High Impedance VM

[jhadhar65]

I don't happen to have a VTVM and the idea struck me regarding the possibility of making a DMM or analog solid state meter high impedance by using a series resistance, such as a 10M resistor, in one probe.  I'm sure if this could be done, I would have read/heard about it by now.  I don't really know how meters work, but I assume they function according to Ohm's Law and the big resistor might throw off the results.  I'm guessing it would be possible to do the math and account for the 10M if needs be, right?  That would be a pain, but for quick jobs like grids, it wouldn't be too bad.

Any thoughts on this idea?

[PRR]

You seem to have multiple misconceptions.

Most VTVMs were 10Meg or 11Meg input.
http://www.tone-lizard.com/VTVM.htm
http://www.cfp-radio.com/documentations/Sylvania-VTVM.pdf -- 13 Meg PDF file!

Most DVMs are 10Meg input.
http://www.multimeterwarehouse.com/m9803spec.htm

So what's the difference? Use your dang $4.95 DMM.

VTVMs are usually a 10Meg divider in the box, and to reduce AC-loading the circuit, a 1Meg in the probe. The 1Meg must be bypassed for AC and Ohms work. If you forget to put it in for DC, the meter reads 11/10 high. A little thought shows that a 22meg at the probe gives 32Meg input and half-voltage readings you can easily correct by eye.

[jhadhar65]

Hmmm... so what's the deal with DC loading?  Seems like I was told somewhere you can't get accurate voltage readings on PI grids with a DMM (and I can't) due to loading... that you'd need a VTVM with it's higher input resistance for accurate measurements?

[PRR]

> I was told somewhere you can't get accurate voltage readings on PI grids with a DMM (and I can't) due to loading...

No. So measure the cathodes. You know that the grid is never far from cathode (much-much smaller than plate-cathode voltage). If the cathode is at 100V, odds are the grid is at 97V, 95V, somewhere in there. Anyway it does not matter. If the plan says cathode should be at 95V, and you find 100V there, nothing is wrong at the grid.

> that you'd need a VTVM with it's higher input resistance for accurate measurements?

The standard VTVM is 10Meg, 11Meg, rarely 15 or 22Meg. Not so different from a DMM. And for the same reason: 99% of stuff works far under 1 Meg impedance, divider resistors over 10Meg are problematic, so ~~10Meg is a sweet-spot for an all-purpose volt meter.

[PRR]

If you "must" know hi-hi-Z points like Fender longtail grids...

There are ways. Real high precision needs some complexity.

But for the accuracy you need in tube-work, and the speed you (don't) need for DIY, this will work good enough.

It takes power from the amp being tested.

It reads 2V-3V low, a factor of the specific MOSFET you get, temperature, some other slop. So before you read a grid, read a nearby point which can be measured with DMM and with this buffer and DMM. In the longtail, the cathodes are an obvious point. They don't sag (much) when you poke a DMM at them, and they are very near grid voltage which keeps other errors small.

Say the cathodes read 80V on DMM, and 77.5V with buffer between cathodes and DMM. So it reads 2.5V low, in this voltage range, today.

Now poke the grid with the buffer and read the DMM. Say it reads 75.5V. But you know the buffer reads 2.5V low. So the true grid voltage is 78V.

Oh: if the Gate+10Meg is not connected to anything, the DMM reading is completely random. The ultra high Gate impedance may hold the last voltage it saw for many seconds, or drift up or down to any other voltage. It only is valid when connected to something.

High-high-Z work is painstaking. The least dirt or moisture spoils the reading. In this case, the Gate pin, the 10Meg, and the wire to the grid being tested, must have LOW leakage to anything. Don't put these critical points on terminals, especially cheap phenolic. I would epoxy the MOSFET to a board or chassis (the tab is electrically "live"; gluing face-down is valid), keeping epoxy off the gate lead area. Solder the 10Meg right to the MOSFET Gate leg (no, you can not hurt the MOSFET with solder). Trim the leads, solder a good insulated clip wire to the other end of the 10Meg. None of this should touch anything! Air is your best friend. Wash around the gate leg with fresh rubbing alcohol, dry over a hot lamp.

The 10K, 220K, and meter-point can be on perfboard etc, though I would be tempted to solder right on the MOSFET, then shrink-tube the contraption so the tab and legs can't short to stuff. Clipleads will let you connect to B+ (not plate!) and ground and DMM.

[jhadhar65]

Holy crap... that is excellent!!!

>If the plan says cathode should be at 95V, and you find 100V there, nothing is wrong at the grid.

That's a good point.

Still, I'd like to be able to measure for the fun of it at least and the MOSFET buffer will be perfect.  Does it matter which MOSFET?

I don't know squat about MOSFETs and I'll have to grab one at Radio Shed.  The only one they've got that says MOSFET is the IRF510.  The IC page is here in case you happen to see something else they list that's actually a MOSFET, too.

Duh, I just saw the specs on the drawing.  So much for detective work.  Anyway, the voltage rating on the IRF510 is only 60V between the drain and the source, so that one won't work.  I'll have to order one.

Thank you very much, PRR, for this!!!

[HotBluePlates]

As a sidebar, I have a couple of VTVM's that could be rigged for a high iinput impedance of 122M. It didn't come from the factory with that input impedance, and the manual states that it will wreck the calibration in some ranges (the manual still provides the instructions for the conversion if the remaining ranges would actually benefit from the modification).

