Tube data sheets - inaccuracy

[Brit01]

Hi guys,


Maybe some of you have seen my OTL circlotron thread but I'd like to start one in regards to tube data sheets and real measurements.

Looking at the 6080 data sheets I set

-40 bias
@ 100 volts

should give me about 60 mA.

I had about 100 volts +- 10 volts
-40 grid bias +- 5 volts


I have 3.9R cathode resistors to avoid current hogging and to measure current.

I measured 140mA and 115mA across the 2 cathodes of 1 triode!!!!!
0.55 volts/3.9 * 1000 = 140
0.45/3.9* 1000 = 115

So definitely an adjustment of the power supply voltage and grid bias needs a large adjustment to get it down to 80-100mA max.

Have other people had similar experiences with data sheet info?

 

[DF96]

Valves can easily be 20% different from the typical values given in a data sheet. Your bias voltages could be up to 10% out too, so together that adds another 21%. So your result could be out by 45%. NOS valves from good manufacturers often measure high.

Nothing very surprising in your results. People brought up on Spice might find this shocking.

 

[SY]

Wait until you play with FETs and find 5-fold spreads in gm and idss...

 

[Brit01]

Wait until you play with FETs and find 5-fold spreads in gm and idss.

I've 2 FETs in this design also as source followers. These are out also.

I need to match the current through them better.

Maybe with a SS sink on them instead of a resistor?

 

[DF96]

Design circuits so accurate matching is not so much of an issue?

 

[Arnulf]

I've 2 FETs in this design also as source followers. These are out also.

How can source follower present a problem ? Unity gain followers are about the only component that works reliably as anticipated without applying loop feedback (well, the intrinsic feedback may have something to do with this ). Others will deviate from published specs, more or less. Take JFETs for example, it is common for datasheets to contain identical diagrams for same component, showing vastly different parameters.

 

[Brit01]

How can source follower present a problem ?

Wasn't sure.

So I can rule out the FET's as causing a problem in differences in grid bias needed to cancel out DC offset?
The current flowing through each FET on each rail doesn't need to be equal?

Got a 10-12 volts difference in grid bias on each rail to balance the tubes and cancel out DC.

 

[kevinkr]

<snip>

Got a 10-12 volts difference in grid bias on each rail to balance the tubes and cancel out DC.

Remember the information in the datasheet is typically the mean (in this case data set average) of a large number of samples, so no individual device is necessarily going to have performance that is reflective of that mean, but any sufficiently large sample should..

This is completely normal both for devices within an envelope, and for device to device.

 

[Sch3mat1c]

I bet your PSRR is about -3, too.

/Something about designing circuits actually worth using... >

Tim

 

[Arnulf]

So I can rule out the FET's as causing a problem in differences in grid bias needed to cancel out DC offset?
The current flowing through each FET on each rail doesn't need to be equal?

Got a 10-12 volts difference in grid bias on each rail to balance the tubes and cancel out DC.

Adjust the driver tubes' operating points (or whatever it is that is driving your source followers) accordingly or do it the SS way (loads of feedback to take care of sloppy design).

 

[Brit01]

oads of feedback to take care of sloppy design

Sloppy design?? how dare you sir!

Got a cathode grounded driver feeding into an LTP with a cascode DN2540 as tail resistor setting the current for a balanced split signal to feed the FET source followers.

I'll have a look at the operational points further of the drivers.

BUT if the LTP is well balanced the operational points of the drivers shouldn't make any difference no? Just affect the linearity/distortion and voltage swing limits.

Well let me play with it anyway and take a closer look.

 

[Arnulf]

Sloppy design?? how dare you sir!

Have a look at common SS ciircuits. Rarely is attention paid to keeping intrinsic offset to a minimum, instead gigantic feedback is used to bring it down into (tens of) milivolts range.

 

[Brit01]

How can source follower present a problem ?

varying the current through the source followers (drain - source) via a trimmer had a dramatic effect on the grid bias.

