Regulate B+ of a parafeed output stage?

[Deke609]

Is it possible to regulate the B+ feeding a parafeed output stage with a solid state reg?  Or is this just inherently a no go?

I have a high voltage mosfet-based regulator designed by Linear Audio. It works perfectly when given a purely resistive load.  470VDC input (filtered through my Kaiju rebuid CLCLC filter) gives a stable 450 VDC output. But it doesn't work when I connect the output of the reg to the plate chokes. Voltage bounces around +/- 1V rms or so -- seemingly at random.  At the designer's suggestion, I tried putting a 110uF cap across the reg output (this might have helped a bit), a 50K resistor across the output, both cap and resistor across the resistor output, and ditto with the OPT disconnected. No dice. voltage is still unstable -- not bad, but not much better than just feeding the plate chokes from the CLCLC filter. 

The reg designer has suggested that the amp might be motorboating -- which I take to mean oscillating. Without the reg I hear no noise at all (and certainly no put, put, put sounds) - so if the amp is oscillating it is at a very low level. I have not listened to the amp with the reg.

It's been suggested that the reg, which I am told sources current to maintain voltage, potentially conflicts with the plate choke, which I understand releases and stores voltage to maintain current.  If that's the case -- that there is simply an inherent conflict between the reg and choke -- then I can stop trying to get the regulator to work, patch up the wiring and get back to listening.

Any thoughts or insights would much be appreciated, including any suggestions as to what type of reg might work or be worth exploring.

cheers and many thanks in advance, Derek

[Paul Birkeland]

We have done this a few times.  Long ago we put out a modification kit for our old Paramour monoblock kit that converted it to a fully shunt regulated #45 amp.  We have typically used a hybrid regulator, as solid state current sources are super convenient and tubes in the shunt device move the heat to the top side of the chassis. 

Put a crappy test speaker on the amp and see what happens.  If the designer of the regulator you're using doesn't mind, you could post it up here and maybe we could spot something.  How much compliance does the regulator need?  What happens if you set the output to 420V?

[Deke609]

Thanks PB. I'll check with the designer about posting his schematic here - it is public, so hopefully he will agree. The reg only needs 2-3 volts compliance - and I am feeding it nearly 20V compliance. Instability does not appear to change with voltage output of the reg. That said, it is very difficult to change the output of the reg when it is connected to the 53H/160DCR plate chokes in my Kaiju rebuild - the reg uses a 5K trimmer pot to adjust voltage, and in the Kaiju rebuild it is very difficult to adjust output voltage with the trimmer resistor. The voaltge seems to "want" to stay where it is. But with a purely resistive load, the reg adjusts easily.

RE speakers - my Kaiju no longer has speaker outs, but I suppose I could clip lead one channel to some speaker wire. But the only speakers in the house are my son's Overnight Sensations -- would be be putting them in jeopardy?  I doubt it, b/c the instability is limited to about +/- 1V RMS - at most. Usually it bounces around with 500mV of 450.

cheers and thanks, Derek

[jan.didden]

Hello Bottleheaders, Derek pointed out this forum to me. I'd be very interested if someone here could shine some light on this regulator issue.
The design info and circuit can be found at my website linearaudio dot nl slash t-reg-hv-regulator.
One thing I hadn't mentioned to Derek is that the regulator circuits need a minimum load of about 2 - 3mA to function, which normally isn't a problem, but may somehow come into play with an inductive load. That is why I suggested to add a 50k load resistor.

Jan

[Paul Joppa]

It has long been claimed that shunt regs are "good because they can take current flow in both directions". That describes a capacitor; does it describe the Linear Audio regulator?

I admit, I can't see a plausible situation where reverse flow would occur except high-level transients at low frequencies, but I've puzzled over it for a long time.

[Deke609]

Many thanks Jan and PJ.

