EEVblog Electronics Community Forum

Electronics => Beginners => Topic started by: newbie666 on September 27, 2017, 08:08:44 pm

Title: Best topology for high gain boost converter
Post by: newbie666 on September 27, 2017, 08:08:44 pm
Hello all,

I'm in the process of making a tube phono preamplifier (no audiophool - just like the challenge and glow of those tubes). Most phono preamps use external PSU to reduce mains induced hum so I thought about using a standard 9V wall wart and stepping up it's voltage to 210V required for B+ and stepping it down for heaters.

Heaters are easy, the challenge is creating a nice, low ripple HV supply. So far I found:
* classic boost converter
* boost converter with coupled inductor
* flyback converter

Out of those 3 I'm only familiar with design of classic boost converters. I followed the datasheet of LTC1872 and I came up with this: (attachment). Now boosting 9V to 210V@10mA means 95%+ duty cycle. It simulates ok but I'm not sure if I'll be able to pull it off in real life.

Can you please tell me if coupled inductor boost converter will be better in this case? I know DC will be less than 95% but I'm not sure how to choose such inductor and if I'll be able to get off-the-shelf one. Creating a custom inductor is out of question as I don't have knowledge / experience / tools to do so.

Same thing for flyback converter. Do standard transformers for flyback converters even exist? Can you point me to something that could be even remotely applicable to my use case?

Thanks.
Michal
 
Title: Re: Best topology for high gain boost converter
Post by: T3sl4co1l on September 27, 2017, 08:34:40 pm
Flyback is way better here.  Use an off-the-shelf flyback converter transformer, for a 120/240VAC to 5-12VDC application.  Absolutely nothing wrong with using a transformer in "reverse"!

There's this,

(https://www.seventransistorlabs.com/Images/Discrete_Tube_Supply.png)

which has much the same functionality as a UC3843, or perhaps the LT1872 as well, I'd have to check.

Winding details are shown (well, not a layer stackup, that would be nice I guess), in case you want to make your own.

(One note: only the HV output is regulated, but it really needs to be both, jointly.  Another feedback resistor, like the 680k from +100, but from +6.3 to the TL431, does this.  Adjust resistor values to get the right output voltage and weighting between outputs.)

Tim
Title: Re: Best topology for high gain boost converter
Post by: newbie666 on September 27, 2017, 08:48:15 pm
Hi,

Thanks a lot for your reply. Would you care to elaborate on this:

Quote
Flyback is way better here.

Why exactly is it better than classic boost configuration? Efficiency? Radiation?
Title: Re: Best topology for high gain boost converter
Post by: Howardlong on September 27, 2017, 10:25:38 pm
It’s down to efficiency. With normal boost converters, the high duty cycle required reduces efficiency. A typical problem with flyback converters is the need for (often custom) coupled magnetics, but as stated you can cheat by repurposing a mains transformer. Magnetics are often expensive commodities in SMPS!
Title: Re: Best topology for high gain boost converter
Post by: langwadt on September 27, 2017, 11:20:02 pm
Quote from: newbie666 on September 27, 2017, 08:08:44 pm
Hello all,

I'm in the process of making a tube phono preamplifier (no audiophool - just like the challenge and glow of those tubes). Most phono preamps use external PSU to reduce mains induced hum so I thought about using a standard 9V wall wart and stepping up it's voltage to 210V required for B+ and stepping it down for heaters.

Heaters are easy, the challenge is creating a nice, low ripple HV supply. So far I found:
* classic boost converter
* boost converter with coupled inductor
* flyback converter

Out of those 3 I'm only familiar with design of classic boost converters. I followed the datasheet of LTC1872 and I came up with this: (attachment). Now boosting 9V to 210V@10mA means 95%+ duty cycle. It simulates ok but I'm not sure if I'll be able to pull it off in real life.

Can you please tell me if coupled inductor boost converter will be better in this case? I know DC will be less than 95% but I'm not sure how to choose such inductor and if I'll be able to get off-the-shelf one. Creating a custom inductor is out of question as I don't have knowledge / experience / tools to do so.

Same thing for flyback converter. Do standard transformers for flyback converters even exist? Can you point me to something that could be even remotely applicable to my use case?

Thanks.
Michal

sure you need that high a B+ voltage?


