Valve Shunt Regulator Design Walk-Through

[Mike H]

There's another possible problem, you've got KT88 wired as triode and you're assuming it will behave as a triode, but it won't, as it's using a tetrode model. The gain will be more than it truly would be triode mode.

In my library All_Valves2.zip > new_triodes.txt I've got KT88 as triode, made it only recently, plus I've just found it's also got the 6N1P from triode_nh.txt!

[IslandPink]

Hmmmm... more
I do have series R for the inductors, Mike, I think 69R .
I replaced the current load with a 2K resistor, which is roughly correct .
All the voltages are still tiny tiny .

[Mike H]

OK well that's stopped the current source making things negative.

Have you followed the Voltage from V1, to see where it goes low?

This is like fault finding a real circuit.

[IslandPink]

There's another possible problem, you've got KT88 wired as triode and you're assuming it will behave as a triode, but it won't, as it's using a tetrode model. The gain will be more than it truly would be triode mode.
In my library All_Valves2.zip > new_triodes.txt I've got KT88 as triode, made it only recently

I was hoping the Stephie KT88 tetrode would be good enough to work properly when wired up this way, maybe not I might have to ask elsewhere then regarding the limits of these models .
In any case, I put in your KT88 triode model instead ...
...and the voltages are all still tiny tiny .

[IslandPink]

There's nothing significant anywhere now Mike - even at the transformer secondary or around the diodes - it's as dead as a Dodo . Maybe there's a power cut caused by the snow .

[Mike H]

Oh wait I've just noticed '.op'

Is that all?

Should be .tran

E.g.:

.tran 0 10 (10 seconds)

I've never used .op

Tried it just now, just get a list of node Voltages in a pop-up.

You should use .tran, creates a second waveforms window. Then you click on circuit nodes to see the waveforms at those nodes.

[IslandPink]

OK, I suppose if you're starting from an AC source, the DC operating point analysis may not 'see' the 50Hz input, and so you get no voltages .

I tried the .tran 0 10 0 analysis , and clicked on the B+ line.
It slowly ( really really slowly ) plotted a voltage curve coming up, as if the power supply had just been switched on . It eventually plateaux-ed out at about 300V . This makes me wonder if a 390AC source is 390V Pk-Pk instead of RMS as I was thinking .

So, I might have to cut the front of the circuit short for efficient DC operating points analysis ( and some of the other stuff too ) and just put a combination DC + AC ripple source there, either just do one choke or maybe both . Hmm, having just done this trace of the B+ line out to 10sec, I wonder if I can analyse the residual ripple and its spectrum ? - I will have a look .

[IslandPink]

Looks like an AC voltage source in LTSpice is specced as Pk-to-Pk , not RMS .
Just re-running it now - should come up to about 425V .
Interesting that the 0-10sec simulation of B+ showed a similar start-up curve to PSUD ( for the passive section ) with a small overshoot and reaching nearly full B+ at about 0.4 sec.

[IslandPink]

OK , not bad, 416V . I could adjust to get more like 450 .
Good news : the KT88 is conducting - about 45mA
Bad news : the 6N1P sections are only conducting 0.1mA .
Hmmm ...
Maybe explains why the ripple is in the 10-20mV level and not 2 to 3mV .

[Mike H]

OK, I suppose if you're starting from an AC source, the DC operating point analysis may not 'see' the 50Hz input, and so you get no voltages .

Yes!

According to help .op only tells you what the initial start-up Voltages are, which in this case are zero, because V1 starts at zero (start of sine)

V1 is a signal source, AND rest of circuit needs time to react to it, ergo needs a transient simulation over some period of time.

Even if you were using a V source as a DC supply, still need time for the circuit to start up and stabilise from that supply. So still .tran.

Looks like an AC voltage source in LTSpice is specced as Pk-to-Pk , not RMS .

Yes!

Interesting that the 0-10sec simulation of B+ showed a similar start-up curve to PSUD ( for the passive section ) with a small overshoot and reaching nearly full B+ at about 0.4 sec.

Yes!

For power supply sims I nowadays tend to delay the input for a second or two, to better see the start-up point, there is a 'delay' parameter in the attributes for V1 SINE

[Mike H]

Next step is insert a proper transformer into the input

[IslandPink]

Maybe so, Mike.
In the meantime, I'm looking at a trace of the B+ voltage I just ran, with all the valves working approx correct now. It seems to have a curious meandering ( admittedly only 10mV or so ) that has a long period, down in the sngle-Hz range .
I wonder if this is real( doubtful) or some subtlety I've got wrong in how I'm running the analysis ?

[Mike H]

Hunting?

Dunno, but I've received your file OK thanks, will have a look when I get a mo.

You've upped the input AC to 680V now?

[IslandPink]

You'd think it's hunting, but could it do that, with the info it has ? - maybe .
Voltage - 680 , he he - the screen cap view of these things when the circuit is shrunk down goes all nuts - the numbers appear to change ! I'll have to post the circuit separately I suppose, when it's important . No, it's at about 550V I think. The resistors above the 6N1P are also NOT 60k and 68k - they are 50k and 56k !

Meanwhile - here's a cool thing - attached.
This is the analysis of Zout that was suggested by Dave Cigna on the Intact Audio site.
Set up two current loads at the end - one for DC ( 0.26) and one for the AC component ( 0.03 ) .
Do a decade frequency analysis ( note, I removed the rectifier section & subbed a DC voltage source here, to get the AC smoothing ripple out of the picture ) .
Analyse V(out)/I(I2) : this is the B+volts over the AC current load .
[ V(out) in this case is V(n003) on my circuit ] . Do this by editing the 'Visible traces' options box to create this function.

Hey Presto ! - you get a direct trace of the Zout in ohms.
Result looks pretty tasty - only 3ohms at 10Hz dropping to 0.6ohms by 10kHz.

[Mike H]

Pretty sure you have to do it with one current source only, not with a second one in parallel.