The definitive "rectifier tube sag" thread

[Chuck H]

Is what I hope this will become

On another thread here a well respected member mentioned that a rectifier tubes resistance DECREASES with current. Well... Doesn't that negate sag? If this is the case then a rectifier tube would act more like a regulator. Which is the opposite of sag. And all this blather from the guru's about rectifier tube compression would be hooey (wouldn't be the first time). Other issues would be the rectifier tube replacement modules (Copper Caps, etc.) or the inclusion of power supply resistance to simulate the rectifier tube effect. Then there's the amps on the market that allow the user to choose between tube or diode rectifiers. In this case I would expect more sag from the diode rectifier since, IF rectifiers tubes actually decrease resistance with current, selecting it would lower voltage but increase dynamics.

Unfortunately this WOULD give credence to all the guru types claims that only a tube rectifier will do for the real vintage sound.

Somehow this subject has been touched on, but glossed over for decades. My hope is that of all the various amp forums out there WE can get it right.

Please opine freely.

[trobbins]

Pretty much all valve diode datasheets that show a voltage drop curve will show a gradual reduction in effective resistance as current increases. That curve in resistance is added to the effective resistance of the power transformer winding, although the tube resistance is the dominant resistance and so will give a slight change in ripple harmonic structure compared to a constant resistance clone.

One advantage of the 'higher resistance' as the valve voltage is close to 0V, is that the transformer winding current ramps down more sedately to zero, and hence reduces dI/dt noise.

However the datasheets don't show what happens when current extends higher than the plot - where the resistance then increases to high levels as voltage saturation starts to set in. The 5Y3GT curve I just looked at shows a plot up to the 420mA steady state peak plate current. During start-up the current is allowed to peak up to 2.5A, where I reckon the resistance has certainly turned the other way - but I can't confirm that at the moment.

During a cranked output, with B+ sagging heavily, I'd be thinking that some amps have their valve diodes conducting beyond their steady peak level.

The 'sound of ripple' is a benefit in a valve amp.

[Chuck H]

After my post I went straight to Duncan PSUD2 and ran some simulations. It seems that Duncan's program corresponds to what you posted. And, to put it in perspective, even when the tube diodes aren't saturated the resistance curve seems somewhat shallow with only about a 10% or 12% difference in sag from idle as current demands approached what would be typical in a cranked guitar amp.

[defaced]

Then there's the amps on the market that allow the user to choose between tube or diode rectifiers. In this case I would expect more sag from the diode rectifier since, IF rectifiers tubes actually decrease resistance with current, selecting it would lower voltage but increase dynamics.

I don't think trends are enough here; I think you need to consider magnitude. A rectifier tube may decrease in resistance with increasing current, but never approach the low to nill resistance values of a diode. Here are some numbers for the 1N4007 (plot from the Fairchild datasheet attached):

0.7v at 0.05 amps = 14 ohms
0.85v at 0.4 amps = 2.1 ohms.
1.6v at 20 amps = 0.8 ohms.

So the trend is the same, but I suspect the magnitude is very different.

[Enzo]

And keep in mind that terms like increase and decrease are not quantitative.

If we consider a rectifier tube to have a resistance, then doubling the current through it would also double the voltage drop. Now what if the resistance of the tube decreases 10% from the increased current. The doubled voltage drop would be diminished a little by that 10%, but would still increase as the current does. So even though the resistance might decrease some amount with current rise, the overall sum total still shows sag.

[nickb]

What this is happening is that the voltage drop increases more slowly with increasing current, not that it negates sag. To negate sag the resistance would have to go negative.

[pdf64]

Yes, while the effect of tube rectifier sage may be regulation of a sort, the actual result will probably end up being less well regulated than if silicon rectification was used.
Stepping back, another factor here though is B+ winding resistance; tube's specify a limiting minimum value, whereas with silicon it can be as low as practicable.

[J M Fahey]

Agree, it may well do so, but the main question is how much and how significant it may be.

