Diode voltage drop stability/accuracy

[mawyatt]

I'd say passing analog signal through a diode will cause distortion. Anyway, what's the use-case, do you have schematics?

Diodes have quite some tempco (2mV/C or so), they self-heat when passing signal, and they are highly non-linear (i.e., voltage drop depends on current). So, without a feedback look it's bad to have in a signal chain.

The diodes are "biased on" with a DC voltage to act as a switch----remove the bias, & the "switch" is off.

The analog signal is normally a much lower voltage level than the bias, hence, it cannot cause the diode to cut off or saturate as this voltage differential between signal & bias ensures it appears in a substantially linear part of the diode transfer function.

Such diode switching allows a transceiver or similar device to use the same special modules (such as filters) in both the transmit & receive modes.
The ham transceiver I am currently fixing uses this technique extensively.

The diodes commonly used in analog signal switching, especially RF, are usually PIN Diodes which have a very long carrier lifetime. The signal period is much less than the carrier lifetime, so the signal (if small) produces little modulation of the diode junction, thus the junction introduces little distortion.

Best,

[David Hess]

I'd say passing analog signal through a diode will cause distortion. Anyway, what's the use-case, do you have schematics?

Diodes have quite some tempco (2mV/C or so), they self-heat when passing signal, and they are highly non-linear (i.e., voltage drop depends on current). So, without a feedback look it's bad to have in a signal chain.

The diodes are "biased on" with a DC voltage to act as a switch----remove the bias, & the "switch" is off.

The analog signal is normally a much lower voltage level than the bias, hence, it cannot cause the diode to cut off or saturate as this voltage differential between signal & bias ensures it appears in a substantially linear part of the diode transfer function.

Such diode switching allows a transceiver or similar device to use the same special modules (such as filters) in both the transmit & receive modes.
The ham transceiver I am currently fixing uses this technique extensively.

The diodes commonly used in analog signal switching, especially RF, are usually PIN Diodes which have a very long carrier lifetime. The signal period is much less than the carrier lifetime, so the signal (if small) produces little modulation of the diode junction, thus the junction introduces little distortion.

The example of diode switching that I gave works down to DC and is often done with microwave schottky diodes because of their low capacitance, but it is also sometimes done with transistors.

PIN diodes used for RF switching can handle much higher power levels.

[TimFox]

PIN diodes for high-power RF switching are run with large DC currents in the ON state (to generate a large stored charge in the junction for low AC resistance) and high reverse voltage in the OFF state (to sweep that stored charge away, making a small capacitance in the depleted junction).
The AC current multiplied by the waveform's period must be small compared to the stored charge to give a constant AC resistance.

[vk6zgo]

I'd say passing analog signal through a diode will cause distortion. Anyway, what's the use-case, do you have schematics?

Diodes have quite some tempco (2mV/C or so), they self-heat when passing signal, and they are highly non-linear (i.e., voltage drop depends on current). So, without a feedback look it's bad to have in a signal chain.

The diodes are "biased on" with a DC voltage to act as a switch----remove the bias, & the "switch" is off.

The analog signal is normally a much lower voltage level than the bias, hence, it cannot cause the diode to cut off or saturate as this voltage differential between signal & bias ensures it appears in a substantially linear part of the diode transfer function.

Such diode switching allows a transceiver or similar device to use the same special modules (such as filters) in both the transmit & receive modes.
The ham transceiver I am currently fixing uses this technique extensively.

The diodes commonly used in analog signal switching, especially RF, are usually PIN Diodes which have a very long carrier lifetime. The signal period is much less than the carrier lifetime, so the signal (if small) produces little modulation of the diode junction, thus the junction introduces little distortion.

Best,

The diodes used as RF switches in the Icom IC575 are 1SS53, which are plain old silicon signal diodes, a close equivalent of the 1N4148.
Many years ago, I repaired several Sanyo CB radios which used diode RF switching, using unknown silicon diodes off an old computer board.
Some designs of radio used silicon diode switching to connect or disconnect different crystals to the one oscillator.

PIN diodes are nice, & obviously to be preferred for professional equipment, but if silicone diodes are biased hard on, they pretty much look like a resistor for small signal levels.

[David Hess]

PIN diodes for high-power RF switching are run with large DC currents in the ON state (to generate a large stored charge in the junction for low AC resistance) and high reverse voltage in the OFF state (to sweep that stored charge away, making a small capacitance in the depleted junction).
The AC current multiplied by the waveform's period must be small compared to the stored charge to give a constant AC resistance.

