Fluke 8800A design-detail

[wiss]

Hi

Can someone explain to me why transistors are used as diodes in discreet designs?

An example is Q10 on the input of Fluke 8800A.

Regards
  Jonas

[SeanB]

Lower leakage current, more controlled forward voltage drop and a well defined reverse breakdown characteristic. Mostly used for the very low leakage and because the case is opaque a very low sensitivity to incident light.

[wiss]

Lower leakage current, more controlled forward voltage drop and a well defined reverse breakdown characteristic. Mostly used for the very low leakage and because the case is opaque a very low sensitivity to incident light.

Thanks! Are bjt's intrinsically better than diodes in the leakage or is it just practical? I know that a bc547 has lower leakage than a 1n4148 but this diode is optimized for fast switching... Can a bjt take sustained reverse polarization? (one could make a ref-amp out of a double-bjt).

[SeanB]

Cheaper than the low leakage diodes and will not really mind fast switching, just be aware of capacitance. Not realyy usable as a reference as the junction is both noisy and degrades with time under reverse breakdown, it does make a very nice broadband noise source. To make a reference you need a proper zener or a buried junction like in a proper zener reference unit.

[drtaylor]

Fluke has long used the venerable 2N3904 as low leakage diodes. Shorting C to B as the anode, and E as the cathode makes a fairly high current low leakage diode. If I recall correctly, you can expect reverse leakage less than 10 pA in that configuration. However, the voltage rating is around 6-7V where the connection zeners. Using this in a protection circuit, this zener action is useful. Using two 2N3904s in opposing series makes an extremely low leakage bipolar zener diode. This is used in many DMMs to protect the ohms circuitry (including the 8060A). Using just the B-E junction makes even a lower leakage diode, but very low current. You'll often see this used in sample and hold circuits where the current flow is minor, and low leakage is key. I've tested some brands of 2N3904 to less than 1 pA in this configuration. No other kind of diode can touch it. Even very expensive FET based diodes will be worse when tested over temperature. So in low voltage circuits requiring an extremely low leakage diode, 2N3904s in various connections are very useful. Best of all 2N3904s are dirt cheap. Some manufacturers outperform others, so testing more than one brand might be prudent. Photodiode effects are usually not a problem with the black epoxy coat used in most TO-92s.

Also the C+B to E connection is quite fast in turn on and turn off. Again, if you can handle the <6V voltage rating. Works great in logic circuits 5V and below.

You can also use the B-C junction as a low leakage diode. This leaks around the 20pA range (if I recall correctly, been awhile). But it has the advantage of working at higher voltage. Also, not as high current as the C+B to E configuration. But still better than any signal diode that is not FET based. Not very high speed either, but I don't recall what the speed is.

The only time I would use a FET based low leakage diode is in higher voltage circuits.

[BravoV]

drtaylor, how is the capacitance on the 2N3904 based diode ?

[Harvs]

Best of all 2N3904s are dirt cheap. Some manufacturers outperform others, so testing more than one brand might be prudent.

They are excellent.  However, in my experience they need to be hand picked.  Even those from the same batch can have at least an order of magnitude spread of leakage. e.g. I've tested two of the same brand at the end of this page:
http://www.electronicsinoz.info/ideas-for-the-lab/testing-reverse-leakage/

At least hand picking them in TO92 package is a trivial exercise if you only need a few.

[drtaylor]

I just looked up the Fluke 8800 schematic. Specifically Q10 is used as a low leakage diode connecting the input voltage to ±22V clamps. This prevents the rather high leakage current of the zeners to be reduced to that of the 2N3904 until 22V + 2 diodes drop is exceeded. It also provides a reference point for the bootstrap circuit. I would guess that CR6 is also low leakage, but I'm not sure why that circuit just didn't use back to back transistors. This whole circuit is interesting as it shows how very high impedance was achieved up to ±20V, a necessity in a truly precision DMM. Basically the Input Amp circuits power supply adjusts (bootstraps) to the input voltage. With precision matched high beta BJTs this circuit achieved >1000M ohms of input impedance for voltages up to ±20V. The leakage current of the zeners CR7 and CR8 would have compromised that input impedance if Q10 and CR6 didn't isolate the zener leakage.

Now days there are higher voltage amps available, but not always in high precision types. So the power supply bootstrapping technique is still a valid approach for precision circuitry.

[drtaylor]

Best of all 2N3904s are dirt cheap. Some manufacturers outperform others, so testing more than one brand might be prudent.

They are excellent.  However, in my experience they need to be hand picked.  Even those from the same batch can have at least an order of magnitude spread of leakage. e.g. I've tested two of the same brand at the end of this page:

Indeed, Fluke had Motorola (now On Semi) 2N3904s selected for low leakage (that is if my memory serves me correctly). Motorola was the original manufacturer of this ubiquitous transistor, along with its PNP compliment the 2N3906. So I tend to trust them more than other brands (trust but verify). BTW, I have found certain brands of surface mount 3904s to have much higher leakage, might be due to a thinner encapsulation allowing light detection. Most brands of MMBT3904 in SOT-23 are OK but smaller packages seem to suffer. If low leakage is the need, testing is required to ensure proper circuit performance.

[wiss]

I just looked up the Fluke 8800 schematic. Specifically Q10 is used as a low leakage diode connecting the input voltage to ±22V clamps. This prevents the rather high leakage current of the zeners to be reduced to that of the 2N3904 until 22V + 2 diodes drop is exceeded. It also provides a reference point for the bootstrap circuit. I would guess that CR6 is also low leakage, but I'm not sure why that circuit just didn't use back to back transistors...

What strikes me as odd is that there is a diode in one direction and a diode-bjt in the other, either way two of the same would be cheaper, right?

[wilheldp]

I just picked up an 8800A off eBay for $25 (local pickup).  Good deal or no?  It is operational (have photos of the display lit up), but no probes and no indication of calibration.

[retiredcaps]

I just picked up an 8800A off eBay for $25 (local pickup).  Good deal or no?

$25 is what I try to pay for a non working Fluke. So anytime I can get a working Fluke for $25 locally, I think it is a good deal.

[wilheldp]

I picked it up this afternoon.  It works like a champ.  I tested it side by side with my Rigol DM3058E with a recent calibration, and the Fluke was within a count of the Rigol on DCV.  I'm planning on starting a new thread with a comparison, mini-teardown, etc.  Suffice it to say that I am pleased with my purchase.