and would like to show my new findings here.

I try simulate a phantom power supply, since the board is not yet hacked (I don't have the right SMT resistors yet), here only with approx. 38V instead of 48V.


My test circuit looks like this, I have installed an RC filter to see what effect I can achieve with it.



What is noticeable (at least to me) is that the switching peaks of the DC DC board rise sharply with a light load (such as an LDC microphone).

Comparison idle vs light load without filter: (20mV/5muS)


Here now with light load and 50mV per division

Now the 470R/4700uF filter comes into play and clearly shows its effect (back in the original 20mV setting!)


To make the signal more visible, now in the highest resolution of my scope 5mV.

For me this is pretty interesting

 

[rock soderstrom]

My learning odyssey continues, this time a test to use the DCDC board as a heater supply for a nice PCL86.

According to the data sheets, the PCL86 needs 13-14.5V with series heating at 300mA. What if you wanted to use this tube (as an example) in a device with only 6.3V DC tubes?


Let's try it out. It works for the most part without any problems. My lab power supply provides the 6.3V DC, the tube and the DC board together draw 0.75A. The output voltage of the board is exactly 13V. The chip stays cool after the (hot) start

This is the only problematic point, the moment of power up, the cold tube is at first a short circuit and makes the XL6009 sweat a lot, it reacts with a soft start mode and draws a whopping 2.5A or more.

After a short time this normalizes, a small series resistor for starting could bring advantages.

The chip is tough, I have tortured it several times and it is still alive.

The scope picture now looks like this, at 0.2V/ 5muS per division, without further filter.

 

[rock soderstrom]

Now with my single-stage, super inefficient RC filter with 18 ohms and 4700uF. Its a disaster in terms of energy, too. The laboratory power supply now delivers 1.26A at 6.3V, because I had to increase the output voltage of the DCDC board by 5V, I had no better suitable resistor of this size. Very warm affair...still 0.2V per division. The chip still stays cool.

Learning by doing.

 

[rock soderstrom]

1muS per division


Edit:

I think this is a valid way to replace the now very expensive ECL86 with the much cheaper PCL86 in old radios and R2R recorders. These boards are very inexpensive and small. There are versions without a display that are even smaller and cheaper.

 

[nashkato]

According to the data sheets, the PCL86 needs 13-14.5V with series heating at 300mA. What if you wanted to use this tube (as an example) in a device with only 6.3V DC tubes?

i guess one would use ecl86 then

 

[rock soderstrom]

i guess one would use ecl86 then

Take a look at today's prices and availability and BTW it was just an example (as mentioned). Thanks for feedback

 

[nashkato]

i'm very sorry if i interrupted your oh so very serious scientific monologue with my ridiculous comment.
stupid,stupid , stupid me !

 

[thor.zmt]

52VDC minus some filtering makes good 48V?

Maybe.

What do you think, are such boards suitable for obtaining a clean phantom voltage from lets say the heater voltage (or a separate DC voltage)?

Their frequencies are low. The XL6009 is a trifle up on most at 400kHz, this still meand triple the values of inductors and capacitors needed at 1.2MHz.

Or in another view, a typical generic design with 08/15 parts values will have 3 times less noise 400kHz vs 1.2MHz, nearly 10 times at 150k vs 1.2M.

The stupid LED Displays add extra noise BTW, never buy anything with these stupid things.

Usually these boards use the cheapest generic electrolytic capacitors, where even the Chinese datasheet calls for ultra low ESR solid electrolyte types. So this needs fixing. Noise will be less.

You basically need to rebuild the board, add filtering etc. to make them usable. It may be worth it in a one off, where alternatives are scarece and only one unit is used.

Using multiple independent SMPS in the same device tends to cause beat-notes (birdies), so if you use more than one, all need synchronising. But these boards usually lack any sync option.

I commonly use ~ 1MHz and up. For a receiptnt design, I used RT9297 @ 750kHz externally synchronized as step-up to 24V with a charge pump voltage doubler to 48V

Post Filter is CLCLC. It is as quiet as a good linear PSU.

