Regulated Voltage Supplies

[bob p]

So - we ready to design this up?

Funny you should ask... I left out part of the story -- after not getting the tranny delivered after two weeks, I gave up and asked for a VISA chargeback, and I bought another transformer on eBay. Murpy's Law prevailing, the original transformer arrived the day after I paid for a second one. So I'm gonna have two transformers on hand and you know what that means -- I'm gonna have to build two big-ass amps.

Let's start off with a pair of 100W Class A1 monoblocks, each running 3 pairs of 6550 in push-pull. If the plate voltage comes out to about 400 VDC, each pair should produce about 34 watts or thereabouts, for total of 100W per channel. I think that the 2 KVA transformer should be able to handle the job just fine, and I've thought about building parallel supplies for each monoblock from the rectifiers on out. See Post # 70. I know it may sound stupid, but everyone's got to try building an insane tube amp at least once in their life, right?

I think that the 100W Class A monoblocks will be easier from a PSU standpoint, and would probably serve better for a HiFi application than a fully regulated Class AB1 amp. Then, for a later project, maybe I'd go off the deep end and build an evil monster Class AB1 bass amp with that other power transformer. <sinister grin>.

[bob p]

back to the subject of inrush limiting -- the PT was opening the breaker today with no load on the secondary. just energizing the coils required enough inrush current to trip the 20A breaker. i'm thinking that energizing the coils is going to be far more demanding in terms of inrush current than powering on the tubes, so i seem to be missing the concept of how relays would be helpful -- even if the tubes were powered on sequentially in pairs, the initial loading of the transformer would still pop the breaker, so it seems like a thermistor would be needed no matter what. or maybe i'm missing something.

thanks again for your help.

[Steve Conner]

Wow, that's one big transformer. It makes the PT for my bass amp project look pretty inadequate.

The blowing of breakers on turn-on is good! ;-) It's caused by core saturation due to the DC component of the turn-on transient. It means the transformer has a nice low DC resistance. It also means you need a soft-start circuit like R.G. suggested. You need to power up the transformer with a big-ass resistor in series with the primary, and then short this resistor out after a few seconds with a time-delay relay.

[Bruce / Mission Amps]

back to the subject of inrush limiting -- the PT was opening the breaker today with no load on the secondary. just energizing the coils required enough inrush current to trip the 20A breaker. i'm thinking that energizing the coils is going to be far more demanding in terms of inrush current than powering on the tubes, so i seem to be missing the concept of how relays would be helpful -- even if the tubes were powered on sequentially in pairs, the initial loading of the transformer would still pop the breaker, so it seems like a thermistor would be needed no matter what. or maybe i'm missing something.

thanks again for your help.

I think Steve's idea is good.
I do a lot of Drive-In movie theater tube amp repairs and the little Altec 1570B amp is a decent example of how to control some inrush current.
It uses a high wattage 5 ohm resistor in the primary, bypassed by BR1 when energized... oh, just look at the schem and you'll see what I mean.
It also has a high voltage saftey switch on the cage that kills primary voltage when you lift the covers off, so don't get confused by that.

[bob p]

Wow, that's one big transformer. It makes the PT for my bass amp project look pretty inadequate.

Sorry Steve -- I didn't mean to imply that mine was bigger than yours.

As it turns out, the second PT arrived today. Funny thing -- the 1.5 KVA GE Transformer is in a much bigger box than the 2.0 KVA Acme, and it outweighs the Acme by 2 lb. IIRC the GE iron uses aluminum windings, and the Acme iron uses copper. Maybe that will explain the weight and size difference.

The blowing of breakers on turn-on is good! ;-) It's caused by core saturation due to the DC component of the turn-on transient. It means the transformer has a nice low DC resistance. It also means you need a soft-start circuit like R.G. suggested. You need to power up the transformer with a big-ass resistor in series with the primary, and then short this resistor out after a few seconds with a time-delay relay.

I powered-up the GE under no load conditions. It didn't pop the breaker, and the 480 VAC secondary voltage only measured 462 VAC with a 122 VAC input. It was so quiet that I couldn't hear anything over the hum from my fluorescent lamps.

Thanks for elaborating on R.G.'s point about inrush limiting.

[bob p]

II do a lot of Drive-In movie theater tube amp repairs and the little Altec 1570B amp is a decent example of how to control some inruch current.
It uses a high wattage 5 ohm resistor in the primary, bypassed by BR1 when energized... oh, just look at the schem and you'll see what I mean.

Bruce, thanks for that scheamtic! I'm kinda surprised that the Altec uses a 5R 50W resistor for that application, but then I have to admit that I didn't crunch any numbers. I've probably got something lying around here that would fit the bill.

