I wasn't complaining about the OPA1611 , just as an example of bipolar NPN input opamps , like the OPA1622 , or AD797, NE5532, OPA1692, ...
Anyone know whether the LM4562-LME49710-LME49860 series are PNP input or NPN . It doesn't say or show in the datasheets. OPA227 has 2 diodes in series , anti-parrallel = 1,2 V pp, maybe PNP like OPA1662 ( has single anti-parrallel diodes) . Even the 2 CMOS : OPA1652 an the OPA 1656 has these diodes.
Typical small signal mosfets can take 20 V GS , so on +/- 15V supply they could have like 4-5 V above the supply , about 8 diodes in series (zeners are noisy) , and still have enough ESD & over voltage protection . Maybe an idea for TI's John124.
Small signal MOSFETs are also usually really noisy. Most input FETs on CMOS op amps are low voltage devices cascoded with high voltage ones to maximize the voltage noise performance of the op amp. The downside is that you usually need to add back-to-back diodes to protect the gate oxide of the input FETs from large differential voltages. The input stage of the OPA192 and OPA191 families (and the lower cost variants, OPA196 and OPA197) is very unique and allows for very large input differential voltages. We also used this design on the OPA189 as pointed out elsewhere.
Regarding voltages beyond the supply rails, a couple things to consider: 10mA is the continuous current rating of the ESD diodes to the power supply rails. Since an audio waveform is neither DC, nor a continuous sinusoid, I highly doubt this would damage the amplifier input. Furthermore, most line outputs have some series resistance, which will limit current into the op amp input. Likewise, I would suggest having some series resistance on a line input, for ESD and EMI protection. And if we really want to get technical here, the primary purpose of the internal ESD protection diodes of an op amp are to protect the device during shipment, handling, and assembly.
Most companies I have worked with do air gap discharge ESD testing on their RCA connectors. This makes a lot of sense, because no one wants their stereo to go up in smoke just because they rubbed their feet on the carpet and touched a line input. Protecting an op amp input from a 15kV air gap discharge requires external protection devices which would also protect the op amp input in the use case you're describing.
And yes, the OPA1656 prototype samples were pretty popular and sold out quicker than we anticipated. We're working to restock them. The full release-to-market of the device will be happening soon as well.
I bought a handful of those from Taobao. They look similar to the picture at that link but not absolutely identical - there are differences on the silkscreen. The components though are exactly the same looking.
On the top side you see two 10k resistors - these create a virtual 0V by spanning the two supply pins (4,8) of the DIL8 plug. The OPA1622's GND pin goes to the junction of these. On mine there is a helpful oversized via to which a wire can be attached which goes to the GND pin (and the junction of the two 10ks).
On the underside the two caps are the usual decouplers - each between +/- and GND. The 10k resistor next to them is doing the 'enable' function and goes between +ve (DIL pin 8) and OPA1622 pin 8.
Not sure if this has been posted anywhere yet, but the full production datasheet for the OPA1656 is now available on the product page: http://www.ti.com/product/OPA1656
Hi John
When will OPA1656 be available either as samples and or full production.
I have a Holton Audio production RIAA Preamp I would like to try these opamps in?
I am currently using the OPA1642 devices in the preamp.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Voltage noise is higher on 1656 at low frequency til 1khz. Benefit is lower dc offset and less distortion also high open loop gain also superior stability. CMRR is not a big issue. It's best use as an alternative to opa1612 when noise is not a big concern. Opa1656 has good phase response at high frequency. Composite amplifier is one of the best application with opa1656. For being itself, opa1642/opa2140 is pretty good.
About the new full production datasheet for the OPA1656 :
This is what I see :
fig 16 : PSRR max 122 dB .... fig 17 : PSRR starting at 144 dB ?
fig 5 : nasty phase wobble at 2 MHz , will cause the same HF oscillation that
Neurochrome warned us about in his HP 1 with the OPA1611 ? (fig 51 : where the
1656 drives a BUF634A like HP 1 but without the servo).
With higher temperatures comes higher noise . Why doesn't any datasheet picture a
graph of noise vs temperature ? Luckily the 1656 comes in a normal SMD package,
what if you heat the chip in that miserable 3 x 3 mm VSON package to 175 C
where it switches off like the OPA1622 ? Must be al lot noisier ?
