• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Rod Coleman CCS/gyrator help

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Wondering if someone can help with an issue I'm having with my implementation of Rod Coleman's DHT supply. Initially, it looked like R4 at 1K was too high to allow Q3 to turn on fully, as I wasn't getting full current out. So as I lowered it further down, eventually to 100 ohms (damn, 1W) I got the expected 5V across a 4 ohm resistor load.

However, I am also getting a little bit of 120Hz ripple through, about 12mV p-p, which I assume means the gyrator isn't working well enough. Please take a look at the operating points; any ideas ?

Raw supply has about 100mV of ripple.

Thanks.
 

Attachments

  • ccs.pdf
    132.1 KB · Views: 1,386
Operating points look OK, but they suggest the MJ15032 needs more headroom than the types I used. I suspect that increasing the raw supply would be the best way to improve the performance. Or short the bootstrap resistor R2 to try it.

The capacitor in the gate circuit may need a few hours under power to reach full performance, if it's an OS-CON or polymer type.

The performance may be better still with a FET (logic level type) and a film capacitor in the C2 position [470n..1uF] and R1 = 820k or so. In that case the raw supply would need to be 12..14V according to the FET's working headroom..

Low base resistor value suggests a disappointing working Hfe for the MJ15032. It may improve with more headroom.
 
The 2SC5511 that I built with are osbolete, but TOSHIBA lists a couple parts that specify equal or better performance, eg

2SC5000: Co 90pF, typical gain 200 at 1A, operates at 10mA base current for 1A collector with about 2V Vce.
2SC4881: Cob 45pF, typ Hfe 280/1A and Vce=1V, 10mA base current for 1A collector with about 2V Vce.

Both are in isolated TO-220, and would work really well in both posiitions.

Don't know how easy these are to procure outside the TOSHIBA sales network. But they may make a better solution than raising the raw supply voltage, if the MJs are not so happy at low headroom.
 
Thanks, Rod.

I have a few thoughts on this. I don't require an isolated package, but if there is benefit to doing so, I can try it out. I don't know if adding the extra volume of the heatsink to the collector compromises capacitance leakage, I would think it would be small. My experience has been the few isolated packages available also carry some detriment in performance.

If I short R2, would it also be suggested to connect C1 negative to circuit common, resorting to the standard capacitance multiplier topology? That bootstrap has me miffed. For that matter, what really is the purpose of the upper portion of the circuit, simply to remove ripple, or to present a high impedance? I don't see how C1 presents a high impedance to the filament. Why not just present raw DC to one end of the filament if I have low ripple already?

It should be easy enough to test each circuit (top, bottom) independently. So I'll experiment with the minimum voltage required for proper operation of Q3, and the actual value of R4 necessary (it shouldn't be 100 ohms, that doesn't even follow the datasheet to an order of magnitude). Testing the upper portion might be a little more tricky, but I'm under the assumption that ripple rejection is the primary measurement.
 
Gyrator section is there to reject ripple, and also to present a high impedance looking out from the filament. There seems to be some effect on the anode current if it finds an ac-leakage path through long wires/bad caps/power trafos in the heating circuit. The idea is to keep it all contained. I should mention that this assertion is unsubstantiated hand-waving; it's audible, but not susceptible to honest analysis.

return C1 to negative if trying the R2 short.

I like the iso-TO220s to prevent shorting accidents. If you're ok without, just find an NPN with low Cob (~ 20..70pF) and high 1A Hfe. Check the Vce condition for the Hfe measurement.
 
Point taken about leakage; I don't think it's just handwaving, there is both theory and merit to the idea. Maybe you haven't actually measured it, but I agree it is there. A must to have either shielded or split bobbin filament transformers.

I was thinking about how to minimize the effect of this leakage if resorting to a single ended CCS approach (raw DC on one side of filament). The SET this will be serving is fixed (but adjustable) bias, so the cathode will be grounded. I assume the raw dc to have leakage capacitance from xfmr/choke to ground, which is a source of nasties to enter into the cathode.

So, if I choose to ground the positive end of the filament, the nasties essentially are at the same potential as the cathode, and there is little tendency to inject themselves into the cathode. The negative end of the filament, elevated 5V below ground, could be a source of problems, but this side is terminated to the high impedance node of the CCS.

Just a thought.
 
There's a good test to be made of that. Some 300B designers, including the Western Electric applications staff, have claimed that the negative filament connexion is best for returning the anode current. If that is right, a PNP CCS on filament +, with Filament-neg grounded would sound better than the NPN. I have seen no explanation for it, though.
 
Well, definitely looks to be poor performance of the transistor at low Vce. Even at 5V performance was mediocre. In addition, low values of the base resistor really did help, with 75 ohm being best. You could see the ripple of the supply impressed across the base resistor, which means ripple on the base drive current. This resulted in ripple across the load, albeit low. One could use a small CCS to drive the base, but I still have the Vce issue.

Based on that result, I didn't even bother checking performance of the upper portion. I'll pursue the source in the + side and just ground the -.
 
