The MAX1771 SMPS Development Thread

[cfcubed]

>  but completely forgot I recommended these

Heh, I didn't see that in this thread.   Anyway they could have been a nice, cheap solution so it was worth a try.  There are not that many off-the-shelf solutions that provide at least 200V, besides googling here's a nice resource:
http://www.tayloredge.com/storefront/SmartNixie/PSU/comparison.html

> is that coming from the SMPS unit itself or is it hash in the DC power or both?

I think its coming from both (would have to scope around to confirm)...  I cut, bent & installed a little sheet metal box over the switcher (in lower-left of pic above) and seemed to reduce the audio level of the freq in the amp's output to "barely there" w/no signal & not perceivable with any signal (YMMV).  So think some of this is RFI and might be reduced though use of a shielded inductor... BTW I'll be ordering the shielded one for Nick's/ogi's solution if I have any issues with that piece.
But think some of this is also coming direct through B+.  Note I'm using dual CTH cap multiplier filters for B+ (the raised protoboard strip along left in pic, schem on cavalliaudio).

I asked TE what chip they are using on these 1363/4 to research this a bit (as the #s on mine were rubbed off/not legible), but haven't heard back.  Again the only chance to use these for audio seems to load them up so they crank their freq up (e.g. 100k across output, tube draw, lower the input V) & shield them, but it seems to me to plateau at the very top edge of audible range.  

Will be curious what you find if you play w/them.  Please let me know if you can read the #s on the switcher.  At least they are great for testing / proof of concept, enough so I'm pursuing the MAX1771 solution, and we can always build a Nixie clock w/'em someday:)

[defaced]

I grabbed my camera and took two pictures of the board with different lighting angles.  It looks like the chip in the upper right is made by STMicroelectronics, but searching for their website and Google gave me nothing.  The other chip 8 pin chip, to the left of the big one in the MS8 package, is made by Linear Technologies, part number LT1619, here's the data sheet: http://cds.linear.com/docs/Datasheet/1619fa.pdf.  I'm thinking that the STMicro chip might be an MCU...

The images are about 1M each in size. 

http://img690.imageshack.us/img690/2945/smps2.jpg

http://img841.imageshack.us/img841/2382/smps1.jpg

[cfcubed]

Nice pictures, thanks.  
Now I have to say that the high frequency "whine" / tone /oscillation I hear from this SMPS in my application is a mystery.  
AFAIK "modern" switchers operate @ freqs well above 20kHz/audible range.  And yes, it is emanating from the SMPS & not somewhere else in the circuit.

That LT doc/chip you found appears to operate @ 200kHz+.  So maybe what I'm hearing is not switching frequency but something else (e.g. ringing / oscillation) as IIRC putting a finger/pressing on the switcher's inductor reduces the sound.  Other posters in this thread are much more knowledgeable about this than I, but maybe I'll fool w/this switcher a bit more & see what I find.  At least until the MAX1771 arrives:)

Perhaps this little $13 #1363 HV supply design is just not that robust/stable, at least not with what I'm asking it to do.
Will be curious what you find if/when you give that 1363 a go in a circuit.

Edit:  Well just found a lot of good (but discouraging) info WRT SMPSs & audio in the "Do SMPS have a home at high end audio?" & other threads over @ diyaudio.  But everyone chooses their compromises so I may keep at it:)

[defaced]

Will be curious what you find if/when you give that 1363 a go in a circuit.

Me too.  I was originally hoping to run both of them from a single supply and stack the outputs for 400v circuits (most high gain tube circuits are 4 stages + CF and start upwards to 400v).  Then I discovered that the grounds for the input voltage are shared with the output, which means I need either two power transformers or a transformer with two secondaries.  Not the end of the world, but it's not what I have on hand.  But, yea, anyway, I'll post whatever I find when I find it.  This will be a nice project once I get my scope up and running. 

[zambo]

so i am not really sure but would it help to go with more of a traditional resistor/cap/res/cap type of power supply fed by a 1363?

[cfcubed]

so i am not really sure but would it help to go with more of a traditional resistor/cap/res/cap type of power supply fed by a 1363?