For ease of reference I attach Jan's schematic for the latest version of his reg.   I don't know how the reg works, but I can offer the following info for making sense of the circuit: (1) Current protection -- the reg has a current limiter controlled by the divider R8 and R6, parallel with R16. The user sets the current limit with R16. My version is slightly different -- there is no R16 and I have set the current limit with R* and R6. Current limit (shut-off) is set for 300mA.  I have noted this info about current protection in hopes of avoiding confusion as to how R6, R8 and R16 contribute to volatge regualtion -- sohrt answer: they don't. (2) Output voltage -- the output voltage is ballparked with R11. U1 generates a reference current of approx. 580 uA.  So for every 100K ohms at R11, you get approx. 58V.  I have a 787K ohm resistor at R11. Final output voltage is set using the 5K trimmer at VR1.  I doubt this latter info is relevant, but I have stated it just in case.

It has long been claimed that shunt regs are "good because they can take current flow in both directions". That describes a capacitor; does it describe the Linear Audio regulator?

I can't speak for Jan, but his following statements to me may answer PJ's question:

Quoting Jan: "I believe that the problem with the large inductive load is the fact that an inductor tends to create a large varying voltage with varying current. If that voltage gets above the T-reg output, it can't 'go' anywhere - a regulator is designed to source current, not to sink it."

To troubleshoot this, I'd like to do 2 things: First, scope the instability of the reg output. Problem - with my X10 probes, my scope can only handle 400 VDC max. And my B+ is 450 VDC. Plus, at X10, it might be hard to focus in on the instability. So my thought is to block the DC with a capacitor and AC couple the scope using X1 (no attenuation) since the instability tends to be less than 1V RMS.  What I don't know, and would appreciate advice on, is how low a frequency I might need to measure -- as I think this will inform the capacitance value I use.  Voltage rating will of course 600V+ to withstand the full 450V B+.

The second thing I'd like to try is building up another of Jan's regs for use in a different parafeed amp: my S-II 45.  If the same instability appears, then I think I can conclude that the issue is with the interaction of the reg and parafeed output stages generally, not just with the particular parafeed output stage of my Kaiju rebuild.  But if it works with the SII-45 amp, then this will point to something funny going on in the Kajiu rebuild -- and I can focus my efforts accordingly.

All thoughts, speculations, suggestions welcome and much appreciated.

cheers and thanks, Derek

[jan.didden]

The T-reg is a series regulator, not a shunt regulator, so cannot sink current. I have never found a case where the load actually returns current to the regulator, but it may be an inductive issue.
Derek, your amp will probably continue to work at a B+ of 400V if that is necessary for the test setup.
If you couple the scope via a cap be sure to take a large one to capture the very slow varying voltage. In your case I would actually DC couple the scope and see if you can see the variations atop of the B+.
Your idea to power another amp this way is a good one and will give us another data point.
A wild guess: at which point do you ground the scope probe ground clip? Is it possible that there is some impedance between that ground point and the grounding of the T-reg, like a CM choke? How does it look with the scope probe (or multimeter) grounded t the same point where the reg is grounded?

Jan

[Deke609]

Thanks Jan.

Derek, your amp will probably continue to work at a B+ of 400V if that is necessary for the test setup.

Good point! That could make things much easier. If I have a suitable value of resistor for R11, I can drop raw B+ to 400VDC and try to regulate to 390V.  Hopefully I can capture the voltage instability.  If not, I will try caps in series with the probes.  My questions are: How "big" should the big caps be? and whether they should be film for low ESR, or whether 'lytic caps will work just fine.

Your idea to power another amp this way is a good one and will give us another data point.

I will try this on the wknd. I will stuff a new board just for this purpose, as I hope to use a T-reg in the other amp in any event.

A wild guess: at which point do you ground the scope probe ground clip? Is it possible that there is some impedance between that ground point and the grounding of the T-reg, like a CM choke? How does it look with the scope probe (or multimeter) grounded t the same point where the reg is grounded?

I've only used my multimeter so far, with COM connected to the same star ground to which the reg -ve output is connected by 7cm of 20 awg wire. There is nothing between the -ve output of the reg and star ground.

cheers and thanks, Derek

[jan.didden]

Derek  I can send you an extra T-reg set if you want.
Another question: did you check the reg input voltage when the output was moving around?
Is the input stable under that condition?