It would not be hard to make a community project of a power supply to cover a small to medium size tube/hybrid circuits with +A, +B, -C (heater, HT, bias) and P48, +/-18V etc. from a USB-C GAN 100W PSU (~ 90W max out), that can be fabbed at JLCPCB.

I suggest that before, but all I got was "nah, meanwell switchers are all I need".

Could you heat an unusual single tube (P or U tubes) with these boards to operate them with other conventional E-tubes?

Yes, you could, with all the caveats above.

...or would it be better to stay away from such solutions, as they contaminate everything in their vicinity with HF noise?

IF the design is competent (not a given in a cheapo Chinese product), these supplies are a great alternative to AC off mains. They can step up, down, create +/- Rails and so on, all from a commercial commodity USB-C power supply with no practical power limited for small projects.

The basic designs are decade old concepts, it is all well understood.

Thor

 

[rock soderstrom]

Thank you for your detailed and qualified feedback! There are some very interesting aspects.

The basic designs are decade old concepts, it is all well understood.

Let's start at the back, that's right. For me, the whole topic was very new and I'm still learning, no doubt about it. So far I've had a few “wow” moments in the last two days and the correlations are starting to become a bit clearer.

Their frequencies are low. The XL6009 is a trifle up on most at 400kHz, this still meand triple the values of inductors and capacitors needed at 1.2MHz.

Or in another view, a typical generic design with 08/15 parts values will have 3 times less noise 400kHz vs 1.2MHz, nearly 10 times at 150k vs 1.2M.

Okay, I'll take that with me now.

The stupid LED Displays add extra noise BTW, never buy anything with these stupid things.

I already suspected that, you can also deactivate (and “calibrate”) them, but they will probably still make noise. You should just leave them off, you don't need them in normal operation.

Usually these boards use the cheapest generic electrolytic capacitors, where even the Chinese datasheet calls for ultra low ESR solid electrolyte types. So this needs fixing. Noise will be less.

You basically need to rebuild the board, add filtering etc. to make them usable. It may be worth it in a one off, where alternatives are scarece and only one unit is used.

Thanks for info, the component quality will match the extremely low price (around 1€ per board). During my research there were also some indications that some of the chips are counterfeited, there are also Youtube videos about this. These boards could certainly be optimized, but everyone has to decide for themselves whether this makes sense or not. I have already found some applications for myself where even these boards can make sense. Auxiliary voltages for things like e.g. relays that you want to install in existing devices would be conceivable, or even exotic heater voltages.

Using multiple independent SMPS in the same device tends to cause beat-notes (birdies), so if you use more than one, all need synchronising. But these boards usually lack any sync option.

That's right, I can confirm that. My first tests show exactly that, I connected one of the boards directly to an existing SMPS power supply of a tube-spring reverb. The result was quite musical...

I commonly use ~ 1MHz and up. For a receiptnt design, I used RT9297 @ 750kHz externally synchronized as step-up to 24V with a charge pump voltage doubler to 48V

okay, interesting.

 

[rock soderstrom]

Post Filter is CLCLC. It is as quiet as a good linear PSU.

Better filters clearly score points here, perhaps it also makes sense to choose an active solution?

It would not be hard to make a community project of a power supply to cover a small to medium size tube/hybrid circuits with +A, +B, -C (heater, HT, bias) and P48, +/-18V etc. from a USB-C GAN 100W PSU (~ 90W max out), that can be fabbed at JLCPCB.

I suggest that before, but all I got was "nah, meanwell switchers are all I need".

This is certainly a very interesting idea and honestly I have been wondering for some time why there is no such DIY (or commercial) solution to buy? I looked around, there are very few combined SMPS solutions especially for tube applications...

The idea is very promising, as the needed requirements are very similar for many different projects. This is interesting even for a rather conservative iron, glass and copper guy like me.+

I believe that a clever design is at least competitive in terms of quality, flexibility, effort and cost, if not better in (almost?) every respect

IF the design is competent (not a given in a cheapo Chinese product), these supplies are a great alternative to AC off mains. They can step up, down, create +/- Rails and so on, all from a commercial commodity USB-C power supply with no practical power limited for small projects.