Having communicated w/someone knowing a lot more than I about this, yes something like a RC or LC PI filter @100kHz 40kHz or so (?) would suppress most modern switcher's freq in output/B+.
But don't think that's the problem we are experiencing (e.g. the coil squeal/oscillation) w/the 136Xs.  Again my total guess is its a not the most robust/stable implementation, e.g. maybe its got a "sweet spot" where it doesn't squeal/oscillate but who knows where that is.  
And that electromechanical(?) emanation did seems to pollute my little test amp.

Maybe someone else will get somewhere w/inexpensive switchers like 136Xs but if I pursue this further it'll be w/apparently more robust implementations like the MAX1771 I referenced.  And maybe shielded coils & an RC or LC output filter.

[zambo]

i am curiouse about the 1771. I got two amps to work without squeal using the 1363 but when i was hooking up preamps it was horrible. The 1 watt and 2 watt amps work great though. weird. I think the 1771 has enough milliamps to make a sweet  amp.

[defaced]

The one thing that would make me very happy is a MAX1771 (or other, I'm not picky) circuit good to 400v and about 10-20ma.  That is a project I don't know enough about to tackle, but want to design very badly.  I read some of that DIY Audio thread and there's some discussion in there I've got to do more research on about the different topologies mentioned. 

[Cliff Schecht]

That's not that hard really. It's making the design compact and low noise (err no noise really) that's the kicker. Getting a good gain and phase margin is the trick, but it's not a very easy to do without lots of feedback knowledge.

I plan on eventually revisiting my HV SMPS designs but as is usually the case I am swamped with school and other projects.

I'd like to eliminate the guesswork by having different schematics that people can build on the same board for their voltage needs, essentially adjusting a few parts for max stability and lowest noise at whatever voltage level one chooses. But as anybody who has done this type of design work would know, it's not very easy to get past about a 10-20x voltage boost off of the input supply. It really depends on the design though as some converters are limited to 50% duty cycle while others can hit close to 100% duty cycle.

[Cliff Schecht]

So I think I'm bored enough to dig up my old design files and finish the design I started forever ago. I got up to a working prototype last time but was never fully satisfied with my work. I think with another years worth of practice and a lot more knowledge when it comes to tubes, I should be able to nail down a solid design that I'd be comfortable sharing with the community.

[cfcubed]

zambo - I got two amps to work without squeal using the 1363  Interesting.  Did you (or would you) share info about those in some other thread?
Guess I've done a lot of 136X nixie PS talk here given this thread's title:)

Cliff - Sounds good & that'd be nice & generous of you:)

BTW I'll see & post what I find after playing around w/the MAX1771 bit coming my way.  Will probably try 220v or so in my tests to keep w/250v rated parts.

[Cliff Schecht]

I finished a 200V, 50 mA design last night (for an 8-16V DC input). I'm going to keep stepping up the voltage by 50V until I get to maybe 400V, but the input voltage requirements will increase as well with this change. Hopefully all of these different levels will fit nicely on the same board.

I probably won't be doing a through-hole version of this design though as the oscillator is running in the 1 MHz range and TH parts don't play well with switchers at this high of a frequency. The parts count is relatively low (again I'm doing current-mode discontinuous designs, they fit our needs well) and shouldn't be more than $10-$20, even cheaper if you are smart and order samples of the IC from TI. I'll resurrect my old thread on this when I have everything prepared.

[Ice-9]

A little OT but !

Cliff , thanks for the great work you have put in on this , but I have to ask.How old are you ?  You keep saying "when I get in from school I will look at this etc". Where I come from we leave school at 16 years old. Your knowledge seems to exceed a young pre 16 year old.

[Cliff Schecht]

I'm 24 and a fresh PhD student, so I'm always somehow busy between that, gigging and building stuff (not to mention the girlfriend..). I think I joined these forums when I was about 16 though..

[taylotjpt]

The 1363/1364 switchers use the Linear Technology LT1619 running at 300kHz into a open bobbin flyback transformer.  Note that the numbers are never removed, however part of the manufacturing includes covering completed units with a polyester conformal coat to keep the windings in place and limit corona leakage from the high voltage nodes.

