LM338 question (solved!)

[Viitalahde]

Today I hooked up a TO-3 LM338 regulator for 1,8A worth of 6,3v filaments. 2x 6SN7 and 2x 7193.

With the load connected, the raw, rectified & smoothed DC is about 13 volts. This voltage I'm taking from series connected 6,3vac/4A & 5vac/2,1A windings (which is not ideal because of the unbalance, but I wanted to try it before installing another transformer). Capacitance used is 22,000uF.

For some odd reason I just can't get the damned LM338 go past 6,0v with the load connected.

I first had 255ohms as R1 and a 1k pot as R2. Worked fine without the load. After connecting the load, output voltage dropped too much. Then I reduced the R1 to 200ohms, which helped me to get to 6,0vdc, but I'd still like to get more out of it in case I want to dial in a full filament voltage. At 6,0vdc the output shows a small amount of ripple, which almost vanishes when reducing the output to about 5,8vdc.

Here comes the question:

Years ago I remember reading something about R1 & higher currents - the R1 should be made smaller than usual or the regulator won't work properly. I can't find any reference to this, so it might be BS..?

I'm returning to this on monday but I find it strange that with some 7 volts of extra voltage to regulate from, I can't get past 6,0vdc..

 

[PRR]

> R1 having also an effect in current limiting the output

No.

> the output shows a small amount of ripple

How bad is the input ripple? Remember you need 2V drop in the chip or you get NO regulation; you want 3V-4V minimum drop (at lowest wall voltage and highest output setting) for sure regulation.

> with some 7 volts of extra voltage to regulate from

1.8 Amps (really more- heaters are cold) at 7V is 13 Watts.

A naked TO-3 is good for 1W-2W. You don't say what heatsink you have. So I can only ask: how HOT is it getting?

> 1,8A worth of 6,3v filaments. 2x 6SN7 and 2x 7193. ... smoothed DC is about 13 volts. ... Capacitance used is 22,000uF.

So why regulate? Wire each pair of 6.3V in series for 12V 0.9A. A couple thousand uFd ought to be ample for 1 Amp load. So trade the 22,000u for a 5,000u, a 1 ohm 2W, and another 5,000u. The output will be 12-some volts of very clean DC. And one less bit of silly Silicon to confuse or disappoint you.

> not ideal because of the unbalance

What unbalance? If there is no other load, and neither winding is over-loaded, it's fine. And working series at 0.9A, the AC current is just ~~1.5A, well under the 2.1A rating and loafing the 4A winding.

 

[Viitalahde]

[quote author="PRR"]How bad is the input ripple?[/quote]

First thing I checked. It was certainly under 0,5 volts, can't remember more to be specific.

[quote author="PRR"]A naked TO-3 is good for 1W-2W. You don't say what heatsink you have. So I can only ask: how HOT is it getting?[/quote]

That one I forgot. The TO3 is in the back panel of a 4HE chassis, pretty heavy steel. I used thermal grease, insulation & whole shebang. The panel gets hot around the regulator, I can hold my finger on it for maybe 4-5 seconds.

I did however keep the regulator on for a pretty long time and didn't go for thermal shutdown.

[quote author="PRR"]So why regulate?[/quote]

Two things. I wanted to make sure the heaters stay at constant voltage and don't get any crap injected into them. The current building where our studio is at has a lot of businesses working with heavy equipment. I'm also considering moving to the countryside, and who knows how the electricity goes there. 2nd, I also wanted to experiment with the sonic effects of different heater voltages. I actually tested & tweaked the amplifier with AC heaters..

[quote author="PRR"]What unbalance? If there is no other load, and neither winding is over-loaded, it's fine.[/quote]

There's no other load. Well, maybe I'll hook up a 12 volt lamp to the raw DC. :wink:

 

[Viitalahde]

Oh, in the very first setup I had 2200uF into 2,2r/25w resistor into 2200uF. I calculated it should do a 3-4v drop and ease up the heat on the LM338. This gave pretty terrible ripple and a hefty voltage drop.

Now that the input capacitor is larger, maybe I should experiment more with a voltage dropping resistor? PRR's post made me think that the heat might just be too much. As the regulator was on for maybe 15 minutes, I did notice the output voltage had dropped slightly. maybe 0,15v.

 

[mad.ax]

I had a look at the ST LM338 datasheet that I found here:
http://www.datasheetcatalog.com/datasheets_pdf/L/M/3/3/LM338.shtml

If I understand this right, Vout = 1.25 X ( 1 + (R2/R1)) + (Iadj X R2)

Iadj is about 50µA so even X1000, we can almost forget it...