Well, definitely looks to be poor performance of the transistor at low Vce. Even at 5V performance was mediocre. In addition, low values of the base resistor really did help, with 75 ohm being best. You could see the ripple of the supply impressed across the base resistor, which means ripple on the base drive current. This resulted in ripple across the load, albeit low. One could use a small CCS to drive the base, but I still have the Vce issue.

Based on that result, I didn't even bother checking performance of the upper portion. I'll pursue the source in the + side and just ground the -.

What about a simple MOSFET gyrator on the (+) side? Given a well-isolated transformer, I don't see why *both* ends if the filament need high AC impedance. The single gyrator circuit will set the DC filament voltage constant, isolate the PS ripple, and provide a high AC impedance to the signal current imposed in the filament. You can make the cathode connection to the (+) or (-) sides according to your religious preference ;-) or use the resistor tap.

Cheers,

Michael
 

Attachments

  • filament-gyrator-supply-zener.png
    filament-gyrator-supply-zener.png
    5.8 KB · Views: 1,465
If you're only using one transistor, remember that the CCS also serves to eliminate the cold-filament surge currents, which are large enough to affect reliability.

I set the CCS to run 4.9V on my 300Bs 5 years ago,, and they're still 4.91V, with no voltage control at all.

If your trafo is well isolated,, the single sided regulator maybe good enough. you could check by addiing an HF choke (2mH ferrite perhaps) and rechecking the sound.

The rectifiers would need careful snubbing to be sure no recovery pulses get through to the fiilament, - keeping in mind that these are current pulses.
 
Well, I only use split bobbin transformers, which should be sufficient enough of isolation for my purposes. Raw supply is made of schottkys feeding a 35mH choke input.

Once the raw dc is running, I break out the scope and check all points in the circuit for ringing. Also have a hall effect current clamp for monitoring current throughout. I have found that the most important item to snub is the transformer, and have not seen (by the scope) improvement with 10nF high VDC caps across the diodes. By zooming waaaay in on the turn on/off transitions of the transformer secondary, I experiment with RC values for a nice smooth waveform with the minimal amount of capacitance and highest resistance possible. With my 56VA xfmrs that results in 10nF 1K. Doesn't seem to be any need to address the choke, everything looks peachy. Ripple at the capacitor bank is a nice clean 120Hz.

Next selection will be an experiment to see how the LT1086 in the positive portion will work. Heatsink is big enough to give it 5-6V of headroom. If this is met with success, I'll consider more elaborate topologies. I am not convinced a discrete bipolar design has the open loop gain necessary. FET probably is the smart approach. Unfortunately, I've always had difficulty understanding how to work with FET's, especially when looking at the datasheet. Tubes, bipolar, no problem. FET's in the active region, confusing.
 
If you're only using one transistor, remember that the CCS also serves to eliminate the cold-filament surge currents, which are large enough to affect reliability.

I set the CCS to run 4.9V on my 300Bs 5 years ago,, and they're still 4.91V, with no voltage control at all.

If your trafo is well isolated,, the single sided regulator maybe good enough. you could check by addiing an HF choke (2mH ferrite perhaps) and rechecking the sound.

The rectifiers would need careful snubbing to be sure no recovery pulses get through to the fiilament, - keeping in mind that these are current pulses.

Oh, I forgot to mention that soft start is provided by the time constant R2, C2 (R3, C3 in the earlier sketch).

Trying a small choke is a good idea. The effect shoulld be measurable as well as audible. It might be best with an RFC and the one transistor... Maybe the choke would take care of any (Schottky) diode switching noise that would want to couple through the Crss or the bias network.

I suppose there will be tweaks needed, and I left out some of the obvious things like snubbers. My aproach lately is to start dead-simple and add only what's found to be needed or what produces a noticeable improvement. It's surprising how simple things can be (sometimes) and not need further improvement.

ZZflux: If you stare at the MOSFET data sheet charts for "On-Region Characteristics" and "Transfer Characteristics" long enough, you should be able to reconcile the MOSFET behavior with what you know about vacuum tubes and BJTs.

Thanks!

Michael
 
Michael:

Alas, it never will make sense, I suspect.

One other question: I see in the LT1086 datasheet it requires an output capacitor for stability, whereas the older LM317 does not require it. I of course want to avoid using the capacitor on the output, as it defeats the high impedance goal. I know many people have used the LT1086 in CCS mode; do you simply dispense with the cap and deal with it? The load is constant, after all.
 
Michael:

Alas, it never will make sense, I suspect.

One other question: I see in the LT1086 datasheet it requires an output capacitor for stability, whereas the older LM317 does not require it. I of course want to avoid using the capacitor on the output, as it defeats the high impedance goal. I know many people have used the LT1086 in CCS mode; do you simply dispense with the cap and deal with it? The load is constant, after all.

my experience is that even the lowly lm317 will oscillate when used as a CCS. I've had terrible luck with them for filament supplies. On the other hand, the CCS circuit here has been totally stable for me.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.