Using a zener diode to lower screen grid supply voltage

[cdemike]

So the recent thread about lowering screen voltages got me thinking about 1970s Marshalls with 6550s. With higher mains voltage now than in the 70s, those amps could be running those 6550s pretty hard with the screens 60-80V above the 440V maximum (if going by the up-rated specifications per the 1970s GE and Tung Sol datasheets), and as others have noted before, reviewing the load lines shows pretty brutal operating conditions already for a pair of 6550s loaded at 3.4K or a quartet at 1.7k. I've seen Andy Le Blanc's voltage divider method of reducing screens to be at lower potential than the plates, but that was a fix primarily aimed at fixing amps using OTs with fairly high DCR causing the plates to idle at lower potential than the screens, and adjusting the voltage divider values to provide that large a drop starts to dip into situations that place a lot of strain on the power supply with low resistance to ground. Alternately, raising the value of the top leg of the voltage divider brings with it all the other issues that come with increasing screen resistor values like unwanted additional compression.

What about using a Zener to lower the screens' supply voltage? Although it still won't likely be the most ideal operating conditions, it seems like this might be some insurance against catastrophic power tube failures after screen grid meltdowns. Plus it, like the voltage divider and unlike a regulated screens supply, would allow for the screen grids to still sag with the remainder of the power supply unlike a regulated screens supply, also helping the amp to stay sounding somewhat close to the way it would stock despite the additional safeguard.

I might try this out in one of my amps, but I wanted to run it by y'all first.

Screen grid idea.PNG

[B Ingram]

So the recent thread about lowering screen voltages got me thinking ...

What about using a Zener to lower the screens' supply voltage? ...

Try it & see if you like it.

________________________________________________________

What is the purpose of the choke between plate & screen filter caps?
- It provides filtering, but the minimum possible voltage drop, to keep Peak Plate Current high.
- The choke is always followed by a cap, and the screen is tacked to the cap (with or without a series screen resistor to limit screen dissipation).

If we are worried that "screen volts are too high" then why do we need a choke? If we are scratch-building, we can knock the voltage down with any method we prefer:
- Dropping/Decoupling resistor between successive filter caps.
- A gas regulator tube (like an 0C3, etc)
- Your Zener Diode.
- A MOSFET wired to impose a set voltage-drop.
- Etc.

If we are focused on vintage amps possibly running too hot with modern outlet voltage, should we really be modifying those? Should we not simply use a variac & knock the outlet voltage down to what the amp can handle? (I know I do the latter with 1960s Vox amps.)

[Stevem]

Would not the use of a Zener there add in switching noise right where we don’t want it?

[cdemike]

I think I will give this a shot sometime this week there's some safety or reliability concern. I can't think of anything specific, including from a dissipation perspective as long as the diode's power rating is up to the task. I figure with a 500V screens node, I'd want a Zener that can stand up to the same current as the screen grid resistors. If those resistors are rated at 1K and 5W, then the maximum breakdown voltage for a 1W Zener would be 14V, and 70V for a 5w Zener.

If we are focused on vintage amps possibly running too hot with modern outlet voltage, should we really be modifying those? Should we not simply use a variac & knock the outlet voltage down to what the amp can handle? (I know I do the latter with 1960s Vox amps.)

Yeah to me this probably makes the most sense, especially since fixes like lowering screens voltage won't really resolve other issues like high heater voltage. I guess if you're giggling with an old amp, though, and don't feel like brining a variac this could make some sense?

Would not the use of a Zener there add in switching noise right where we don’t want it?

You're right -- this would add switching noise, and I was curious about the magnitude of the switching noise after you raised this. I assumed that the maximum magnitude would be a small fraction of the breakdown voltage, but I'm reading that I was way off and that it can be as high as tens of volts if the Zener is actually operating under avalanche breakdown. My understanding is that the transconductance of beam tetrodes' screen grids is generally very low, though, so maybe this wouldn't be too much of an issue on a practical basis? There also could be some protection given push-pull phase cancellation. If it is still a major issue, though, that could explain why this isn't already something in common use.