Title: Re: Best topology for high gain boost converter
Post by: Audioguru on September 28, 2017, 12:09:05 am
Phono? Oh yeah, I have one that I used 55 years ago.
Vacuum tubes? Oh yeah, my audio system used them 55 years ago.
Title: Re: Best topology for high gain boost converter
Post by: basinstreetdesign on September 28, 2017, 01:44:38 am
A SMPS booster for a high-gain phone ckt.  Sure, that can work.  I, myself, have used a classic boost ckt for a low-power guitar amp, 12V -> 125V @ about 85 mA with some success.  Guitar amps have a high gain, too.

The thing I found about this approach is that while the boost ckt may work in the ultrasonic frequency range, the duty cycle or frequency of it, what ever is controlled for regulation, will be modulated by a random (maybe Gaussian) noise on top of the controlling signal.  The frequency content of this noise extends right down to 0 Hz and up past the audible range.  This noise will be applied to the high-gain amp through the B+ even if filtered between the PSU and the input stages of the amp.  Since a phono amp must have in excess of 40 dB of gain at low frequencies (better yet, 60 dB) and at the same time have many more dB of noise-free response below that  to the worst-offender noise signal, then getting rid of the PSU noise is extremely difficult.  You're one ace-in-the-hole is that the frequency characteristic is RIAA, so reduces high frequency hiss greatly.

My guitar amp has a constant hiss which is just acceptable after as much filtering I could put on it.

The more power you want to produce, the harder it will be to filter it.  The noise travels through conducted paths and through radiation.  You may need to give the switcher its own box, be it a steel box or a copper-clad box with only two holes in it, power input and output.

Best of luck!
Title: Re: Best topology for high gain boost converter
Post by: T3sl4co1l on September 28, 2017, 02:19:01 am
Well, careful -- you're talking about two kinds of noise here.

Low frequency noise, arising from jitter in the switching circuit (it doesn't always switch exactly on time), noise in the primary supply (particularly mains ripple, if powered offline or with a transformer, but also cross-talk from other loads, e.g., engine and accessories in a 12V automotive situation), and the noise floor of the error amplifier (controller), all contribute to the output noise, in the passband.  (The passband being, everything that's let through the output low-pass filter: this will cut at some fraction of the switching frequency.)

LF noise has the advantage that, the controller is usually quite quiet to begin with (microvolts would be noisy!), and it uses negative feedback to stabilize the output, meaning low-frequency noise (like mains ripple) is particularly well filtered by it.  High frequency noise might not be so well filtered (the amount of negative feedback necessarily drops off at higher frequencies, allowing more noise through), but it is easier to filter with a passive (LC) filter, and there are usually fewer sources of it.

(That said, speaking of automotive: alternator whine is right in the middle range, where it might be accidentally audible.  It's also around peak auditory response, which doesn't help!  On the upside, cars aren't usually very quiet, so it would be hard to notice, eh?)

Above the output filter's cutoff, switching noise and RFI is filtered.  Or it's supposed to be, anyway.  Filtering and shielding of high frequency content can be tricky, and proper design must be followed!

And there's always the matter:

How much filtering do you need?  How much noise can you tolerate?

We can put numbers to everything, here.  Nothing need be left to guesswork or tradition! :)

The nice thing about a tube amp is, the gain per device, and the number of devices, and the gain at high frequency, are all so low that it's very difficult to receive RFI noise.

SS amps have trouble with this (because physics), and it's often immediately obvious when filtering has been neglected.  Ever had the experience of a set of powered speakers, playing some local radio station, just sitting there plugged in with nothing playing?  Alas, they didn't spend the six cents to install filtering, and guess what, you have to listen to some annoying radio station, ever-so-quietly, now... yep...

Tim
Title: Re: Best topology for high gain boost converter
Post by: newbie666 on September 28, 2017, 01:48:24 pm
Noise discussion is interesting but it's a bit ahead of me for now but from what T3sl4co1l posted there are two issues:

* low frequency noise from mains / surroundings - this is applicable regardless if I use "normal" or smps psu. In my case I'm planning to power the device from a 9V wall-wart or maybe even 5V battery pack, preamp will be built in a steel / aluminium case etc. I should not get too much interference

* for HF noise - i was thinking about dealing with it with a post-dc-dc regulator. Even a simple floating[1][2] lm317 can give me 30db of ripple attenuation at 500kHz, not to mention a newer generation of LDOs like TPS7Axx that reach 60dB @500kHz. That really makes me think that dc-dc converter can be a better option than a traditional linear PSU even for sensitive applications like tube phono preamp.

(I still need to figure out how to use TPS7Axx in a floating configuration as, unlike lm317, it needs connection to ground - any tips?)