My point is that while PIN diodes rely on stored charge for switching, current mode switching using diodes does not.  You do not have to rely on stored charge and recovery time to make an RF switch with a diode.  Diode bridge mixers work this way.  Many oscilloscope channel switches work this way.

[exe]

Or connect the collector to the +supply rail and have a hFE buffered "diode" for a positive peak detector with an NPN... (You could do the same with a PNP & negative rail for a negative "peak" detector)

Diode is a floating device, you can just reverse NPN for negative rail)

[Kim Christensen]

Or connect the collector to the +supply rail and have a hFE buffered "diode" for a positive peak detector with an NPN... (You could do the same with a PNP & negative rail for a negative "peak" detector)

Diode is a floating device, you can just reverse NPN for negative rail)

I know... I was thinking of using it like an active detector (Collector connected to power). Basically, a common collector amplifier without bias. (class-B)

[TimFox]

For a different purpose, I have used the other diode (collector-base) of small-signal transistors for a low-leakage with reasonable breakdown voltage diode in analog or protection circuits.
The characteristic of that diode is tabulated in the transistor data sheet as I_CBO, when the emitter is open-circuited.
Another possibility for very-low-leakage diodes is the gate-channel (source connected to drain) diode of a low I_DSS JFET.

[Infraviolet]

I came across an idea in Horowitz and Hill (diagram shown at https://electronics.stackexchange.com/questions/68019/compensating-the-forward-voltage-drop-of-a-diode-signal-rectifier) for compensating for a diode's drop.

Consider that circuit, but with a cap between the output and ground, it would then at as a peak detector with R2 acting to drain away voltage over time if too long between peaks.

Would it also work to compensate for temperature and other variation in the diodes' properties (assuming they were a paired diode built in a single IC or were discrete diodes very close together on a PCB)?

I had a go at simulating it and found it definitely had trouble maintaining proportionality between the smoothed output voltage of the detector and the amplitude of the incoming signal, it seems this circuit magnifies the way in which diode voltage drops vary with current in passing through the diode.

What about that circuit but using BJT NPNs as "diodes" (  c+b-->|--e  ) instead, as had been discussed here?

Is there any way to easily achieve the type of negative feedback that an op amp peak detector provides to account for diode variations of manufacturing/temperature/current using BJTs instead of op amps? (for my project working at 3MHz for which I've no very-fast op-amps available, and the peak detector configuration for an op-amp really exposes any slowness)

Thanks

[MrAl]

Hello,

It looks like that circuit lowers the input impedance 10 fold which could be a problem.  Maybe the two bias resistors would be better off at 10k each or maybe just the resistor that goes to +5v stays at 1k while the other is increased to 10k.  Look at the circuit with and without the bias resistors and bias diode and notice the difference other than the conduction point of the series diode.  With the extra diode and bias resistors the input impedance gets lowered quite a bit.  The whole idea is to bias the second diode with a source that varies with temperature just like the main diode, so both diodes have to be the same and have to track over temperature to some degree.  Increasing R1 to 10k should not bother that but would allow the input impedance to stay 10 times higher.
There will be a difference in currents between the two diodes though, to get that better there may have to be an adjustment mechanism introduced into the circuit also.

[David Hess]

For a different purpose, I have used the other diode (collector-base) of small-signal transistors for a low-leakage with reasonable breakdown voltage diode in analog or protection circuits.

That is a good way to get a high voltage high current low leakage diode which will not be available otherwise.

[TimFox]

For a different purpose, I have used the other diode (collector-base) of small-signal transistors for a low-leakage with reasonable breakdown voltage diode in analog or protection circuits.

That is a good way to get a high voltage high current low leakage diode which will not be available otherwise.

For years, I had used the CB junction of low-noise audio transistors (2N2484, 2N4250, etc.) in analog systems.
When I needed a similar device for a high-impedance radio-frequency circuit (5 to 10 MHz), I had good results using the CB junction of an RF transistor (2N918, IIRC).

[Infraviolet]

MrAl, thanks. Are the relative impedances from the choice of resistors going to cause variation with things like temperature, or only variation with current? What do you mean by a temperature variable source do to the biasing of the extra diode, isn't the idea of that circuit that the extra diode itself provides a temperature varying way to bias the main diode so it will be held on the conduction threshold regardless of how the diodes both change, so long as they change together? Having both diodes of the same time from the same production batch would be enough here?