The basic designs are decade old concepts, it is all well understood.

+1

 

[rock soderstrom]

It would not be hard to make a community project of a power supply to cover a small to medium size tube/hybrid circuits with +A, +B, -C (heater, HT, bias) and P48, +/-18V etc. from a USB-C GAN 100W PSU (~ 90W max out), that can be fabbed at JLCPCB.

Even at the risk of repeating myself. I would very much welcome it if this idea becomes reality. I have no idea how I can help, but I hope more people jump on this bandwagon.

Think about how a clever solution could make life easier. You set what you need and take care of the (amp) circuit instead of building a dedicated power supply for the umpteenth time.

 

[ccaudle]

20mV/5muS per division, Vout=24V

I try simulate a phantom power supply

Notice the high frequency spikes in the scope display. The way you have wired the components together with clip leads makes large (high inductance) loop antennas, so you cannot trust the measurements. Any time you are working with switching power supplies you ideally have components wired to a solid copper plane, or at the very least keep the wire connections tightly twisted and the components located as closely together as possible.
With your setup as shown you are measuring not only the electrical noise of the regulator, but also the sensitivity of the loop antenna formed to the magnetic fields from the regulator.

 

[thor.zmt]

Better filters clearly score points here, perhaps it also makes sense to choose an active solution?

That depends on the precise active solution. With (R)LC we have a very predictable results.


Above is one of my SMPS Designs.

It is to > 90% a "cooking" SMPS based on a Fairchild/On-Semi Chip with 67kHz switching frequency, app note circuit.

The key differences are in small details.

Instead of a heavily "slugged" loop with TL431 and a turnover at a few 100Hz for maximum stability, this uses a triple loop design, optimized for maximum reliably stable loop gain. Predictably this lowers noise a lot more over a much wider bandwidth, while still using TL431, just a few RC are used to "bend" the amplitude and phase response "just so".

Triple loop means that there is a "fast loop" that closes for high frequencies above the turnover set by the 431, a "medium speed loop" where the 431 AC behaviour dominates. Both of these loops connect to the first foler after the rectifier.

And finally a "slow loop" which is for DC and up to few 100Hz (this mirrors the usual "slugged single loop") that connects after the LC filtering and maintains DC and signal envelope dynamic control.

There is C+C -> L + C -> L+L + C filter after the rectifier diiode (it's s flyback converter) with DC feedback from the output. Without the triple loop design this would not be stable. So we get a 4th order lowpass function for switching noise.

Theseeasures dramatically lower noise. As is, using just a few extra L/C and tiny SMD R + C over a completely generic design, noise in the 10's of microvolt, instead of 10's of millivolt at rated load.

The transformer is wound in ways that allow the "Y" capacitor to be removed (the lab insisted on having one so 100pf is fitted instead on 2.2nF but could be removed) without causing interference. PCB layout is optimized for the smallest possible noise loops.

Finally, as coup de gras, a small PCB is fitted (bottom left hand corner, vertical PCB), that acts as active noise canceller on the output, which lowers noise to around the self noise of a NJM5534, which is used as the main loop element of this circuit.

So yes, active filtering can work well, you just need to make sure the solution doesn't cost too much power and is efficient at all frequencies that need to be dealt with.

So, using pretty generic SMPS circuits with just a smidgen of Knoff Hoff, as we say in the Fatherland, can give a large Vorsprung durch Technik.

For reference, here an AC output spectrum of the above PSU at 50/100% current.



For comparison a "low noise linear supply" from a Chinese Vendor:



From here:

http://audiohobby.pl/index.php?topic=14751.0

This is certainly a very interesting idea and honestly I have been wondering for some time why there is no such DIY (or commercial) solution to buy?

It's pretty trivial. Another one of my designs:



This is an EL84 PP Amplifier (Think Dynaco ST-35 approximately) with some solid state sections for DAC (768kHz/32Bit + DSD512), Phono (including low output MC pickups) etc.