It sounds like the time-delay relay would be a great way to automate the turn-on process so that it's foolproof. I guess that another alternative would be to do this sort of thing manually, with separate panel switches that interrupt the primary and the secondary connections. Although a series of power-on switches and a defined power-on sequence definitely increases the Frankenstein Factor, it does render the amp a bit less foolproof to operate and subjects it to the risk of operator error.

[bob p]

Even though it's only rated at 3/4 of the KVA, the GE is surprisingly bigger.

[R.G.]

bob, I'd like you to measure the DC resistances of the primary and secondaries of the trannies as closely as you reasonably can. They may be too low to easily measure.

I did some simulation work, and it looks like an 8-10H inductor is better than a 1H for doing the voltage lowering directly with a variable load. You might be able to get a class A without regulating.

[bob p]

bob, I'd like you to measure the DC resistances of the primary and secondaries of the trannies as closely as you reasonably can. They may be too low to easily measure.

I don't really have anything that can accurately measure ohms down into the single digits. I don't know how helpful the measurements would be, but here they are. (bear in mind that my "primary" and "secondary" taps are reversed, since I'm using the PT backwards):

ACME 2KVA:  
120 V "primary" = 0.5R
480 V "secondary" = 2.0R

GE 1.5 KVA:
120 V "primary = 0.5 R
480 V "secondary" = 2.8 R

As far as specs go, Acme says that the regulation on all of their iron is in the range of 2% to 4%. I don't have any model specific specifications for the Acme, so 2% to 4% is my best guess. I haven't been able to find any regulation figures for the GE at the GE Industrial web site, but I'm guessing it would have to be 5% or better.

[bob p]

I did some simulation work, and it looks like an 8-10H inductor is better than a 1H for doing the voltage lowering directly with a variable load. You might be able to get a class A without regulating.

What tools are you using to do your sims? When I did my models with PSUD for the Class AB (variable load) amp, I liked inductors in the 5-10H range. The results that I got using the Crowhurst swinging choke nomograms were pretty close. (I wish I had discovered the nomograms a lot sooner!)

For a 100W 2x6550 P-P Class AB1 amp, 560 plate volts, the idle currents were about 125/250 mA mono/stereo, and the full load currents were about 375/750 mA mono/stereo. Data comes from this KT88 spec sheet, at the top of page 4. Some of those currents include 25 to 50 mA of additional current that I added to account for preamp tubes. See Post # 9.

The Crowhurst nomograms yielded the following choke specs for the 100W stereo Class AB1 amp: (one PSU rail for both channels)

Imin  = 0.225 A
Imax  = .750 A
Iopt  = .450 A
MaxDC = 560 V
Lopt  =  2.5 H

For the 100W 6x6550 Class A1 amp, 400 plate volts, for TWO channels, the Crowhurst nomograms yield:

 Imin  = 1.4 A
Imax  = 1.4 A
Iopt  = 1.4 A
MaxDC = 400 V
Lopt  =  500 mH

If we split the PSU rail, so that one transformer provides AC to a separate rectifier circuit for each channel, the Crowhurst nomogram yields the following data for ONE channel: (Post 70)

Imin  = 0.7 A
Imax  = 0.7 A
Iopt  = 0.7 A
MaxDC = 400 V
Lopt  =  1 H

I honestly don't know how close or far off these data would be from your models. Not knowing the internal Z of the transformers, I had to guesstimate it based upon a regulation figure of 2% to 5%, and IIRC I used values of maybe 4R. Refining my PSUD sims with those assumptions, it looked like the Class A amp's power supply rail might give us a DC voltage that wasn't too far off from our 400V target.

[R.G.]

What tools are you using to do your sims?

It's one of the spice derivatives. Pretty good if you know how to put in the parasitics and not trust the built in models.

When I did my models with PSUD for the Class AB (variable load) amp, I liked inductors in the 5-10H range. The results that I got using the Crowhurst swinging choke nomograms were pretty close. (I wish I had discovered the nomograms a lot sooner!)

Yeah, inductors from 8H and up gave an output just over 400Vdc from 150ma to 1.5A.

The inductor needs some damping. I put a 1K in parallel with it and that damped the ringing after a step load change down to about 1 cycle. I was using 4 ohms before the diode bridge and 4 ohms after. That was probably pessimistic.

There is an oddity about startup. At power on, the transient output shows a rise to a peak of about 700V, then tailing off as it goes into L-C filter mode. I think you need to use filter caps that can stand 750V or so. Effectively, 350 to 450V caps in series-parallel. I used four 100's in the sim. Interestingly, with that much inductance in series, the cap doesn't affect ripple much. It's down in the 10V range for all the loads I checked.

For a 100W 2x6550 P-P Class AB1 amp, 560 plate volts, the idle currents were about 125/250 mA mono/stereo, and the full load currents were about 375/750 mA mono/stereo.