TimFox, David Hess, what are typical reverse breakdowns for the C-B junction of an NPN?  I see breakdowns quoted in datasheets for E-to-B (usually 5 or 6V), and I see C-to-E foward breakdowns given (often 45 to 75V), but not much about B-to-C breakdown. Would either of the junctions of a transistor be a particularly good choice for getting consistent voltage drops with current/temperature/manufacturing variation in that 2 diode setup? For which reasons? Thanks

[TimFox]

For a low-leakage diode, I usually shorted the BE junction of the transistor and used BC for the diode.

For an example of one that I have used, see the datasheet for the 2N2484 NPN:  https://www.mouser.com/datasheet/2/68/2n2484-31162.pdf
In that datasheet, both V_CEO (collector-emitter, base open) and V_CBO (collector-base, emitter open) are specified (as maximum ratings).
Both equal 60 V.
The CB leakage current is specified < 5 nA at 45 V, and < 10 uA at 60 V.  (60 V from a different manufacturer's datasheet)

I have seen other transistors where these voltages are different, typically V_CEO < V_CBO .
In the "CEO" case, the leakage current from the collector flows into an open-circuit base node and is multiplied by beta.

[David Hess]

TimFox, David Hess, what are typical reverse breakdowns for the C-B junction of an NPN?  I see breakdowns quoted in datasheets for E-to-B (usually 5 or 6V), and I see C-to-E foward breakdowns given (often 45 to 75V), but not much about B-to-C breakdown.

Vcbo is the breakdown voltage of the collector-base junction with the emitter open, and Vces is the breakdown voltage of the collector-base junction wit the base shorted to the emitter.  Both are usually but not always higher than the Vceo.

Would either of the junctions of a transistor be a particularly good choice for getting consistent voltage drops with current/temperature/manufacturing variation in that 2 diode setup? For which reasons? Thanks

Separate transistors will have the same variation in forward voltage drop as separate diodes, although there are dual transistors with close matching available.

[Wallace Gasiewicz]

Quote from: exe on June 09, 2023, 05:33:33 pm

I'd say passing analog signal through a diode will cause distortion. Anyway, what's the use-case, do you have schematics?

Diodes have quite some tempco (2mV/C or so), they self-heat when passing signal, and they are highly non-linear (i.e., voltage drop depends on current). So, without a feedback look it's bad to have in a signal chain.

The diodes are "biased on" with a DC voltage to act as a switch----remove the bias, & the "switch" is off.

The analog signal is normally a much lower voltage level than the bias, hence, it cannot cause the diode to cut off or saturate as this voltage differential between signal & bias ensures it appears in a substantially linear part of the diode transfer function.

Such diode switching allows a transceiver or similar device to use the same special modules (such as filters) in both the transmit & receive modes.
The ham transceiver I am currently fixing uses this technique extensively.

The old Kenwood 930 and 940 transceivers used this technique a lot and are famous for their RX and audio qualityThe diodes used are just cheap regular old diodes like 1N4148

[Infraviolet]

I could exploit that "switch" related effect then by putting the say a 2.5V bias on the diode's input, capacitor coupling the signal to here, and then having the diode's output running to ground? I can see how having the diode always strongly on, rather than the Horowitz and Hill diagram way of being just on the threshold of turning on, would mean the signal amplitude didn't cause much change in current, so no voltage drop changes here, but what about compensating for temperatrue effects, I guess this method wouldn't provide any asistance on that matter?

P.S. If I wasn't clear enough earlier, I have multiple channels feeding fast signals to peak detectors (envelope detection) to track signal amplitudes. It is too fast for any of the op amps I have to hand, so I' using the passive diode, capacitor and discharge resistor type peak detector. The crucial thing is to ensure that temperature variations or manufacturing variations between the diodes (or transistors serving s diodes) on the separate channels do not cause the amplitudes of the signals on the different channels to individually lose different amounts of voltage between the amplitude of the incoming signal and the amplitude measured on the peak detector. They all need to lose the same diode drop voltage amount, or all lose none at all though I don't think that is possible without op amp usage.

[Kim Christensen]

The crucial thing is to ensure that temperature variations or manufacturing variations between the diodes (or transistors serving s diodes) on the separate channels do not cause the amplitudes of the signals on the different channels to individually lose different amounts of voltage between the amplitude of the incoming signal and the amplitude measured on the peak detector. They all need to lose the same diode drop voltage amount, or all lose none at all though I don't think that is possible without op amp usage.

What kind of signal levels are we talking about here? A 50 millivolts offset, when measuring a 50V signal, is only a 0.1% error. But if you wanted to look at a 0.5V signal, the error would be much higher at 10%.