You ended up using 200 Ohms for R1 and 1K for R2 which gives:
1.25 X (1+(1000/200))= 1.25 X (1+5) = 1.25X6 = 7.5V
Not so far from what you get...

The typical application note from the same datasheet gives 120 Ohms for R1 and 5K adjustable for R2, in order to get a variable 1.2>25V regulator.

You may try it...

Axel

 

[sahib]

[quote author="Viitalahde"]
Years ago I remember reading something about R1 & higher currents - the R1 should be made smaller than usual or the regulator won't work properly. I can't find any reference to this, so it might be BS..?

I'm returning to this on monday but I find it strange that with some 7 volts of extra voltage to regulate from, I can't get past 6,0vdc..[/quote]

No it is not BS. Strangely enough I am about to test the dual PSU that I put together with LT1033 and LT1085. However it is mostly to stop the output rising under no load. On LT 1085 R1 is between 100 and 120 R. On LT1033 it is between 100 and 120 R.

On LM338 R1 is generally 120 ohms. If you look at the example schematics. 240R is specified for LM138. Try it with that.

Also the drop out voltage may well require over 7V, try increasing input voltage.

 

[bcarso]

I'm surprised that the voltage setting R's would be affected, but who knows.

What I would like to see is a succession of good old constant resistance test loads and what the output voltage does under those conditions. The turnon surge of cold filaments is brutal and might have something to do with the problem.

 

[PRR]

> didn't go for thermal shutdown.

I forget how that chip works, and how it may interact with hot-wire load. But it might not cut-out, it might just shave-voltage and keep running.

> under 0,5 volts... heavy steel. ...grease, insulation & whole shebang. The panel gets hot around the regulator, I can hold my finger on it for maybe 4-5 seconds.

Hmmm. That's not too hot.

I dunno.

 

[Martin B. Kantola]

Strange... Some obvious questions: The voltage before the regulator always stays as 13V with the full load? Have you tried a larger R2?

I would try a ramping turn-on, easier on the filaments and deals with the surge through the regulator too. But again, once the heaters are hot it should give you the right voltage...

Martin

 

[Viitalahde]

[quote author="mad.ax"]You ended up using 200 Ohms for R1 and 1K for R2[/quote]

R2 is variable 0-1k. The strange thing is that it works fine without the load. I've never seen a load affect the output voltage that much, unless the input voltage is way too small.

[quote author="mcoz"][quote author="Viitalahde"]
Years ago I remember reading something about R1 & higher currents - the R1 should be made smaller than usual or the regulator won't work properly. [/quote]No it is not BS[/quote]

Wow, you're the first one I hear this from. I quess I just have to try it and see. The ripple voltage that qoes away after reducing the output to 5,8vdc gets me curious too.

[quote author="mcoz"]Also the drop out voltage may well require over 7V, try increasing input voltage.[/quote]

Hehe, that's going to be a problem. :razz:

[quote author="Martin B. Kantola"]The voltage before the regulator always stays as 13V with the full load? Have you tried a larger R2?[/quote]

Yep, the ~13vdc doesn't change at all and the scope says it's pretty nice & smooth.

[quote author="Martin B. Kantola"]I would try a ramping turn-on, easier on the filaments and deals with the surge through the regulator too. But again, once the heaters are hot it should give you the right voltage... [/quote]

This is the first time I'm trying DC filaments, and the first thing I noticed was that visually checking, the start-up seems softer by checking out the glow. With the meter connected, I can clearly see how the regulated voltage climbs slowly up to 6,0vdc. There must be a good deal of current going in.

 

[mad.ax]

Viitalahde, the schemo you posted comes from: http://diyaudioprojects.com/Technical/Voltage-Regulator/
right?

somes quotes from that source:
"For the LM338 and LM350, 120 ohms (for R1) is typically used. However, other values such as 150 or 220 ohms can also be used. Refer to the adjustable regulator datasheets below for more information."

Also, that same page provide an online calculator. When given the values you used, R1=200 Ohms and R2=1K, it returns exactly the same result that I posted before: 7.5V

You need R1=120 Ohms and R2=1K2 in order to get a 13.75V output, so use a 2K or 5K adjustable and you're done!