For this one, I think I'd go capacitor/regulator. I'll dink around with the inductor-first filter, but I think that getting it stable over load variation (and there's a lot of that in an AB amp) and moreover making it not ring might be harder.

For the 100W 6x6550 Class A1 amp, 400 plate volts, for TWO channels, the Crowhurst nomograms yield:

 Imin  = 1.4 A
Imax  = 1.4 A
Iopt  = 1.4 A
MaxDC = 400 V
Lopt  =  500 mH

Imin  = 0.7 A
Imax  = 0.7 A
Iopt  = 0.7 A
MaxDC = 400 V
Lopt  =  1 H

I'll use those and see what turns out.

I honestly don't know how close or far off these data would be from your models. Not knowing the internal Z of the transformers, I had to guesstimate it based upon a regulation figure of 2% to 5%, and IIRC I used values of maybe 4R. Refining my PSUD sims with those assumptions, it looked like the Class A amp's power supply rail might give us a DC voltage that wasn't too far off from our 400V target.

I don't think it matters a lot. I want to put reasonable values of parasitics in. I'll vary the values and see which ones change things. Most of the parasitics don't matter too much unless you want three-decimal-place accuracy. We're after 1 1/2 to 2 decimal places accuracy.

Unlike some folks, I prefer to put in theoretically perfect components, then dirty them up with parasitics instead of relying on someone else's guess about what the parasitics are. Then I can blame myself for not getting the sim right...

[bob p]

It's one of the spice derivatives. Pretty good if you know how to put in the parasitics and not trust the built in models.

So far I've only used Duncan's PSUD. I've been thinking about things like TubeCAD and Spice. I've actually opened a thread on them, but so far nobody seems to have noticed it.

Yeah, inductors from 8H and up gave an output just over 400Vdc from 150ma to 1.5A.

The inductor needs some damping. I put a 1K in parallel with it and that damped the ringing after a step load change down to about 1 cycle.

I noticed the ringing when I was doing my PSUD models. Because PSUD lacks the ability to model resistances in parallel with the inductors, my only option was to model insane amounts of capacitance in the filter.

There is an oddity about startup. At power on, the transient output shows a rise to a peak of about 700V, then tailing off as it goes into L-C filter mode. I think you need to use filter caps that can stand 750V or so. Effectively, 350 to 450V caps in series-parallel. I used four 100's in the sim. Interestingly, with that much inductance in series, the cap doesn't affect ripple much. It's down in the 10V range for all the loads I checked.

Are you modeling a single stage L-C filter, or a dual stage L-C-L-C filter? My dual stage models looked better.

I did notice the voltage peak to about 750 V in the first 1/4 second or so. It wasn't there in the models that had a big load at startup, but it was there on all of the models that used "standby" parameters at power-on. I guess you could prevent that the big surge by using another resitor/relay, or by just adding a big load to the supply at startup, but I have been thinking about totem-poled caps -- like the CDE 380/381 series snap-ins.

[bob p]

100 W Class AB1. For this one, I think I'd go capacitor/regulator. I'll dink around with the inductor-first filter, but I think that getting it stable over load variation (and there's a lot of that in an AB amp) and moreover making it not ring might be harder.

I ran into lots of problems with ringing in all of my models for this circuit. The voltage drop (and the subsequent ringing) when I modeled the shift from idle to the full load condition was really amazing to see. Needless to say, I thought it was way to ugly, and those models are why I started this thread about getting help with regulated supplies.

Thanks again.

[R.G.]

I guess you could prevent that the big surge by using another resitor/relay, or by just adding a big load to the supply at startup, but I have been thinking about totem-poled caps -- like the CDE 380/381 series snap-ins.

My preference is to use the series caps. I like the snap-ins. I used snap-in radial caps in totem pole on the Workhorse amps to ensure that no combination of startup, nonworking tubes, etc. would ever over-voltage the caps.

Which brings up a delicate point. Almost all tube amps get powered up at one time or another with no tubes or no output tubes in them. That means that the load goes effectively to zero. The regulated option works fine then, but the L-C filtered version becomes a C-only filter unless you also install a minimum load bleeder. That would waste a lot of power in this setup.

My belt-and-suspenders genes want to make the caps able to take the 750V at startup on a continuous basis, just so the no-tubes case does not take them out.

[R.G.]

I ran into lots of problems with ringing in all of my models for this circuit. The voltage drop (and the subsequent ringing) when I modeled the shift from idle to the full load condition was really amazing to see. Needless to say, I thought it was way to ugly, and those models are why I started this thread about getting help with regulated supplies.

Putting a 1K in parallel with the inductor really squelched that.

I still think I'd go with active regulation for the AB case. The class A amp can use just a big inductor, I think.