It used a SMPS (again based on Fairchild/On-Semi IC's) providing +B, Heaters, Logic Supply and +/-12V for solid state circuitry of DAC & Phono stage.

These days I'd use synchronised SMPS with Heater and HT separate, simply because it was such a PITA to make that design from going "kaboom" at turn on, when the heaters are cold.



Again, extensive LC filtering kills noise very efficiently.

Idle draw just low enough to pass the Mains Harmonics regulations a decade ago.

A full service manual is at hifi engine, if anyone is interested in the design details.

I looked around, there are very few combined SMPS solutions especially for tube applications...

I noticed.

Every time I propose something, nobody wants anything to do with this.

Except one or two people, who sadly cannot help making a community open sauce design with some input from me and who are far to few to make this a valid revenue project.

The idea is very promising, as the needed requirements are very similar for many different projects.

As SMPS are regulated, selecting different Heater and HT voltages is trivial. Each Switcher IC is current, voltage and thermally limited to deliver a certain maximum power.

If we set the output to 120V or 360V only matters insofar that 360V will have less (1/3rd) maximum current.

Thus such a PCB can be very universal. Multiple heater voltages, all selectable 6.3V (should be set to ~ 5.7V), 12.6V (should be set to ~ 5.7V) or arbitrary values with up to 4A are trivial.

HT is a little less trivial, but still not that hard, Greinacher showed a good option.

Efficiency is usually > 90%, as we can easily find solutions capable of sustained multiple 10's of watt, there is not much of a real limit in this.

Thor

 

[trobbins]

The XL6009 module is quite versatile, but it uses 50V e-caps so perhaps don't try and go above circa 40V. I've used it with a spare 5Vac heater winding to generate a negative bias voltage up to about 40Vdc max. If using a heater winding then you may need to use a Schottky diode rectifier, and add extra input e-cap to augment the on-board 220uF 35V e-cap to generate circa 6Vdc and avoid UVLO and to suppress input mains frequency ripple sufficient to keep output ripple below about 30mVrms depending on load. I measured output ripple for my application at about 90mVpp 400kHz ripple spikes that can be attenuated by an LC filter (preferably an smt L part of a few hundred uH). The module may show regulation instability for certain levels of input and output voltage – one example showed instability for input between 8 to 14Vdc with no output load and output set for >28Vdc – the instability was an output saw-tooth ripple voltage of circa 100-150mVpp at circa 90-330Hz, with 400kHz ripple spikes on rising edge of saw-tooth, but that is an unlikely situation.

For low loads like phantom powering or just 1-2 12AX7 heaters, then I'd recommend a MT3608 based small cheap module.

When powering heaters then I'd not worry about initial hiccup or other quirky operation as the heaters come up to temp, that is to be expected and of no tangible stress.

For higher B+ dc supplies then I'd recommend the 'small 'Royer' style module with its full bridge output (or change to doubler) and with added input and output e-cap filtering, or a more powerful inverter module with multiple tap secondary that can be rectified and filtered for a range of B+ levels. Both supplies need a fast diode like UF4007, and minimal loop layout of parts.

Such cheap ebay modules are an excellent diy choice imho, and on close inspection of pcb layout and operation perform as well as can be expected of whatever switchmode tech is being used. About the only aspect I've noted is judicious addition of smt caps in some locations to minimise switching current loops to smallest practical region on the pcb.

Btw, when showing any type of switchmode waveform you really should show a photo of your probing scheme/setup, as you are likely not probing correctly.

 

[rock soderstrom]

Hey @ccaudle , @trobbins and @thor.zmt thanks for your feedback! I really appreciate that, even if I don't (yet) always understand everything. I am slowly feeling my way forward...

The way you have wired the components together with clip leads makes large (high inductance) loop antennas, so you cannot trust the measurements.

Many thanks for the tips. I think you can see that I took a pretty unconventional approach, to put it bluntly, I just put it together and started playing around with it to gain experience.