But! Due to the bootstrapping effect on most long-tail pair, split-load inverters and cathode followers (there are a number of ways to bias these circuits, and not all use bootstrapping) the input impedance at the grid could be several hundred-Meg and quite higher than even 122M. So you would still be likely to load the circuit down and measure a voltage significantly less than what is really present.

Which is why some of the old schematics showed voltages readings, but then also gave a spec for the voltmeter used (like 20,000 ohms per volt). You were likely to not get the same voltage reading as what was written on the schematic, but you could also assess whether you were in the ballpark.

[jhadhar65]

You're right there and I didn't know some of that... thank you!  It may be futile in some cases, but I think I'll build the buffer anyway and keep it around.  I'll play with it and see how I like it.  I need to get a good VTVM at some point anyway... and I will... at some point.

[PRR]

> the input impedance at the grid could be several hundred-Meg and quite higher than even 122M.

I got curious what the actual effective DC impedance is and what errors result.

This is good 'ol 5E3, a typical self-bias cathodyne, which jacks itself up in the same way as a Fender long-tail (but is easier to draw).

Three versions: naked, my 11Meg Heath VTVM, and HBP's 122Meg-mod VTVM.


When you are not looking, grid really sits at 52.2V. When I poke it, I see 20.6V, less than half of "right". With a 122Meg meter, we see 45.5V, 15% low.

So then I plotted meter-to-ground loads for meters from 1Meg to 1,000Meg.


For -this- circuit, the DC input resistance is about 17Megs. For low-low error we need a meter much-much-much higher than 17Meg. 11Meg sucks. 122Megs sucks less, but still skews the reading enough to be interesting/confusing.

Even at 1,000 Meg (an unusual and unlikely value), we have 2% error from meter loading. 2% accuracy is more than good enough for tube audio amplifers, and utterly meaningless error for gitar work, but can be significant in some other field.

[jhadhar65]

What I'm doing is tossing the idea around about building myself a tube tester/matcher of sorts.  I thought about having a variable control voltage capability for obvious reasons, but I guess I'm stuck for how to measure it.

I've got other projects ahead of this one, so I'm just mulling it over at this point.


EDIT:  Well, wait a second.  I wouldn't have to use PI topology anyway - even for tubes that will be destined for PI slots, would I?  If not, then it really doesn't matter about the meter.  I could wire in any old meter for continuous readings.  Do real tube testers have sockets wired as a PI for testing tubes going into that position?

[HotBluePlates]

I wouldn't have to use PI topology anyway - even for tubes that will be destined for PI slots, would I?  If not, then it really doesn't matter about the meter.

Exactly. The thing that is a real killer in the (mostly) phase inverter topologies is when you self-bias a LTP or split-load by returning the grid resistor to the "ground-side" of the self-biasing (cathode) resistor, which is then sitting atop a load or tail resistor. Bootstrapping results in a much higher than expected input impedance, and you calculate it in a roundabout manner.

No need to worry about all that stuff if you have known voltages that you will apply to a typical common-cathode stage.

And the "Mod-Meter" I spoke about is a Hewlett-Packard 425A Microvolt-Ammeter. If I really needed to figure a grid voltage, I'd probably not measure volts at all. That particular meter can measure picoamps, but could be understandly touchy in that range. I'd probably set it up to measure the low-low current through the grid resistor, measure the resistance separately, and measure the voltage across, say, a long-tail pair "tail resistor". Adding the measureable volts across the tail resistor to the calculated volts across the grid resistor should come reasonably close to the real grid voltage.

For other d.c. voltage readings, I have a Hewlett-Packard 412A which has the 122M input impedance on most ranges stock. (EDIT: The 412A has 100M impedance on the 100mV range, and 200M on all ranges from 300mV to 1kV. 10M and 30M is the impedance on the lower voltage ranges.) Even with the high impedance, the voltage calibrator originally used to calibrate this meter notes that you'll experience some loading effects due to the input impedance. Fortunately, the calibrator is a pretty low-impedance output, so the error is small on most ranges.

[jhadhar65]

>If I really needed to figure a grid voltage, I'd probably not measure volts at all. That particular meter can measure picoamps, but could be understandly touchy in that range. I'd probably set it up to measure the low-low current through the grid resistor, measure the resistance separately, and measure the voltage across, say, a long-tail pair "tail resistor". Adding the measureable volts across the tail resistor to the calculated volts across the grid resistor should come reasonably close to the real grid voltage.

That's a good idea!

>Bootstrapping results in a much higher than expected input impedance, and you calculate it in a roundabout manner.

Yeah, I remember that.  I've got a MV in a design I've been working on that uses a pot in place of the 1M grid resistor on a concertina.  Because of the bootstrap situation, FYL helped me work out the new pot value.  As a result, the pot became 100k instead of 1M.  This contrasts with a common grounded k, where effective impedance is only the resistance value of the grid (leak) resistor, right?  It's the difference in impedance between the meter and the grid resistance that determines loading - the higher the effective resistance, like for an LTP, the greater the effect of loading.  If that's true, then I would assume there still must be some loading even on a grounded k's 1M (or so) grid, but that loading must represent a current equal to or less than leakage.  Otherwise, measuring those would also be a problem.  Does that sound right?

[HotBluePlates]

Sounds right to me.