Jan

[Deke609]

Thanks for the offer, Jan, but I am all set. I now have two T-regs built, and another two unbuilt ones. If all else fails, I can use them with CCS-loaded driver stages in other projects.

I did look at the input and output simultaneously using two multimeters (no scope). They both bounce around, but not entirely in sync (but that could be because I used two different models of multimeter: a good Fluke and really cheap meter). But the magnitude of the voltage variation looks to be about the same at the input and output.  My first suspicion was that the reg wasn't regulating at all and simply dropping voltage.  Maybe that's the case?  And so isn't sufficiently loaded? 

Another odd thing (at least to me): as I think I mentioned in my first post, when the T-reg is loaded with the parafeed output stage of the 300B amp, it is very difficult to adjust the output voltage with the 5K trimmer -- I can turn it multiple times with negligible effect. But with the purely resistive load, it adjusts easily.  I don't know whether that's a symptom of the reg not regulating or of the choke fighting the reg, or possibly something else that I haven't thought of.

I have started the process towards testing a T-reg in my parafeed 45 amp - but am too tired after a day of home reno to work any more on it (fixing squirrel damage to my house trim under the roof overhang - who knew squirrels can chew wood like beavers?).  The B+ power supply in the 45 amp has two branches - one for each channel - so I will try regulating only one channel. And if it works, I will try regulating only one channel in the first amp. Just to see.

cheers, Derek

[jan.didden]

That is very strange that the input voltage also fluctuates. Do you remember what the approximate input-output difference was?

Something else occurred to me when I woke up in the middle of the night ;-) why not run the 300B parafeed with the plate choke shorted? The amp wouldn't mind, all DC conditions remain the same. It wouldn't amplify but you don't need an input signal for this.
I wonder whether in this case the output also varies?

Jan

[Deke609]

That is very strange that the input voltage also fluctuates. Do you remember what the approximate input-output difference was?

I just assumed that it is my mains instability. My 45 parafeed amp does the same thing with no reg on the B+ of the output stage. I verified this yesterday when taking measurements for fixing Vout of the new T-reg board.

I don't recall the precise difference between Vout and Vin. But that is easy to test and I can do so today.

Something else occurred to me when I woke up in the middle of the night ;-) why not run the 300B parafeed with the plate choke shorted? The amp wouldn't mind, all DC conditions remain the same. It wouldn't amplify but you don't need an input signal for this.
I wonder whether in this case the output also varies?

Another good idea! Many thanks. I will try that now and report back.  The DC conditions will slightly change b/c I'll be taking the 160 ohms DC resistance of the choke out of the equation - but that should be ok.

cheers, Derek

[Deke609]

I am completely baffled. Same instability with the choke disconnected and the T-reg feeding the 300B anode directly.  Although just this second I realized that I also disconnected the cap-coupled output transformer from the anode as well - not sure if this would affect things. My guess is "no", since the coupling cap blocks DC and I assume the plate resistance of the tube provides a sufficient load to the T-reg. But that's just my weakly-educated guess.

So perhaps the amp is oscillating? 

Or I've messed up how I wired the T-reg to the amp circuit.  Two things come mind regarding how I've wired the board to amp: (1) I am using the +ve Vin of the reg board as a shared B+ connection for both the driver stage (hybrid shunt regulated and CCS-loaded 5670 tubes) and the 300B output stage; (2) since the -ve Vin and -ve Vout of the reg are directly connected by a board trace, I am using the -ve Vin as a shared connection to star ground (0V and chassis/earth ground) for the reg and the last cap in the power supply -- a 1500 uF film cap.  Does anyone see a problem with either of these two wiring arrangements?

Edit: and I retested the T-reg with a purely resistive load (50K) and it worked perfectly. Feeding the reg from the amp's power supply, with a 50K resistor across the reg output, driver stage still sharing a common B+ connection at Vin, and output stage disconnected.  Is there any chance that the 50K resistor is creating the appearance of regulation, and not the reg itself? 

I need to get back to squirrel-proofing my roof, but I will return to this this evening. Provided I have the energy, I hope to test the new T-reg in the 45 parafeed amp.

All thoughts, suggestions, speculations are most welcome and appreciated.

cheers and many thanks, Derek