This is what I see :
fig 16 : PSRR max 122 dB .... fig 17 : PSRR starting at 144 dB ?
fig 5 : nasty phase wobble at 2 MHz , will cause the same HF oscillation that
Neurochrome warned us about in his HP 1 with the OPA1611 ? (fig 51 : where the
1656 drives a BUF634A like HP 1 but without the servo).
With higher temperatures comes higher noise . Why doesn't any datasheet picture a
graph of noise vs temperature ? Luckily the 1656 comes in a normal SMD package,
what if you heat the chip in that miserable 3 x 3 mm VSON package to 175 C
where it switches off like the OPA1622 ? Must be al lot noisier ?
It's easily stable in composite design with proper techniques. Nothing to worry about. Magnitude is low at that range. No problem at all. I have designed opa1656+tpa6120 composite amplifier with 0.00010% thd at 1khz with 8ohm load 125mw. No issue.
It has the potential to do so. Thus, any circuit designer who uses the OPA1656 will need to exercise the due diligence and check the stability of the design. That's the same for any other opamp. No news here.
Actually, the HP-1 used an LME49600. Different chip.
Yep. That's how electrons operate.
What if you took a blowtorch to the IC and ran over it with a tank? Would it be noisier? Would you please provide a plot?
The OPA1656 is specified for an operating temperature range of -55 ºC to +125 ºC with a junction temperature below +150 ºC. Why would you possibly expect TI to provide data that shows the noise performance of the IC when it's operated well outside its operating range?
There are many reasons why noise vs temperature graphs would be excluded from the data sheet. One reason is that it's a preliminary data sheet of a pre-production part. Maybe it's work in progress... Another reason is that measuring noise at temperatures other than room temperature is actually quite difficult and time consuming. If key customers aren't asking for those data, why spend the time providing them? Time is money. Time spent on an irrelevant task is time that could have been spent on a relevant task.
While I was at TI, I spent quite a bit of time measuring phase noise of high-end PLL/VCO ICs (LMK04800 and LMX2581 for example). The mechanical vibration of the thermal chamber - and especially the puffs of CO2 used for cooling - would be enough to disturb the measurement. Our lab manager came up with a way to turn off the temperature regulation on the thermal chamber during the measurement. Smart solution, though, it did limit the test time to about a minute at -40 ºC. Much beyond that and the temperature would start drifting too much.
Testing anything below freezing can be a pain - in particular in humid environments. You get ice forming everywhere. I recall a test that ran for three weeks at -40 ºC. By the time the test was done, a block of ice about 10 cm on each side had formed on the connector panel of the thermal chamber. It was pretty spectacular.
I agree that the VSON package is miserable for the DIYer, but everybody else has no trouble using the package. The VSON package has a much better thermal connection to the outside world than an SOIC would have, which allows for a powerful chip to be packaged in a small package. You need that exposed DAP to get the thermal resistance down. That's life in the modern world. We DIYers will need to adapt (or go extinct).
Tom
Not everyone has the means to check whether it is stable.
A blowtorch , Tom (neurochrome) ?
I think on 2 occasions Johnc124 wrote here on diyaudio about designing the chips so they cut the output at 175 C and back on around 160 C . See confirmation on page 15 of the OPA 1622 , which is in my amateur opinion , way too high temperature , and the hysteresis of 15 C way too small.
OPA1622 quiescent current is about 2 times 3 mA x 30 V (2 x 15 V) = 180 mW
A 32 ohm HP , maybe around 20 mA rms 1V rms out , is estimated at least another 500mW (both L & R ) , that's 680 mW for a 3 x 3 mm chip , even with the thermal pad only soldered to some copper print surface. Try heating up an SMD resistor soldered to a large copper printed board with 500 mW . I'm sure many would do that , I wouldn't , especially with a low noise premium amp like the 1622 . I really like the OPA 1622 /INA1620 with its enable input so you don't need a print relais on its output , but the size and dissipation sucks. Look at the 8 pen DiL they offer "on the net" with the OPA1622 on ,with no big copper surface to cool ....maybe ok if you use it with a very low voltage , but not on anything higher than a 2 x 12 V supply.
