My original thought.
Maybe easier: Use split supplies, ground input, then use variable resistor to +. When output lifts from zero, that's the maximum sink current. ...Jim Thompson
My original thought.
Maybe easier: Use split supplies, ground input, then use variable resistor to +. When output lifts from zero, that's the maximum sink current. ...Jim Thompson
-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
Jim Thompson wrote in news: snipped-for-privacy@4ax.com:
Nice, that fits with a vague idea I had earlier too. I'll set that up later and get you some results. With two tweaks per sampling it will take longer but I like the method.
Jim Thompson wrote in news: snipped-for-privacy@4ax.com:
I think that test might be impossible. I used a single rail as before (my supply arrangements won't allow a split AND variable supply to test with.)
I grounded the voltage follower input to negative as for the first test. A DC offset appears on the output, about 3.2mV at 4V supply, rising in nonlinear slope to about 6.1 mV at 36V supply. Even if this didn't make things intractably tedious, I found that a Li-ion cell in series with as much as 47K between output and negative ground causes a few mV added to the offset on the output, and 83µA to flow through the output pin, a value that is very constant from 4V to 36V supply. It suggests that the method fails with no definable point at which output voltage rises due to loss of current sinking. Even with the offset there to confuse things, it should have been possible to see a discontinuous change in the output voltage with respect to current change around 1 mA if the datasheet is to be beleived. There isn't one, or if there is, its signal is swamped by noise from offset and whatever else is going on. I saw the same result with two IC's and didn't try the third.
I know that Google isn't all that at times, but beat this for a new low: No results found for "measuring sink current"
What to do, what to do....? Surely there IS a way? If not, how and why does anyone specify this for an op-amp? After today's debacle it will be some time before my head clears enough to solve this.
OK. Single supply. Ground Input. Load Output with resistor to V+. Lower resistor value until Output is at +1V. Record value of resistor and of V+. Do that for V+ = 5V, 10V, and 15V. That should be enough data.
(I have the basic core running to specification. All I need is to add the sink current.) ...Jim Thompson
-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
Jim Thompson wrote in news: snipped-for-privacy@4ax.com:
That won't work either. I set up for 5V supply, found resistors such that a
1K pot (all I have available) in series will find the 1V pullup on output. What happens is that at just over 200mV it latches up to 2.3V or so! There's a lot of hysteresis there, I can reduce that to about 1.5V at which point it suddenly collapses to 147mV.I remember the datasheet saying that the voltage follower needs at least 3K9 as current limit, so I tried a 4K7 to see if this fixed the problem, but it doesn't. The ONLY clearly consistent feature of both attempts as this test is to reveal some hint of nonlinearity. I doubt anything I have will nail it.
I did try a second IC with near-identical results, then set to 10V supply, and found that the total resistance needed to do it was about 2K less. The start point of the jump was a similar voltage, and the end point over 8V.
I think we're going to have to permanently do without it, or go with the datasheet nominal declaration. If what you need is a linear(ish) quality, this amp doesn't seem to have it.
That dynamic current sink is a crock... as I already opined. If the schematic is anything resembling that shown in the data sheet, it's definitely a latch (positive feedback).
Where in the data sheet is the 3K9 requirement? ...Jim Thompson
-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
Jim Thompson wrote in news: snipped-for-privacy@4ax.com:
Certainly looked like one. If it's of any use, I could maybe do the test using 100 mV instead of 1V. It may even let me do 200 mV if it latches at above that on all supply voltages. (Doubtful).
Intersil datasheet CA3140, CA3140A, September 1998, File Number 957.4 (Excerpt starts near end of page 7. Whether this is much different from yours I don't know).
"As mentioned previously, the amplifier inputs can be driven below the Terminal 4 potential, but a series current limiting resistor is recommended to limit the maximum input terminal current to less than 1mA to prevent damage to the input protection circuitry. Moreover, some current limiting resistance should be provided between the inverting input and the output when the CA3140 is used as a unity gain voltage follower. This resistance prevents the possibility of extremely large input signal transients from forcing a signal through the input protection network and directly driving the internal constant current source which could result in positive feedback via the output terminal. A 3.9k resistor is sufficient."
That's NOT output current limiting.... read carefully. ...Jim Thompson
-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
Jim Thompson wrote in news: snipped-for-privacy@4ax.com:
I realise that, though I don't fully understand it either. I thought what it was meant to avoid was the result of a fast swing of the output feeding a transient back to the inverting input, in response to one arriving on the non-inverting input. Whatever it is exactly, I didn't think the resistor would fix the latchup problem in that test because that is a result, not a cause, in this case directly from the pullup. I just tried that resistor because it was suggested, so NOT trying it seemed like a bad idea. In practise, I've often used a CA3140 as a voltage follower with a direct link feedback, no resistor, and never run into trouble. I didn't know about that
3.9K recommendation until a couple of days ago.Jim Thompson wrote in news: snipped-for-privacy@4ax.com:
Go, or no go?
Put 10K in the feedback just to make sure it isn't latching up via the input ESD structure.
Use a resistive divider at the input to establish a midpoint input.
Before any loading, verify that output is also midpoint.
Can you get your hands on a decade resistor box?
Start with it set at least 20K
Connect it between output and V+ (thru you ammeter)
Gradually reduce resistance value avoiding massive steps by backing up to a high setting before down-ranging.
Record current at which output node lifts.
This CA3140 may well be a POS, but your query has stirred my mind to improve my generalized OpAmp to include:
(1) Output dead-band (2) Output current limiting, V+ and V- values independently set (3) Supply currents (4) Swing limits (including load effects)
All with just 2 simple TANH functions... convergence guaranteed... all derivatives exist :-) ...Jim Thompson
-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
I missed mentioning (5) ROUT... independent values for sourcing and sinking :-) ...Jim Thompson
-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
Jim Thompson wrote in news: snipped-for-privacy@4ax.com:
I can't go further on this. I just spent three hours with resistors, tweezers, a magnifier, I can barely see to type now. No resistor box, just a pin deck and various resistors, and op-amps that may or may not be broken now for all I know.
What results I did get were weird. First, I got VERY few results, running ten miles is less stressful. :) Trust me, I know.
Sample 1: V mA Ltch mA Lift
5.12 0.760 10.01 1.074 15.01 1.206 20.02 1.588 25.02 1.560 30.02 1.842 35.00 2.048 1.310That's it, and it's enough for me to conclude that going further will not yield a pattern to grasp at all.
First, that latchup still occurs, even with 10K for feedback resistor. Second, when the lift starts to occur, it is sharp, at a lower current but still a latchup, just a smaller one. There is nothing neat, linear or predictable about any of this. With supply voltages above 20V, the current FELL from a slightly higher peak before latchup occured, as resistance from out to V+ was gently lowered.
If this hasn't tried your patience as much as mine, please can you take a brief look here:
Jim Thompson wrote in news: snipped-for-privacy@4ax.com:
Despite my new wariness about this amp, and the intractability of getting any clear picture of current sink limitations, I'm still pleased about that, and keen to try the model. :) It's worth having purely because that IC is so widely used, and many people might still reach for one whenever they want a very high resistance input and a single rail supply, especially as they probably have one somewhere.
I'll post my subcircuit as it stands right now and you can see if it remotely matches your measured performance.... unwrap the usual suspects...
****************************************************************** ****** PINS: 1 2 3 4 5 6 7 8 ****** .SUBCKT CA3140 NULL1 INN INP VN NULL5 OUT VP STROBE Q_Q11 N_1 N_1 NULL5 CA3046_ORG Q_Q12 N_2 N_1 NULL1 CA3046_ORG C_CIN INP INN 4pF GDC_I2 VP STROBE VALUE {(1+TANH(2.2976*(2*V(VP,STROBE)/1V-1)))*200uA/2} Q_Q17 VP STROBE N_3 CA3046_ORG R_R11 N_4 N_5 20 R_R9 N_6 N_4 50 R_R10 N_7 N_4 1K Q_Q19 N_8 N_7 N_5 CA3046_ORG Q_QD7 VP VP N_6 CA3046_ORG R_R8 N_3 N_9 1K GDC_I3 N_3 VN VALUE {(1+TANH(2.2976*(2*V(N_3,VN)/1V-1)))*2uA/2} Q_Q13 STROBE N_2 VN CA3046_ORG R_R5 NULL1 VN 500 R_R4 NULL5 VN 500 F_F1 STROBE VN VF_F1 1 VF_F1 VP N_8 0V Q_Q18 N_5 N_9 OUT CA3046_ORG G_G2 OUT VN VALUE { (TANH(3.66*V(OUT, N_9)-1.83)+19/17)*8.5mA } C_C1 STROBE N_2 12pF G_G1 N_2 N_1 VALUE { (TANH(2.2/188mV*V(INP, INN))+1)*25uA } GDC_I1 VP N_2 VALUE {(1+TANH(2.2976*(2*V(VP,N_2)/1V-1)))*50uA/2} ****************************************************************** .MODEL CA3046_ORG NPN IS=3.860200F BF=120 NF=1.04845 VAF=61.1026 IKF=-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
Jim Thompson wrote in news: snipped-for-privacy@4ax.com:
It has about -0.6V of output spiking not seen in real ones when a 5V
200 KHz square wave goes negative on the inverting input of a differential amplifier, and a small kink in the slope when the input signal rises. This kink occurs slightly earlier on the slope if the load is reduced from 1K to 10K. It's definitely slightly odd. Generally the frequency response seems about right, judging by effects on square waves at 1 MHz, but they're spikier than the real ones would be.(All this bearing in mind that LT1215 was showing very close agreement between model and built circuit in virtually identical context (only the compensation caps were absent in the CA3240 version as it doesn't need them, and this is true for the current model too, where they make no difference to the wave other than a very slight recuction in that spiking).
Please let me know if you have suggestions for that laser driver's op-amps.
What puzzled me is I couldn't get a match between DC open-loop gain and the gain-bandwidth crossing point... I get about 11MHz GBW with
100dB at DC, so I generally don't trust the data sheet... which says ~4MHz. My 11MHz probably accounts for the "spikier" transient behavior. ...Jim Thompson-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
Jim Thompson wrote in news: snipped-for-privacy@4ax.com:
The negative-going overshoot is large, maybe 100 times what it would be in a real one at 200 KHz. I deliberately slowed the input square wave slopes to
1µs (not that much flat-time left after that, at 200 KHz), and it didn't help much. In a context with very little parasitic inductance or capacitance no op-amp should have this problem.What DID help a lot was choosing the resistors in the first stage differential amp of that laser driver circuit to be higher than those the LT1215 wants for fast accuracy. Instead of 3K3 and 680R, I tried 33K and 6K8, which removed the 'undershoot' but rounded off the wave far too much, so then I tried 10K and 2K, (ratio close enough), and the result is very good. But as far as I know, the CA3140 would not be that sensitive to these changes in resistance.
What are the values of the zeners in that original schematic? Also, what are the three or so most important spice parameters to be used in the internal models for the BJT, MOS, and diode models?
Also, what other op-amp might work for that laser driver circuit I linked to? (Single rail, fast slew, dual stage, pluggable replacement for CA3240, LT1215...)
The CA3140 datasheet has conflicting statements, low sink current, yet
18mA short to V+, so I scaled to that short current. Thus a suggestion, in this line...G_G2 OUT VN VALUE {(TANH(3.66*V(OUT, N_9)-1.83)+19/17)*8.5mA} ^^^^^ Reduce this value--------------------------------------^^^^^
Until peaking matches your measurements.
...Jim Thompson
-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.