For now I'm trying to build a flyback converter in ltspice based on ltc3873. If you have any other chip recommendations feel free to post them here.  Anyway, my biggest issue for now is finding a suitable transformer. Can you guys point me in the right direction? Honestly I don't even know what to search for.

Edit: found a note about using TPS7Axx as a floating reg[3]. I'm not sure if I really understand that design. Let's say we ignore all the current sharing stages (I don't need more current that TPS7A can supply) and also ignore control signal part, then, where I would generate virtual ground from?

[1] https://electronics.stackexchange.com/questions/246204/lm317-maida-style-regulator (https://electronics.stackexchange.com/questions/246204/lm317-maida-style-regulator)
[2] http://www.pmillett.com/HV_reg.html (http://www.pmillett.com/HV_reg.html)
[3] http://www.ti.com/lit/df/tidrpx9/tidrpx9.pdf (http://www.ti.com/lit/df/tidrpx9/tidrpx9.pdf)
Title: Re: Best topology for high gain boost converter
Post by: T3sl4co1l on September 28, 2017, 09:33:07 pm
Quote from: newbie666 on September 28, 2017, 01:48:24 pm
* for HF noise - i was thinking about dealing with it with a post-dc-dc regulator. Even a simple floating[1][2] lm317 can give me 30db of ripple attenuation at 500kHz, not to mention a newer generation of LDOs like TPS7Axx that reach 60dB @500kHz. That really makes me think that dc-dc converter can be a better option than a traditional linear PSU even for sensitive applications like tube phono preamp.

BUT, do you need it?

Remember:

Quote from: T3sl4co1l on September 28, 2017, 02:19:01 am
And there's always the matter:

How much filtering do you need?  How much noise can you tolerate?

Even if you know very little about your signal, amplifier, and load, very basic guesses can be made!

Quote
(I still need to figure out how to use TPS7Axx in a floating configuration as, unlike lm317, it needs connection to ground - any tips?)

It's a lot of bother to go to, without knowing whether you need it or not!

Quote
For now I'm trying to build a flyback converter in ltspice based on ltc3873. If you have any other chip recommendations feel free to post them here.  Anyway, my biggest issue for now is finding a suitable transformer. Can you guys point me in the right direction? Honestly I don't even know what to search for.

Search a distributor:

Quote from: T3sl4co1l on September 27, 2017, 08:34:40 pm
Flyback is way better here.  Use an off-the-shelf flyback converter transformer, for a 120/240VAC to 5-12VDC application.  Absolutely nothing wrong with using a transformer in "reverse"!

https://www.digikey.com/products/en/transformers/switching-converter-smps-transformers/168?k=transformer (https://www.digikey.com/products/en/transformers/switching-converter-smps-transformers/168?k=transformer)
:)

Tim
Title: Re: Best topology for high gain boost converter
Post by: David Hess on September 29, 2017, 02:35:08 am
I would use a push-pull inverter to generate the high voltage and regulate if necessary on the primary side but regulation should not be necessary.  This avoids storing energy in the inductor or transformer core so it may be smaller and the noise will be lower.  The problem then becomes finding a suitable 20:1 transformer.  I have a pile of small 3C8 pot cores that I have been rewinding which were originally used for gate/base drive transformers.  Some small signal audio transformers might be suitable.

A flyback regulator would be the simplest otherwise.  As you found, a step-up converter has an untenable duty cycle.
Title: Re: Best topology for high gain boost converter
Post by: newbie666 on September 29, 2017, 06:52:05 am
Quote
Quote from: newbie666 on Yesterday at 11:48:24 PM
* for HF noise - i was thinking about dealing with it with a post-dc-dc regulator. Even a simple floating[1][2] lm317 can give me 30db of ripple attenuation at 500kHz, not to mention a newer generation of LDOs like TPS7Axx that reach 60dB @500kHz. That really makes me think that dc-dc converter can be a better option than a traditional linear PSU even for sensitive applications like tube phono preamp.

BUT, do you need it?

I didn't say I do, merely pointing out that I don't really see noise as an issue here. Experience may prove me wrong of course. Since I don't know that much about smps' I'm doing breadth-first search in this space that's why I was asking how to implement a floating regulator with one of those super low noise LDOs ;)

I'm moving forward with calculations for flyback converter.
Code: [Select]
Vin	5       [V]
Vout	210   [V]
Iout	0,02  [A]
f       250   [kHz]
 

gives me
Code: [Select]
DC	  0,807692308     [%]
Ipeak(p)  2080	        [mA]
L(p)	  7,766272189	[uH] <- for boundary operation

Which looks sensible. I'll look for a transformer now. Can not thank you enough for your link T3sl4co1l


@David Hess
Using push-pull inverter would imply rectification and filtering on the secondary side, do I understand you right? What's the benefit then over flyback converter?
Title: Re: Best topology for high gain boost converter
Post by: newbie666 on September 29, 2017, 06:45:42 pm
Ok so after struggling for a while I selected Wurth 750841012[1] as my transformer.