I used a trimmer on 1 to match them.

 

[wakibaki]

I measured 140mA and 115mA across the 2 cathodes of 1 triode!!!!!
0.55 volts/3.9 * 1000 = 140
0.45/3.9* 1000 = 115

Welcome to engineering!

w

 

[Brit01]

intrinsic offset to a minimum,

Do you have some info regarding intrinsic offset?
Trying to google it but not coming up with much. Even in wiki.

 

[Sch3mat1c]

"Intrinsic offset" isn't a keyword that I'm familiar with. "Input offset" is a typical specification of amplifiers, referring to the offset required for a neutral output, or alternately, the voltage drop across a voltage follower. For transistor op-amps, this is typically a fraction of Vbe, depending on how well matched the input transistors are, and how matched their quiescent currents are.

For a balanced amplifier (i.e., an LTP), it is somewhere in the gain region, which might cover a whopping 5V for tubes, depending on how many stages are present. Offset is a DC thing, and most tube amps are not DC coupled, so it's not even an issue (and those that are, are usually of silly design -- circuits that only work because tubes are as accurate as they are!).

Tim

 

[DF96]

"intrinsic offset" is not a keyword, but a piece of English. It means that offset which is intrinsic i.e. built-in. I assume it means the offset arising directly from the Vbe mismatch, or a circuit current mismatch. There are other sources of offset too (e.g. thermal).

 

[Brit01]

Interesting. Thanks guys

I'll be focusing on each stage in detail and optimizing their balance and operating points very carefully.

 

[Arnulf]

"intrinsic offset" is not a keyword, but a piece of English. It means that offset which is intrinsic i.e. built-in. I assume it means the offset arising directly from the Vbe mismatch, or a circuit current mismatch. There are other sources of offset too (e.g. thermal).

Yes, precisely, I'm not a native English speaker

If a circuit is well constructed and adjusted one can get away with very little offset added to the signal in completely DC-coupled amplifier (which, due to lack of any schematic, is what I'm guessing is what Brit01 is tinkering with: LTP driving source followers driving output tubes, everything DC-coupled). But most SS projects neglect that and rely on loads of openloop gain and negative feedback to get rid of said offset.

 

[Miles Prower]

Maybe some of you have seen my OTL circlotron thread but I'd like to start one in regards to tube data sheets and real measurements.

Looking at the 6080 data sheets I set

-40 bias
@ 100 volts

should give me about 60 mA.

I had about 100 volts +- 10 volts
-40 grid bias +- 5 volts


I have 3.9R cathode resistors to avoid current hogging and to measure current.

I measured 140mA and 115mA across the 2 cathodes of 1 triode!!!!!
0.55 volts/3.9 * 1000 = 140
0.45/3.9* 1000 = 115

So definitely an adjustment of the power supply voltage and grid bias needs a large adjustment to get it down to 80-100mA max.

Have other people had similar experiences with data sheet info?

Doanworrybouddit. All practical components deviate from their specs. What counts is if the deviation is within the expected range. This is true of both solid state and hollow state.

It's especially true of types like the 6080. This was intended for use as a series pass regulator, not an amp, and so care in matching the sections wasn't taken to the same extent as a device intended primarily for balanced operation. Still, if your plate and bias voltages are correct, and you're getting more than twice the expected plate current, I'd check those 6080s again: they might have gone gassy.

It's for that reason you should include bias adjustments, so you can bring the actual plate currents to the design nominal values. Forch, hollow state is a good deal more forgiving than solid state. That's a good thing since back in those days they didn't have precision, 0.1%, laser trimmed resistors. +/-20% was common, +/-10% pretty much standard, and +/-5% was "precision".

Also, 3R9 cathode resistors aren't gonna be of much help in preventing current hogging. Resistors that small will do for transistors, but not triodes. 100R would be more like it.