The measurement results, like the whole setup, are certainly not suitable for generating reliable data. However, as the conditions for all the “measurements” were equally bad, I can at least compare the results with each other. The criticism is of course justified.

I don't normally work with high radio frequencies, I try to avoid them in my amps . I realize that 400KHz switching frequency means exactly that. My research has shown that in this frequency range, for example, NDB (Non-Directional-Beacon, or german "Funkfeuer") takes place, i.e. radio transmitters that are used in the aeronautical or maritime sector as a basis for navigation.

The consequence is probably that every conductor (wire, measuring lead, conductor track, crocodile clips, component legs, etc.) becomes an antenna, which should be avoided as best as possible. This is probably also the reason for the preferred use of SMT components.

IMHO the good thing about radio waves is that they can be shielded relatively easily. I used a simple SMPS in a tube spring reverb and was able to protect the extremely sensitive transducers of the spring pan from interference with a sheet metal box around the PSU, even though the distance was very close.

Btw, when showing any type of switchmode waveform you really should show a photo of your probing scheme/setup, as you are likely not probing correctly.

Thank you, that goes probably in the same direction. What needs to be considered, how do you do it right?

The XL6009 module is quite versatile, but it uses 50V e-caps so perhaps don't try and go above circa 40V.

Yes, I noticed that too. For further experiments I will install better caps (see @thor.zmt ) and also with a higher voltage rating.

I've used it with a spare 5Vac heater winding to generate a negative bias voltage up to about 40Vdc max. If using a heater winding then you may need to use a Schottky diode rectifier, and add extra input e-cap to augment the on-board 220uF 35V e-cap to generate circa 6Vdc and avoid UVLO and to suppress input mains frequency ripple sufficient to keep output ripple below about 30mVrms depending on load. I measured output ripple for my application at about 90mVpp 400kHz ripple spikes that can be attenuated by an LC filter (preferably an smt L part of a few hundred uH). The module may show regulation instability for certain levels of input and output voltage – one example showed instability for input between 8 to 14Vdc with no output load and output set for >28Vdc – the instability was an output saw-tooth ripple voltage of circa 100-150mVpp at circa 90-330Hz, with 400kHz ripple spikes on rising edge of saw-tooth, but that is an unlikely situation.

Thanks for info! UVLO is Undervoltage-lockout, what I have just learned. For those who are interested, take a look at this.

For low loads like phantom powering or just 1-2 12AX7 heaters, then I'd recommend a MT3608 based small cheap module.

Okay. I'll have a look at that.

When powering heaters then I'd not worry about initial hiccup or other quirky operation as the heaters come up to temp, that is to be expected and of no tangible stress.

Roger.

For higher B+ dc supplies then I'd recommend the 'small 'Royer' style module with its full bridge output (or change to doubler) and with added input and output e-cap filtering, or a more powerful inverter module with multiple tap secondary that can be rectified and filtered for a range of B+ levels. Both supplies need a fast diode like UF4007, and minimal loop layout of parts.

That's interesting, I'm not sure exactly what you mean by that. I don't think you're talking about the Royer Trippler PSU (which I know well), right? Can you post a link?

Such cheap ebay modules are an excellent diy choice imho, and on close inspection of pcb layout and operation perform as well as can be expected of whatever switchmode tech is being used. About the only aspect I've noted is judicious addition of smt caps in some locations to minimise switching current loops to smallest practical region on the pcb.

Even if there are better ones (I assume there are), you can do a lot with them from my point of view and these things definitely invite you to experiment. +1

 

[trobbins]

Thread on the 'Royer' module: https://www.aggh.net/discussion/index.php?topic=54659.15

Google and reading is your best friend for 'probing smps', such as:
https://www.mwrf.com/technologies/t...kills-for-switch-mode-power-supply-evaluation
https://www.electronicspecifier.com...-techniques-for-measuring-power-supply-ripple
https://www.ti.com/lit/ta/ssztb25/s...24437&ref_url=https%3A%2F%2Fwww.google.com%2F