Martin's sugestion to try a ramping turn-on is more than valid in your app, if you scroll down the LM338 data sheet, you'll find an application note: "Slow Turn-On 15V Regulator"

Axel

 

[Viitalahde]

[quote author="mad.ax"]Viitalahde, the schemo you posted comes from: http://diyaudioprojects.com/Technical/Voltage-Regulator/
right?[/quote]

Nope, I just looked at the datasheet.

[quote author="mad.ax"]You need R1=120 Ohms and R2=1K2 in order to get a 13.75V output, so use a 2K or 5K adjustable and you're done![/quote]

Hehe, but I want 6,3v output. :wink: That's why I feel this is odd - without load it goes there, with load it doesn't even if there's plenty of input voltage.

I'll reduce R1 on monday and see how it goes..

 

[mad.ax]

I assumed it came from there because in the schemo you posted, it says: "http://diyaudioprojects.com"

I wouldn't make conclusions of unloaded measurements, because the LM338 regulator want to see a minimum load...

I don't know why this regulator is so much "load dependant" but I keep thinking there must be a serious reason why the datasheets advise a 240 Ohms value for R1 with LM138 and 120 Ohms with LM338...
Another hint is given under the Load Regulation section in the datasheet, looks like if you add to the value of R1, then your load regulation falls down...

Axel

 

[Viitalahde]

Yep, R1 must be the heart of this problem.

 

[Martin B. Kantola]

[quote author="mad.ax"]You need R1=120 Ohms and R2=1K2 in order to get a 13.75V output, so use a 2K or 5K adjustable and you're done!l[/quote]

Exactly, you can probably even leave R1 at 220 Ohms but need to increase R2 to 2K. Or add some resistance in series to the adjustable.

Martin

 

[Martin B. Kantola]

[quote author="Viitalahde"]Hehe, but I want 6,3v output. :wink: That's why I feel this is odd - without load it goes there, with load it doesn't even if there's plenty of input voltage.[/quote]

With your original values (255 Ohms and 1k) you can barely get 6V according to the equation... So, works exactly as expected...

Martin

 

[Viitalahde]

[quote author="Martin B. Kantola"]With your original values (255 Ohms and 1k) you can barely get 6V according to the equation... So, works exactly as expected...[/quote]

..Which I soon changed to 200ohm + 1k, as told previously.

[quote author="Viitalahde"]I first had 255ohms as R1 and a 1k pot as R2. Worked fine without the load. After connecting the load, output voltage dropped too much. Then I reduced the R1 to 200ohms, which helped me to get to 6,0vdc[/quote]

The reason why 255+1k worked without a load must be that the pot measures near 1k18.

 

[Martin B. Kantola]

[quote author="Viitalahde"]..Which I soon changed to 200ohm + 1k, as told previously.[/quote]

Yes, my point was that it was working almost as expected, but yes, 200 Ohms should work in theory at least, which makes you problem strange. Maybe it's a tolerance issue? 2K should be one the safe side.

[quote author="Viitalahde"]The reason why 255+1k worked without a load must be that the pot measures near 1k18.[/quote]

Quite possibly, but the LM338 (as all regulators of these type) always needs some load to function properly anyway. Found out the hard way when I was trying to use the LM317LZ with a tiny load.

Martin
PS. Terveisiä sateiselta Pohjanmaalta... Tämä kesä ei ikinä lopu...

 

[sahib]

[quote author="Viitalahde"]

[quote author="mcoz"][quote author="Viitalahde"]
Years ago I remember reading something about R1 & higher currents - the R1 should be made smaller than usual or the regulator won't work properly. [/quote]No it is not BS[/quote]

Wow, you're the first one I hear this from. I quess I just have to try it and see. The ripple voltage that qoes away after reducing the output to 5,8vdc gets me curious too.
.[/quote]


LT data sheet mentions that a bit but it generally refers to the required minimum no load current.
http://www.linear.com/pc/downloadDocument.do?navId=H0,C3,P1239,D3905

Although LM339 data sheet gives 5V drop out voltage you never know, you may require some more under load. Therefore check the input voltage under load to see if there is a drop. Or at least what does it drops to. Also if the temperature around the circuit is high you may consider using precision resistors to tighten the voltage divider network.

When I first used L4970 I had the same problem. I was trying to fix the output at about 5.5-5.6V but soon learned that the damn thing required 11V turn on voltage.

Cemal

 

[mwkeene]

[quote author="PRR"]
And one less bit of silly Silicon to confuse or disappoint you.
[/quote]

Haha!