If the screens' B+ node is adequately filtered and sufficiently decoupled from the phase inverter and further downstream preamp nodes, AC bypassing the Zener with a fairly large capacitor could shunt the noise back into the B+ rail. With 1K screen resistors, 8uF would place the corner frequency below 20 Hz.

[sluckey]

Would not the use of a Zener there add in switching noise right where we don’t want it?

The zener will only switch on once then it remains on until you turn the amp off. It's working with dc voltage unlike the rectifier diodes that work with AC voltage and switch on/off at a 60Hz rate.

[syscokid]

Watching this thread with interest!

Yeah to me this probably makes the most sense, especially since fixes like lowering screens voltage won't really resolve other issues like high heater voltage. I guess if you're giggling with an old amp, though, and don't feel like brining a variac this could make some sense?

A good friend of mine who saw time in the studios, told me that the Amprx Brown Box is quite prevalent in the studios. A bit pricey though…

IMG_1892.jpeg

[R.G.]

Think about the Vintage Voltage Adapter from geofex: http://www.geofex.com/article_folders/v ... ntvolt.htm.

It's a common problem, and a smallish (compared to the PT) auxiliary transformer can fix that without a lot of tinkering, no variac needed, and without dissipating the power spent in chain-of-diodes voltage droppers.

[nuke]

I built a simple voltage bucker into an all metal Tripp-Lite outlet/surge box. It's just a small transformer wired in series with the line. I use it for all my old tweed amps, especially my 1960 Bassman. It really only wants to see 110 vac on the mains.

One of the interesting aspects of the Silvertone 1484 I restored a few months ago, is the voltage multiplier power supply. The highest multiplier point is 425v for the plates, the screens are on the next leg down, and provide about 350v for the screens. The amp makes about 27 watts from 2x 6l6GC/5881, but with the nice, stiff rail for the screens, retains good punch. The reduced power and screen input should provide for long output tube life. It still had the original "Silvertone" RCA 6L6GC black-plates in when I bought it at a garage sale 20 or so years ago. They were in absolutely perfect running condition, lasted a long time in another amp.

I think there's some advantage to reducing the screen supply, especially now that a pair of output tubes is about $50 USD or more. If you can find some reasonable way to create that power node in your amp without spending a lot of money or wasting a bunch of current.

Back in the day, they didn't really care how long the tubes lasted, since they were relatively cheap and available, and everyone wanted more and more power. So they just ran them right up there.

Little different today, not as much need for power output, but the wallet output has gotten a lot heavier.

[cdemike]

Gave this a shot and it sounded good! The amp I tested this on isn't a 70s Marshall like mentioned in the original post (just wanted to test the idea), but the amp is a Marshall-style build running 2x KT88s at around 520V on the plates. This amp sags that way I would expect a Marshall to (generally about 50-70V lower than idle at the OT CT depending on how the power tubes are biased), so I think it should be a reasonable analogue for what this would look like in an actual Marshall.

My multimeter died right before I got a complete set of voltage measurements, so I didn't get accurate absolute voltages. However, I was able to determine that the plates consistently stayed positive compared to the screens from a DC perspective. That was also true in the arrangement I had before -- the typical 5w 1K flying screens resistors mounted to each socket but with a shared 750R resistor connecting both flying resistors and the screens nodes as additional protection. So in other words, I don't think there was a specific benefit from the perspective of keeping the screens more negative compared to the plate in terms of overall DC voltage. However, being able to tune the magnitude of the potential difference was pretty interesting, especially given how many amps push the screens well past their datasheet maxima as with the 6550s in 1970s Marshalls.

I thought it was interesting that I also didn't notice a big increase in noise from a qualitative/subjective standpoint. I have a new multimeter in the mail, so I'm planning to measure the difference in idle noise when it arrives.