The bigger SMD packages like the OPA1656 has no thermal pad but you could put some aluminum on top of it , besides the VSON pad is only 1,6 x 2,4 mm , that's awful small. Opamps like the 1656 or 1622 with their high current output should have better packages to keep them cool .
About the thermal noise :
So they give graphs or numbers for offset (fig 21) ,offset drift (fig 20) , bias current ,
CMRR , PSRR , slewrate , output current , AoL vs temperature and boast about 2,9 nV/rHz , but no noise figure at 80 C or 100 C (or even 160 C , the temperature the 1622 gets enabled again ) ? Weird. Is the noise lineair or exponential ( like input bias/offset current of Jfet input opamps) . What's the difference between Jfet / bipolar and like the 1656 , CMOS input noise at higher temperatures ? What are they hiding ?
Not everyone has the mean to do so? Not everyone has to be an engineer.Not everyone has the means to check whether it is stable.
A blowtorch , Tom (neurochrome) ?
I think on 2 occasions Johnc124 wrote here on diyaudio about designing the chips so they cut the output at 175 C and back on around 160 C . See confirmation on page 15 of the OPA 1622 , which is in my amateur opinion , way too high temperature , and the hysteresis of 15 C way too small.
OPA1622 quiescent current is about 2 times 3 mA x 30 V (2 x 15 V) = 180 mW
A 32 ohm HP , maybe around 20 mA rms 1V rms out , is estimated at least another 500mW (both L & R ) , that's 680 mW for a 3 x 3 mm chip , even with the thermal pad only soldered to some copper print surface. Try heating up an SMD resistor soldered to a large copper printed board with 500 mW . I'm sure many would do that , I wouldn't , especially with a low noise premium amp like the 1622 . I really like the OPA 1622 /INA1620 with its enable input so you don't need a print relais on its output , but the size and dissipation sucks. Look at the 8 pen DiL they offer "on the net" with the OPA1622 on ,with no big copper surface to cool ....maybe ok if you use it with a very low voltage , but not on anything higher than a 2 x 12 V supply.
The bigger SMD packages like the OPA1656 has no thermal pad but you could put some aluminum on top of it , besides the VSON pad is only 1,6 x 2,4 mm , that's awful small. Opamps like the 1656 or 1622 with their high current output should have better packages to keep them cool .
About the thermal noise :
So they give graphs or numbers for offset (fig 21) ,offset drift (fig 20) , bias current ,
CMRR , PSRR , slewrate , output current , AoL vs temperature and boast about 2,9 nV/rHz , but no noise figure at 80 C or 100 C (or even 160 C , the temperature the 1622 gets enabled again ) ? Weird. Is the noise lineair or exponential ( like input bias/offset current of Jfet input opamps) . What's the difference between Jfet / bipolar and like the 1656 , CMOS input noise at higher temperatures ? What are they hiding ?
It's rock stable on its own. And can easily achieve 80+ phase margin with composite design running 10nF load with proper compensation techniques.
If you don't know how to compensate, stick with recommended configurations. End of story. It's the same with every other chip. Opa1656 is one of the most stable one with tons of open loop gain. You just aren't going to find better ones.
Also there are documents and video tutorials on Ti website to teach you how to design circuits that can drive capacitive load and check stability.
You just won't see the same setup working with ad797. It's far difficult to work with and has much less robustness in terms of potential in composite circuits.
On the topic of thermal shutdown: there is quite a bit of "art" that goes into choosing a thermal shutdown temperature, and most of it has nothing to do with a customer's application. Of course we want to chose a thermal shutdown temperature that prevents damage to the IC. But this shutdown temperature also needs to be high enough that it doesn't impede many of the accelerated aging tests we do during the qualification process, before we have released the device. A 25C difference in oven temperature can change a 300 hour test to a 1000 hour test. Also consider that the thermal shutdown trip point is not trimmed in production, meaning it will vary from device to device and lot to lot. You have to select a thermal shutdown temperature so that the majority of the units in your production distribution have shutdown points well above temperatures which would normally be encountered in operation or in qualification.