Here's a full schematic of the converter:
[attachment]

I have a couple of questions:

1) What do I do with unused aux? Should I just leave it unconnected?
2) There's a lot of ringing during startup - the more voltage I want on the output the more severe ringing is. Should I be concerned?
3) I can see inductor peak current settling down at around 2.2A which is more or less in line with my calculations. However there's a big current spike when transistor turns on. Should I be concerned about it? Will it cause some extra radiation? How do I minimize it? If the spike is at around 7.5A it means that mosfet needs to handle that current?



[1] http://www.mouser.com/ds/2/445/750811137-259565.pdf (http://www.mouser.com/ds/2/445/750811137-259565.pdf)
Title: Re: Best topology for high gain boost converter
Post by: Kleinstein on September 29, 2017, 07:54:46 pm
The current spike during turn on can be a problem, possibly triggering some over-current protection / turn off. If the real system also behaves that way depends on parasitic elements. One might need things like a snubber or a gate resistor for the MOSFET.

The compensation of the regulator does not look  very well optimized: there is still quite some ringing - though this is a different possible problem.

My personal choice for such a converter would be more like a push pull, or resonant converter (e.g Royer type). This is especially true if the voltage ratio is approximately constant. Anyway I would consider a lower voltage for the tubes - at low power/low frequency there is no need for such a high voltage. However the choice of tubes is not that large any more.

Anyway a tube amplifier for phono is kind of odd, phono inputs often want rather low noise amplification and sometimes low impedance, thus not really suitable for tubes.
Title: Re: Best topology for high gain boost converter
Post by: newbie666 on September 29, 2017, 08:23:01 pm
I added 50R gate resistor helped! the spike is not gone but is now limited to peak inductor current. BEAUTY.

I need to admit I have no idea why this happens, does anyone care to explain why there's a current spike and why gate resistor helps? Also some guidelines on how to choose a value for one would be great.
Title: Re: Best topology for high gain boost converter
Post by: T3sl4co1l on September 29, 2017, 09:31:12 pm
Heh, well, you're not going to get away with an MBRS340 in real life...

Two things: gate capacitance and load (diode) capacitance.

When the controller turns on the transistor, it rapidly changes the gate voltage.  I = C * dV/dt, and dV is 5V and dt is small, and C is constant, so I is large (compared to the average gate current).  This isn't usually directly visible in the source resistor's voltage, unless its value is also large-ish (~1 ohm).

When the transistor turns on, it yanks the drain voltage down.  This discharges the drain capacitance as well, but that current flow happens internal to the transistor, so doesn't cause any current flow in the source resistor (however, it does dissipate power, adding to switching losses).  Everything external to the transistor, however, does draw a current.  The diode has capacitance, and it is multiplied by the transformer ratio.  That's probably what you're seeing.

A large gate resistor isn't necessarily the best idea.  Slow turn-on can be okay for a flyback, but turn-off should be fast.  You can add a diode in parallel with the resistor (pointing towards the controller) to tweak this.

Real transformers have equivalent leakage inductance and parasitic capacitance.  Did you manage to find a SPICE model for that part?  Or did you make one up, with k=1 or so?

The datasheet specifies leakage 17uH typ., primary referred (which is your secondary).  Or 0.155uH referred to your primary.  So what will really happen is, when the transistor turns on, the leakage acts in series to the diode capacitance, and you get a series resonant tank, which (assuming B340 typical characteristics, because again, a real diode won't handle 100V across it*) gets you on the order of 200pF (secondary referred), or 0.022uF (primary referred).

*At a quick check, LTSpice has "Vpk=40" in the diode model, but doesn't do anything with it ... I guess the ever-trusting Mike Engelhardt expects his users to check this themselves.  The model happily drops gigavolts without complaint...

Zo = sqrt(0.155u / 0.022u) = 2.66 ohms (resonant impedance)
Fo = 1 / (2*pi*sqrt(0.155u * 0.022u)) = 2.72MHz (resonant frequency)

Since Rds(on) << Zo, expect a strong resonant spike, appearing as current at the transistor (the spike you see, will actually be ringing), and voltage at the diode.