As a parallel example: bacon may have, say, 90% fat content.
Any bacon showing , say, 85% fat may truthfully be labelled "low fat", "reduced calories" , "healthier", etc. , point it's *almost* as artery clogging as the regular one.

Same with tube rectifier "negative resistance", only way for it to be important is if were so high that tubes supplied constant voltage between, say, 50mA and 200 mA (say, from idle to full power) and in that case yes, the supply would behave as if regulated absolutely negating sag.

I guess nothing like that happens, not even close.

[Chuck H]

What I wasn't allowing for, that confused me and has been covered very well at this point, is that the resistance shift is so small as to be almost indiscernible from a plain ol' resistor. I'd probably consider the difference negligible enough to ignore, but there's just enough difference with some higher voltage drop tubes (like a 5u4) that I wouldn't fault anyone for applying the "mojo" label. So... I guess we ended up in the same place WRT whether the "rectifier tube difference" matters.

[trobbins]

The 5Y3GT plate resistance, as deduced from the plate voltage/current curve, starts at >1kohm, and levels at about 210 ohm for current greater than 200mA. A typical PT with secondary CT may have a winding resistance of something in the range 100-350 ohm. That makes the start and finish of the current conduction waveform each half-cycle in a valve diode supply quite soft in comparison to a clone using say a constant 220 ohm resistance and ss diode. That would likely reduce the higher frequency harmonic content in the ripple voltage waveform.

[pdf64]

Empirically and from memory, in a regular '50W' tube guitar amp, compared to a GZ34 (vintage and current production), with silicon diodes, at idle about 80 ohm gives about the same VB+, whereas at full crank it's about 20 ohms.

[Pedro Vecino]

...Ted's COPPERCAP™ produces a linear-voltage drop, not the exponential-voltage drop that actually occurs with vacuum rectifiers, ie:

Vd = (Ip/K)^(2/3)

...so, there's basically similar voltage drop at idle and maximum current, but certainly not the same voltage "sag" across the middle of the current range, for sure!

...the difference between a straight-line between two points and a cantenary-line hung between the same two points...the straight line is a constant voltage, while the cantenary line will ALWAYS be lower (more sag) thru the middle portion, but ramping back to the same voltage at both ends, ie: idle and max, like I mentioned before.

...the linear voltage will sound more 'tight' and crisp, while the exponential/cantenary voltage will exhibit the 'warmth' and 'squish' that we typically associate with vacuum tubes (and, here with vacuum rectifier tubes) and blues playing.

...measuring the power-supply voltage (B+) under load (swept sinewave, 100-6KHz) with a real GZ34 and Ted's replacement, will demonstrate the small but noticable difference in B+ (and thus likewise Vp) levels at the plates (yes, my stock '69 Super Reverb OT isn't perfectly balanced).

Topic: http://music-electronics-forum.com/t10343/

[eschertron]

The 5Y3GT plate resistance, as deduced from the plate voltage/current curve, starts at >1kohm, and levels at about 210 ohm for current greater than 200mA. A typical PT with secondary CT may have a winding resistance of something in the range 100-350 ohm. That makes the start and finish of the current conduction waveform each half-cycle in a valve diode supply quite soft in comparison to a clone using say a constant 220 ohm resistance and ss diode. That would likely reduce the higher frequency harmonic content in the ripple voltage waveform.

Fascinating topic, glad I stumbled across this thread

Do we know that the change in resistance is purely an electrical phenomenon (and so, instantaneous) and not thermo-electric so that the rectifier's change in effective resistance has some hysteresis that lasts over several mains cycles? I'm thinking of light bulbs, et al, that increase resistance when hot. Not sure if there's a similar - but opposite - effect that can be going on here?

[pdf64]

It was discussed on MEF as few years ago, someone clever wrote that the voltage drop followed a power law. I think that's what it was, I'm afraid I didn't fully grasp it. I've searched in vain for it.

[eschertron]

I charted the Vd = (Ip/K)^(2/3) curve and it looks like this to me:



...and note the curve is self-similar for any value of Ip or K

As concluded in the linked thread on the copper cap, there's a sharper slope at low current, but above that it flattens out a bit. Enough to hear?