Selecting the hysteresis amount is somewhat nebulous as well. Really, you're just picking what frequency you want the device to thermally oscillate at. In most fault conditions the device will shutdown and restart constantly before the source of the fault is discovered and removed. On some parts, the thermal hysteresis is so small, and the thermal time constants of the die are so short, that its not immediately clear if the oscillation at the amplifier output is a control loop issue or a thermal issue (yes I've seen thermal oscillations at many megahertz). So when selecting a thermal hysteresis point, it's nice to be able to differentiate between the two conditions. All of the above comments about process spreads also apply here.
Noise versus temperature graphs: I love how a plot not included in the datasheet has to indicate that we're hiding something. Well you heard it here first: our op amps obey the laws of physics and will get noisier when they get hotter.
What a plot being absent from a datasheet really means: the business didn't feel that the cost/benefit ratio was good enough to include it. Or to put it another way, that the time and money required for the engineer to take the data wouldn't be paid-back in increased product revenue or decreased customer support time. Do we always get this right? No, and quite often we go back later and add information to datasheets. They are living documents that are updated over time. The other graphs you list are part of a standard suite of automated tests that are taken for all op amps from that particular product line (the precision amplifiers group at TI in Tucson).
Selecting the hysteresis amount is somewhat nebulous as well. Really, you're just picking what frequency you want the device to thermally oscillate at. In most fault conditions the device will shutdown and restart constantly before the source of the fault is discovered and removed. On some parts, the thermal hysteresis is so small, and the thermal time constants of the die are so short, that its not immediately clear if the oscillation at the amplifier output is a control loop issue or a thermal issue (yes I've seen thermal oscillations at many megahertz). So when selecting a thermal hysteresis point, it's nice to be able to differentiate between the two conditions. All of the above comments about process spreads also apply here.
Noise versus temperature graphs: I love how a plot not included in the datasheet has to indicate that we're hiding something. Well you heard it here first: our op amps obey the laws of physics and will get noisier when they get hotter.
What a plot being absent from a datasheet really means: the business didn't feel that the cost/benefit ratio was good enough to include it. Or to put it another way, that the time and money required for the engineer to take the data wouldn't be paid-back in increased product revenue or decreased customer support time. Do we always get this right? No, and quite often we go back later and add information to datasheets. They are living documents that are updated over time. The other graphs you list are part of a standard suite of automated tests that are taken for all op amps from that particular product line (the precision amplifiers group at TI in Tucson).
This looks like an amazing IC. I'm going to go ahead and order some.
John, perhaps you can help with something about which I've always been curious
opa: opamp
16: audio
56: i believe is the actual number
opa1656 is a part that is derived from opa2156. I think 21 means precision.
It's not always systematic, but in this case:
OPA: Op Amp (or grandpa in German )
16: Audio
5: CMOS Process technology (OPA1652 is CMOS, so I figured I would put my CMOS audio developments here)
6: Next numbers up after the OPA1652/54 family, that leaves OPA1655/56/57 for single/dual/quad of the new device
In the case of precision amplifiers they use a different nomenclature, for instance the "2" indicates a dual channel op amp on OPA2156. We did chose the "56" to tie the two developments together (like OPA1611 and OPA211).
OPA: Op Amp (or grandpa in German
)16: Audio
5: CMOS Process technology (OPA1652 is CMOS, so I figured I would put my CMOS audio developments here)
6: Next numbers up after the OPA1652/54 family, that leaves OPA1655/56/57 for single/dual/quad of the new device
In the case of precision amplifiers they use a different nomenclature, for instance the "2" indicates a dual channel op amp on OPA2156. We did chose the "56" to tie the two developments together (like OPA1611 and OPA211).
Everybody has access to TINA-TI and the simulation models made available for free by TI. Similarly, they can access my Taming the LM3886 - Stability page for hints to simulating stability. They can also read Chapter 8 in Franco. Finally, they can check the transient response of the amp using an oscilloscope and a signal generator (which could be as simple as a computer sound card).
Tom
I especially enjoyed the change National made from LM (everything) to LME (audio), LMP (precision), LMX (frequency synthesizers), LMK (clocking - which somehow is different from frequency synthesizers), LMV (general purpose opamps), etc. I think it was a good move, even if the transition was a bit rough.
Tom