The best way to dampen this is to not excite it in the first place**, or add a R+C across the diode.

**Rather impractical at 2.7MHz.  However, if you could decrease the diode capacitance -- say by using a high voltage diode like UF4007 -- you'll also raise the resonant frequency, which at least helps, even if it might not be high enough.

The transformer itself has capacitance, too, and it's not clear how much.  Probably, a diode under 20pF isn't going to achieve any more improvement, which is about what a UF4007 has, so you can expect sqrt(10) times higher or 8.5MHz ringing.

The R+C can be calculated from the resonant impedance.  R = Zo.  Secondary referred, so, (10.46)^2 times 2.7 or 295 ohms (330 is a standard value).  C should be a few times the offending capacitance, which is about 20pF (for UF4007), or 60pF.  100pF will be fine, too.

And yes, you can leave the aux winding completely open, no worries.

Tim
Title: Re: Best topology for high gain boost converter
Post by: newbie666 on September 30, 2017, 07:00:38 am
Thanks T3sl4co1l for excellent explanation. Diode is my mistake - i haven't checked breakdown voltage. I'll look for one with 300V+ breakdown and the smallest capacitance then. I've quickly replaced it with another model in spice and I see the spike decreased 2x.

Can I bother you one more time for explanation of this part:

Quote
Since Rds(on) << Zo, expect a strong resonant spike

Why does that happen? No need for lot's of typing if you don't have time, can you tell me what's that called and I check it on wikipedia / google :)


EDIT: yes, wurth provides spice models that are supposed to be quite accurate for all their transformers.
Title: Re: Best topology for high gain boost converter
Post by: IanMacdonald on September 30, 2017, 08:42:46 am
The most likely noise problem here will be with SMPS ground-side currents. If these are allowed to get into the phono ground they will wreak havoc.
Title: Re: Best topology for high gain boost converter
Post by: David Hess on September 30, 2017, 10:46:21 am
Quote from: newbie666 on September 29, 2017, 06:52:05 am
@David Hess
Using push-pull inverter would imply rectification and filtering on the secondary side, do I understand you right? What's the benefit then over flyback converter?

The output is essentially continuous so filtering is much easier and the ripple current is much lower.
Title: Re: Best topology for high gain boost converter
Post by: T3sl4co1l on October 01, 2017, 12:53:57 am
Quote from: newbie666 on September 30, 2017, 07:00:38 am
EDIT: yes, wurth provides spice models that are supposed to be quite accurate for all their transformers.

Oh, good!

Oh, great... they encrypt it.

So I can't check the veracity of their model, unless I reverse engineer it.  Oh, but there's text in there that says I can't reverse engineer it.  Well, fuck 'em for trying, it's a freaking transformer anyway!

So, let's see...

Hahah, they put k(aux,sec) = 1.0.  That'll give people some unexpected problems with cross-regulation...

It's a linear model, so doesn't include saturation.  (Not surprising, but it's not hard to add.  Still no excuse to encrypt. ::) )

Looks like they model winding DCR, primary inductance, core loss (probably?), pri-sec leakage inductance, and equivalent primary capacitance.  Saturation, sec-aux leakage, skin effect, pri-sec capacitance, and higher modes (winding resonances and such) are not modeled.

Rpri pk = 684.4kHz, 39.45k ohm
Zpri Lmag = 92.76kHz 984.87 ohm
RDC pri 3.203 ohm
Zpri SC pk = 4.191MHz 399.44 ohm
RDC sec 0.02 ohm
RDC aux 0.482 ohm

So that gives:
Lpri parallel = 1689 uH
Rpri parallel = 39544 ohm
Cpri parallel = 32.02pF
Lpri series = 45uH
Rpri series = 3.203 ohm
Rsec series = 0.02 ohm
Raux series = 0.482 ohm

And the ratios as given in the datasheet.

These values are for the low frequency equivalent transformer model, like so,
https://electrical4u.com/electrical/wp-content/uploads/2013/03/equivalent-circuit-of-primary-side-of-transformer.gif
where Z1 is made of Lpri and Rpri series, and Xo, Ro are made of Lpri, Rpri and Cpri parallel.  The secondary side looks the same, but has only R1 (the respective DC resistance).  The transformer is ideal (k=1), because Lpri series takes up the primary referred leakage (of course, such a model misses sec-aux leakage).

If you draw this up on a sheet of paper, you can use transformations to see the series resonant circuit formed with Lpri series, Cpri parallel, and whatever load capacitance there is (i.e., the diode's CJO).  That's where that comes from. :)

Tim