[Infraviolet]

I want to ensure that multiple separate signals, any of which can vary in size from 60mV (peak to peak) to 2.8V (peak to peak) all lose the same amount of voltage to within +/-20mV when each pasing across their respective diode (or transistor used as a diode).

So if we imagine two signals:
One is a sine with 120mVpp and the other a sine with 2000mVpp, I'd want each to give a peak detected level, after the diode of 120-X+/-20mV and 2000-X+/-20mV, respectively. Where X is identical for both channels, however much of a drop it is, and I can level shift with a coupling capacitor where necessary to work around this, but the channels can also have a component to that voltage drop which can vary by as much as +20mV or -20mV.

Would using a transistor array or diode array guarantee this sort of accuracy, again the point is to have all signals, whatever the amplitude of each may be at any given time, drop by the same amount of voltage (to within error margins), rather than specifically setting what amount of voltage can be dropped. As long as they all drop the same that is good.

[David Hess]

Would using a transistor array or diode array guarantee this sort of accuracy, again the point is to have all signals, whatever the amplitude of each may be at any given time, drop by the same amount of voltage (to within error margins), rather than specifically setting what amount of voltage can be dropped. As long as they all drop the same that is good.

A monolithic diode or transistor array will meet those specifications.  Matched diodes or transistors will also meet those specifications.

Note that the current through the junction will alter the forward voltage drop by about 60 millivolts per decade of current change, so increasing the current from 1 milliamp to 10 milliamps will increase the drop by 60 millivolts.  If the junction current is constant, like from a current source or a high value resistor, then this drop will not change with voltage level.

Low AC impedance across the junction can be provided with a bypass capacitor.

[Vovk_Z]

I want to ensure that multiple separate signals, any of which can vary in size from 60mV (peak to peak) to 2.8V (peak to peak) all lose the same amount of voltage to within +/-20mV when each pasing across their respective diode (or transistor used as a diode).

If that is so impotent maybe consider some active rectifier circuit?

[Infraviolet]

Wouldn't a bypass cap across the diode stop the peak detector from working?

Would placing a high value resistor before the diode be the only way to eliminate that 60mV/decade? Would it not also form a potential divider with the cap and res on the other side of the peak detector and thereby significantly reduce the magnitude of the output voltage level (though thankfully by a consistent amount)? I'll give it a go.

Vovk_z, unfortunately the signals I am peak detecting are too fast for my op amps. That would have been the easy solution, but not the one available here. There is no way I can see to get an op amp with 6V/us slew rate to serve in an active peak detector on a 3MHz incoming signal, even though this peak detected output won't even need to vary as fast as 1KHz (the magnitude of the incoming 3MHz signal does not vary between min and max at even 1KHz).

[David Hess]

Wouldn't a bypass cap across the diode stop the peak detector from working?

Yes, but in other applications a bypass capacitor is used to reduce AC impedance across the diode.

Would placing a high value resistor before the diode be the only way to eliminate that 60mV/decade? Would it not also form a potential divider with the cap and res on the other side of the peak detector and thereby significantly reduce the magnitude of the output voltage level (though thankfully by a consistent amount)? I'll give it a go.

No, increasing the source resistance adds another variable voltage drop.  If switching voltages, then the diode should be driven with a low impedance source, and a constant current drawn through the diode.

Vovk_z, unfortunately the signals I am peak detecting are too fast for my op amps. That would have been the easy solution, but not the one available here. There is no way I can see to get an op amp with 6V/us slew rate to serve in an active peak detector on a 3MHz incoming signal, even though this peak detected output won't even need to vary as fast as 1KHz (the magnitude of the incoming 3MHz signal does not vary between min and max at even 1KHz).

Linear Technology published data on the speed of various peak detection circuits and their accuracy, but I am on vacation and do not have access to those notes.  There are some suitable very fast operational amplifiers now.  There are also some circuit topologies which can take best advantage of slower operational amplifiers.

[Infraviolet]

"There are also some circuit topologies which can take best advantage of slower operational amplifiers."
I'd be interested to see some diagrams for those topologies when you're back, thank you.

What sort of circuit should be placed on the peak detector output's side of the diode to make the current constant? And by constant I assume this means constant throughout the cycle of a waveform going in to the input side of the diode, not just current draw to be constant when averaged over time at any given input amplitude?

[Kim Christensen]

Basically what you are making is a multichannel RF-probe or RF-millivoltmeter. I doubt you'll be able to be very precise without some sort of calibration routine.
This application note may be helpful.
As you can see, it's not going to be a completely linear response at different RF power levels, but it's not terrible: