LM4562, LME49710, LME49720 Start-up Behavior

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

LM4562, LME49710, LME49720 Start-up Behavior

Post by mediatechnology »

I became curious about a few reports on the web regarding single-supply start-up of the LM4562 and the identical LME49720 and single LME49710.

I initially suspected a common mode limitation or polarity reversal but quick tests confirmed this was not the case.
The most recent post was a guitar input using a LME49710 which quite likely had a high-value bias resistor to prevent pickup loading, some form of rail-splitter, AC-coupling and 9V operation. The schematic was not provided. Since it has a similar input I decided to plug an LME49720 into The Waveulator's front-end replacing a TL072.

I realize that the noise current of the LME49720 operating into a 1M Rbias resistor might be excessive and not the best choice for a guitar front-end but it made a good test circuit that was similar to the poster's.
This stage is not a follower: It has 6 dB of AC non-inverting gain. The DC gain however is unity.
The schematic is similar to the one here: viewtopic.php?f=6&t=606
Sure enough, just like the post elsewhere, the LME49720 latched-up on start-up.
Lowering the bias resistor value to 100K, from 1M, was the only way I found that would prevent it.

Conditions During Latch-Up

V+ is 9V, V- ground.
During latch-up:
In+ is +926 mV
In- +950 mV
Out is +827 mV.

If the circuit is latched-up, temporarily lowering the bias resistor from 1M to 100k will unlatch it.

The LM4562 and LME49710-family Have Unusually Low Input Bias Currents for a Bipolar Input Audio Op Amp.

The typical Ib is 10 nA, the maximum Ib for the LME49710 is 72nA.
These are outstanding specifications.
To put it in perspective, the NE5532 has a 1 uA typical Ib.

The datasheet for the LME49710 has no sign for the bias current; for PNP inputs the bias current should be flowing out of the IC.
The lack of a sign for the bias current specification suggests that the LM4562-family uses some type of bias current compensation.

US Patent 7649417 issued to National Semiconductor may provide clues:
http://www.waynekirkwood.com/images/pdf ... Op_Amp.pdf

With Low Bias Current Why are High-value Bias Resistors Causing Latch-Up?

Low bias current should permit higher values of bias resistors.
10 nA across 1M produces only 10 mV of error across the bias resistor.
This is an insignificant I*R drop in the bias resistor of an AC-coupled preamp.

It appears that during start-up however the bias current is significantly higher.
100K works: 1M doesn't.

What may be happening is that the initial bias current for the input pair has to be satisfied externally until the compensation circuit kicks in.
Once the initial start-up bias current requirement is satisfied, the quiescent bias current is quite small.


What Are the Limits to Values of Rbias?

I built the following test circuit and admit testing only ONE LME49710 sample.
The results however correlate closely to a different sample - an LME49720 - in the guitar input.
The results may not be indicative of a broader population of parts but the results are instructive.

Image
Test Circuit LME49710 Bias Current Start-up

In the circuit above the V+ and V- lines are switched simultaneously.
Though shown as a dual supply circuit it is identical to a single-supply circuit with ground serving as the rail-splitter.
The bypass capacitor is 100 nF.
A hard-wired follower is shown: A follower constructed with a 1K RfB||20 pF had slightly worse start-up performance. (20 pF was required for stability.)
TP2 is monitored during start-up, TP1 is for future use to measure the magnitude of Ibias during turn-on.

The Results

The first thing I found was a supply voltage sensitivity.
Lower voltages required lower values of Rbias (Rb) for proper start-up.


Rb vs +/-V vs Latch-Up:

100K +/-2.5V Latch
100K +/-3V Latch
100K +/-3.5V OK (note 7V is near a 9V battery's end-of-life)
100K +/-4.5V OK (9V single-supply)

1M +/- 4.5V Latch (9V)
1M +/-6V Latch (12V)
1M +/-15V Latch (30V)

At the minimum supply voltage of +/- 2.5V:

10K +/-2.5V OK
47K +/-2.5V OK
68K +/-2.5V Latch

Conclusions:

The LM4562, LME49710, LME49720 and LME49740 offer excellent performance as an NE5523 and NE5534 replacement.
The designer applying them should be aware however that the low bias currents offered do not necessarily permit impedance scaling to the point that start-up bias currents, which appear to be several times higher than the quiescent, are not satisfied.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: LM4562, LME49710, LME49720 Start-up Behavior

Post by JR. »

I have not paid close attention to the evolution of these uber-opamps other than to note the unusual low distortion and high output drive current. (this makes them candidates for front ends around wide dynamic range A/D, mic preamps with low value feedback Rs to keep noise down, etc).

I miss the good old days when people would tell us or publish at least rough schematics of the inside.

For the last 24 hrs I have paid a little closer attention to these three and they do appear to be ne plus ultra "bipolar" opamps. They do tricks to reduce input bias current but at the end of the day they are still bipolar input devices.

I am not sure what exactly to make of this, but the differential input impedance of these opamps are all 30k Ohm. For comparison the differential input impedance of an AD743 (bifet) is 1x10^10 ohms.

What this tells me is that in use the open loop input impedance may not be as high as we might think. This 30k impedance to the other input is generally bootstrapped up by the output and NFB resistor making the - input follow the plus, but in the extremes of operation, 30k impedance between inputs.***

It is instructive that all of the application circuits for these show relatively low impedance sources. One of the RIAA preamps shows a 470 ohm termination, perhaps used for noise analysis but clearly wrong.

=======
I concur that these look like great replacements for the old soldier 5534, but for the record the 5534 was never a strong candidate for high impedance buffers. That's why the good lord provided us with Bifet opamps.

I do not mean to trivialize this latch-up issue and they should at least mention this in the app notes (or fix it). :oops: :oops:

I can imagine how to prevent the latch up with clamps, but I can not imagine these being optimal for such high source impedances.

These need to go on the good list for high performance "bipolar" opamps. We need another list for high performance Bifets. Who knows by now they may even have some low noise Mosfet opamps. I really haven't been paying close attention. Back in the day the RCA mosfet opamps were noisy, but processes have gotten better.

JR

*** Wanye, for chuckles have you tried driving one of these (after it starts up) from a high source impedance? A simple inverting amp with 1M input and 1M feedback could have 20+ dB of noise gain. (or not) :lol: :lol:
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: LM4562, LME49710, LME49720 Start-up Behavior

Post by mediatechnology »

They do tricks to reduce input bias current but at the end of the day they are still bipolar input devices.
Exactly. And they should be used as such.
I am not sure what exactly to make of this, but the differential input impedance of these opamps are all 30k Ohm.
The NE5532/NE5534 datasheet also show a 30-300K "Rin" which one could assume is a differential input impedance.
The NE and LME-series also have differential diodes.

EDIT: This limits the high differential impedance required for peak detectors, comparators, clamps etc.
When the op amp inputs are held in null during linear operation Rin and these diodes are bootstrapped as you point out.
I tried lowering the differential impedance resistively however to see if it affected start-up but it didn't appear to help.
I concur that these look like great replacements for the old soldier 5534, but for the record the 5534 was never a strong candidate for high impedance buffers.
Exactly. The low Ibias specification of the modern bias current compensated audio op amp is great to reduce offsets developed across the bias resistance but we still have the noise current in addition to the start-up input current requirement to temper application of these parts in high-impedance applications. Best to keep the impedances low for numerous reasons.

The post that made me look at this read:
I chose the device for its high input impedance and also because it did not require offset nulling.
Then:
It sounds good but I encounter a strange behavior when the circuit is disconnected from and then reconnected to the power source (single supply, regulated 9VDC). The circuit does not turn fully on (very low output, lower than unity). It would require for the circuit ground to be disconnected from the power supply ground to "reset" the circuit and enable it to fully power up. Could this be because the op amp is designed for dual-supply and I am using it with a single supply?
My recommendation in this application - a guitar preamp - was to use a TL071...
ricardo
Posts: 262
Joined: Sun Apr 10, 2011 7:24 am

Re: LM4562, LME49710, LME49720 Start-up Behavior

Post by ricardo »

I now remember coming across this situation and the cure was just the same .. reducing the input bias resistors. Can't remember where though.

TL07x has similar behaviour but for different reasons. I see TL07x is OK in one of your applications.

Wayne, if you are right (and I think you are) then LM4562 should be OK in another application where EVIL latching occurs in TL07x .. LF filter circuits which can be overloaded. I grovel in awe :o

Actually LM4562 must be OK for da LF filters cos it has input protection diodes .. which is how I 'cured' the problem with TL07x.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: LM4562, LME49710, LME49720 Start-up Behavior

Post by mediatechnology »

Thanks Richard.

I did look for polarity inversion first but the input common mode range, when exceeded, doesn't produce inversion in the LM4562. In my recollection that was the TL072's latch-up mechanism.

The AC-coupled line amp didn't exhibit latch-up with a LME49860 despite the high value of the shared common mode resistor but I think the input caps may have been supplying start-up current. Left open that circuit could also latch. That may be one place where a 5532 (or BiFET) is a better choice. viewtopic.php?f=6&t=557
ricardo
Posts: 262
Joined: Sun Apr 10, 2011 7:24 am

Re: LM4562, LME49710, LME49720 Start-up Behavior

Post by ricardo »

Scott Wurcer has confessed that his AD797 has EVIL input bias cancellation :o

I've not heard any reports of yucky behaviour from this but I think makers who use this EVIL should indicate Ib as +/- XX nA in their datasheets.

Like JR, I moan the passing of schematics in OPA datasheets .. even if some AD ones (like AD797) speak with forked tongue. :D

I LIKE big Ib in OPAs. It allows me to reliably bias my EVIL electrolytic coupling caps :mrgreen:
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: LM4562, LME49710, LME49720 Start-up Behavior

Post by mediatechnology »

I LIKE big Ib in OPAs.
And the advantage we think we might get from low bias current bipolar OPAs - higher bias resistor values - we don't actually realize if we have to make the bias resistors low to have good start-up behavior.
I do realize that there's the bias current noise issue that prevents bipolar OPAs from having high bias current resistor values but I have to wonder if the LM4562 for example would perform just as well without Ib cancellation.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: LM4562, LME49710, LME49720 Start-up Behavior

Post by mediatechnology »

Just found this interesting post on the TI e2e forum. Nothing new here but it confirms what many of us have suspected regarding the input topology, bias current cancellation and part numbering system.
Hello Mark-san,

Thank you for taking LM4562 on your product. Yes, LM4562 is taking most common archtecture as you mentioned. The input stage is PNP diff pare with bias cancellation ckt and NPN turned arround current miller as output of input stage. This output is connected to 2nd stage( integrator) input. And then the output of 2nd stage connect to output stage (Emitter follower) with current limit.

I hope this information helps for your concern.

FYI,

LM4562 and LME49720 are the same chip and same perfomance. originaly this op-amp released as LM4562 but this device number "4562" was already exsist in competitor device name. Then we assigned new name as "LME49720" but LM4562 was already in mkt. So keep the device name as it is. So you can use LME49720 as well.

Best regards

T NAOKAWA
http://e2e.ti.com/support/amplifiers/au ... 67540.aspx
The author responding to the post is Toyojiro Naokawa.
Apparently Mr. Naokawa is (or was) a National Semiconductor employee.
User avatar
carlalex
Posts: 11
Joined: Sat Sep 19, 2015 6:12 am
Contact:

Re: LM4562, LME49710, LME49720 Start-up Behavior

Post by carlalex »

Hi,

New to this forum, and I came across this thread when looking around to understand the behavior of the lm4562 during start-up. Maybe my post should be a new thread "Designing with the LM4562" since I have some other questions unrelated to start-up, but I found this thread while googling.... (And by way of introduction, I'm an audio design amateur. My background is aerospace engineering at school and software engineering by trade. My circuit design experience is simple to moderate.)

My current project: designing some loud speakers with integrated amplifiers for home use. I am using off-the-shelf hypeX amps and power supplies, and an entrodev power supply for the auxiliary circuit, which is what I a designing from scratch.

The auxiliary circuit I am currently designing does a number of things:
- It's an active cross-over and balanced input stage built around lm4562 opamps. The design is heavily influenced by "the design of active crossovers" by Douglas Self.
- It controls power on signal detect, and power down after delay. The signal detection is based on the njm2072, with a 555 governing my timer.
- The power management is currently designed to control the HypeX amps, but I would like to extend it to power down the cross over and all of those opamps that would be doing nothing but sucking power when everything else is in stand-by. (An idle lm4562 on +/- 15V rails runs warm and draws about 10mA.) That way during stand-by I could get by with only powering the logic circuits on the +5V rail, and a single op-amp on the +/-15V rails that is buffering for the signal detector. It is pursuing this goal that I started considering op-amp start-up behavior and came across this thread.

So, my immediate questions:
1) regarding the start-up behavior described in this thread, should I expect this behavior if I am using low impedance stages? All of my op-amp stages have been designed with impedance between 600 and 2k ohm to keep noise down. Ultimately, the question I am trying to answer for myself is whether I can get away with letting signal hit my opamps before they are powered up, or do I really need to build in timers and relays into my signal path to manage this? As these are likely to be plugged in an left to the sensor to control power on/off 99% of the time, I don't want to burn my house down with something that latches, burns out an IC, and lets chaos ensue one out of every 50 start-ups. The simple thing for me to do is just cut the rails to the part of the circuit I want to disable. If I leave these portions connected to signal, it would be expected that signal would hit the audio path before and during power-up while my logic circuits are deciding to turn on the power on for a few milliseconds, and it would be expected in this case that input would be present while the audio path op-amps are suffering through whatever transients occur when the rails are reconnected. If I need to withhold signal for a few seconds, I have more work to do.

2) are there other wacky start-up (or shut down) behaviors I need to consider? (I'm an amateur, nothing is too obvious to state.)

3) I have other questions, but don't want to overload this post :). I'm still wrapping my head around oscillation. (what does it look like on a scope? will it be a small amplitude at a high freq? Obvious and rail to rail? at what freq?)

4) Also un-related, but something I saw that I really don't understand.... I've got my control circuit, RF and DC blocking stage, and a simple balanced buffer to feed the control circuit all bread-boarded up. I was looking at this circuit on a scope and I noticed that when the input was 0v (signal generator set to 0v), I would get a small (about 4mVpp) and very fast (90 MHz) signal. Schematic and picture of my oscilloscope attached. I have no explanation for this. Is this a glimpse into the op-amp internals? Noise from my scope bouncing around? Is it something I need worry about? I think I ruled out the control circuit; I see this even when I remove that signal path. Stranger still, I see this when I completely power down both my signal generator, and the circuit under test! (This buffer stage is not in the final audio path, so it's just thrown together with some higher resistor values I had on hand, and I'm currently experimenting with various gains to tweak the njm2072's sensitivity.)

Thanks in advanced for any help!

--- Carl
Attachments
AsBuilt-MysteryHighFreq.png
AsBuilt-MysteryHighFreq.png (51.22 KiB) Viewed 31419 times
Scope-highfreq.png
Scope-highfreq.png (1.7 MiB) Viewed 31419 times
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: LM4562, LME49710, LME49720 Start-up Behavior

Post by mediatechnology »

Thank you for joining us!

The low impedances in your circuit shouldn't cause any problems with LM4562 start-up. I see 6K81 resistors around the op amp.
With the "normal" values of resistors you typically see with bipolar op amps I've never seen one latch.

I also don't think, with the current limiting of the 6K81s you'll have any issues with signal appearing before power though you definitely want to test for that.
One trick you could (should?) use is to put signal diodes from the op amp input pins to steer signal to the rails if the rail voltage is exceeded.
(For an example of that look at figure 14 on the THAT1200 datasheet. http://www.thatcorp.com/datashts/THAT_1 ... df#page=10)

The unused op amp in your drawing would benefit from a 1K resistor in series with the non-inverting to ground to limit current during startup.
The LM4562 aka the LME49720, the 5532/34 and others in both series all have back-to-back diodes to limit Vdiff to +/-0.6V.
The 1K provides current limiting for the input clamp diodes.

BTW when you buy your LM4562s check the LME49720 pricing.
They're the exact same part with different numbers and sometimes different pricing.

I think that 96.7 MHz signal you see is an FM broadcast carrier.
You are quite likely in the transmitter's near-field.
Post Reply