Generally, audible noise is a function of three things:
1.  If the input is not cleanly bypassed the input voltage is pulled at a frequency dependent on the inductance of the wiring and can actually destroy the device due to over dissipation of the switching FET.  The 330uF capacitor supplied should be placed as close to the power pins as possible:  Source-to-Capacitor-to-Switcher
2.  If the load is very inductive, the 1uF output capacitor and the load can form a parallel resonant tank circuit.  If the resonant frequency is any close upper or lower harmonic of the 300kHz switching frequency the device will pump the load and possibly destroy the rectifier or switching transistor because of capacitor over voltage or diode reverse leakage.  The solution to this is to insert a series resistance to lower the Q or increase the output capacitance to push the resonant frequency below the lowest resonant subharmonic of the switching frequency.
3.  Secondly, this switcher uses a PFM control scheme that skips cycles when the load does not pull the output down below the recycle threshold resulting in a "frequency" of operation less than the 300kHz internal setting.  The core can emit sound via magnetostriction if the PFM beat frequency is in the audio range.  To disable the power saving burst mode and force continuous switching, a resistor (About 130 ohms) can be inserted between the LT1619 pin 5 and the 10mOhm current sense resistor.

See current data sheet for precautions against conductive shielding and "Fingers" on the transformer core or FET.

jt

[defaced]

Thank you for the confirmation and additional information about the operation of your SMPS, Taylor.  I'm really looking forward to using your devices for my project. 

[cfcubed]

The 1363/1364 switchers use the Linear Technology LT1619 <snip>
Generally, audible noise is a function of three things:
<snip of loads of great info>
jt

Wow.  Thanks for all the great info & suggestions from someone who knows:)  Will consider these esp. as I'm going back to the 1363/1364 switcher for my little project.  If/when I do that & get anywhere I may start a new thread w/different title.

The pre-built MAX1771 addressed one issue for me (no high freq artifacts, no audible (electromechanical?) emissions) but introduced a more problematic one (a low freq buzz).  Tried various filtering networks both @ input & output w/that MAX1771 implementation and could not reduce the low freq buzz substantially in my application.  
Let me be clear though, YMMV and I'm not an EE.  My total guesses WRT my MAX1771 issues would be a ground problem stemming from the way it was wired in on the bench, e.g. perhaps that implementation/architecture is more sensitive to this.  IMO the precise behavior of these HVSMPSs in specific scenarios, esp. in cases for which they were not explicitly designed (e.g. audio), can be difficult to predict (esp for a layman:).  

[cfcubed]

The 1363/1364 switchers use the Linear Technology LT1619 running at 300kHz into a open bobbin flyback transformer. <snip>
...  To disable the power saving burst mode and force continuous switching, a resistor (About 130 ohms) can be inserted between the LT1619 pin 5 and the 10mOhm current sense resistor.

In the event one wants to try the above, I suppose lifting pin #5 and bridging it back to its pad using an approx 130R resistor be a way to do this:



P.S. sorry again for this slightly OT post, but while we're on the subject....

[defaced]

Any time someone engineers a solution for me, it's not off topic, it's appreciated.  Thanks!  This is something I will very likely do when I play with these. 

[cfcubed]

Any time someone engineers a solution for me, it's not off topic, it's appreciated.  Thanks!  This is something I will very likely do when I play with these.  

But lifting pin #5 and bridging it back to its pad using an approx 130R resistor is easy to say, not to do - scratch one #1363  
Found I could not lift a pin on these little buggers, best I can do is the entire 4-pin chip side tilting the chip up (solder "wipe" method)... Then bridge 3 back down & one use 130R.  In theory.  If PCB pads aren't lifted, etc.  Not sure that's possible either w/my pedestrian gear.

taylotjpt - If you visit here again could you please point out/mark the Rsense resistor in defaced's or my pic (perhaps easier to lift) or(<<-- nope, that won't work either as the 130R must be between pin #5 & the FET/Rsense juncture) give other thoughts on this?  Perhaps its just not possible w/"normal" gear on these tight little SMD puppies.