Definitely a comparative band-aid fix compared to using a variac, R.G.'s vintage voltage adapter, or other means of supplying the amp with a lower potential AC source. But in terms of other uses, I thought it could be helpful in newer builds where the screens are getting pushed pretty hard or where the builder is shooting for a specific operating point that requires lowering the screens. For example, I know Martin has voiced concerns over the operating point placing the load line below the knee in builds similar to mine where big bottle tubes have screen voltages near the plates and with typical Marshall output transformer ratios. A quick look at the 6550 datasheet shows bringing the screens down to the datasheet maximum 440V in an amp with the same B+ brings the load line much closer to the knee.

[martin manning]

The amp I tested this on isn't a 70s Marshall like mentioned in the original post (just wanted to test the idea), but the amp is a Marshall-style build running 2x KT88s at around 520V on the plates. This amp sags that way I would expect a Marshall to (generally about 50-70V lower than idle at the OT CT depending on how the power tubes are biased), so I think it should be a reasonable analogue for what this would look like in an actual Marshall.

With 520V on the plates (sagging to ~460) Vg2 should be about 100V below the plates at full power to avoid cooking the screens. What primary impedance are you running?

[cdemike]

With 520V on the plates (sagging to ~460) Vg2 should be about 100V below the plates at full power to avoid cooking the screens. What primary impedance are you running?

Primary impedance is 3.4K for 2x 6550A's. New multimeter arrives today, but the DMM on life support shows 513V on the plates and 505V on the screens at idle with the resistors. With a continuous strong signal that drops to 460V on the plates and 405 on the screens. Large transients drop that to 458V on the plates and 395V on the screens. I converted it back to the resistive setup to get a better comparison of baseline and do some other voicing tweaks, and I thought I'd revisit the Zeners after responding to this post and hearing any feedback about operating points.

A quick review of the online load line plotters I'm aware of shows pretty different results, and I suspect some of that may come down to differently modeling the tubes' behavior at higher screens voltages than the datasheets provide. Results are at the bottom of this post.

Reviewing the load lines operating with the 100V guideline you mentioned, it looks like doubling the primary impedance (i.e., graduating the 4 and 8 ohm taps to 8 and 16 ohm loads, respectively) produces a nice looking load line per one of the simulations, passing though the lower portion of the knee. The second simulation still looks pretty unfavorable, though I'm not aware of early 70s export-spec Marshalls having a reputation for melting screen grids as compared to domestic versions with EL34's. That leads me to think the more optimistic simulation may be closer to reality, but going off amps' service reputations isn't exactly rigorous.

If this looks reasonable, I'm happy to give this a try to see how it sounds. I recently installed the 6550A's to compare the sound vs KT88s, and I'm happy to compare either/both with the screen grid resistors versus with the Zener on the screens node. My power supply seems a little stiffer than some 100w power supplies I've seen documented, but if it is in the ballpark for "typical" operation with big bottles, I'd be curious how sustainable this kind of operation would be since the screens are sagging well below V(g2) max.

Sim 1: https://bmamps.com/ivds.html
Sim 2: https://www.vtadiy.com/loadline-calcula ... alculator/

(Near) Stock Operating Conditions
Sim 1: "full sag" conditions (continuous strong signal)

6550 full sag.PNG

Sim 2: "full sag" conditions (continuous strong signal)

6550 full sag 1.pdf

Sim 1: Idle

6550 idle.PNG

Sim 2: Idle

6550 idle 1.pdf

Screens Lowered by 50V Zener, Primary Z = 6.8K
Sim 1: "full sag conditions" (continuous strong signal)

6550 full sag with revisions.PNG

Sim 2: "full sag conditions" (continuous strong signal)

6550 full sag with revisions 1.pdf

[martin manning]

This is from my Excel load line plotter. Looks like you are ok given the reduced voltages at high power conditions.

[martin manning]

A new feature I've been tinkering with for the past few days... Dumping out coordinates at data sheet reference Vg2 for plotting measured data or design values on the factory curves. Retro engineering ;^)