This is puzzling. I would expect the x891 to be as least as good as the 4562 in low- and balanced-Z applications, and way better above 10kHz:
For comparison, the OPA1622:
This performance is reached because of the dedicated GND pin to which the internal compensation is referenced to, thus a bit unfair.
No. TI discontinued the LME49710 a few years back, brought it back for a while, only to then kill it.
It baffles me that nobody checks the information available directly from TI. Here's what I see:
Both devices have active status at TI. Active means in current production.
Devices don't go from active to EOL overnight. When a device is obsoleted, a product change notification goes out to large customers and distributors. That notice says, "this device will be discontinued on XXX date." At this point the status changes from active to NRND (Not Recommended for New Designs). NRND means that TI will continue to make the device, but that it is on the chopping block for discontinuance. If you use the device in your products, you should find another device to stuff in that socket or place a lifetime supply order now. Once the device is discontinued, it will be listed as EOL (End Of Life). You might still be able to buy it from TI, but they won't be making more of them. Once they run out, you're stuck dealing with distributors who deal in obsolete parts such as Rochester. And when they run out, there is no more.
Tom
Yes- you won’t get EOL on an opamp without NRFNDs. Discretes however an entirely different matter 😉
Yes. That is unfair and means it’s not a standard pin out! I remember Scott Wurcer talking about some of this stuff.
Very good PSRR BTW
😊
It's pretty remarkable the level of information that flows when a product change is made. I've received NRND notices from Mouser because a manufacturer changed or discontinued a package option. The manufacturer no longer offers the capacitor in bulk. You now have to buy them on cut tape. Oh no! Not that! The horror!!
Tom
Tom
Yup.. good call. I've tried them and they are great. But... no DIP packages and the 1612 (my favorite) is also now pricey as well :-( I'm still looking.OPA1612 and OPA1656 are worth trying
OPA1602 is a cost-effective BJT allrounder with a well-balanced set of specs.
I did look at the TI pages, as you would've seen in the 1st post of this thread if you'd read it before posting.
There are various reasons why I am spooked:
i) All variants of the LM4562 / LME49720 / LME49860 disappeared over night from TI stock
ii) Ti are suggesting a replacement IC that has over 5x the slew rate of the above for all 3 ICs (maybe they are witholding stock, forcing people to try the 2891?). The 2891 appears to be primarily designed for video and my own experiments so far have not equalled the 4562 / 49720's performance. I do not think it's a suitable replacement and suspect quite a few people will struggle to tame the 105V slew rate (I recall the OPA627's speed causing concern with some DIYers). My tests with the 2891 showed a perfect 20K square and absolutely flat FR (to 1MHz as a VF), yet overall THD+N was not as good as the 4562.
iii) The price for DIP variants of 4562 / 49720 appears to have increased exponentially (see Bonsai's point re: wafer fab)
iv) Ti hasn't been honest with us ever, regarding the popcorn / burst noise problem with the 4562 / 49720 / 49860 ICs. Numerous people have tried to get admissions from them, yet they were met with silence (I personally know designers who've raised the issue with them).
Yes - I am a masochist by carrying on with the 4562 and its variants. Despite having to vet them for 1/f issues, it's given an overall balance of strengths that other ICs do not seem to offer for the price. Point (iv) above is why I posted here, in order to see if anyone knows anything. I don't recall seeing much warning before TI chopped the NE5534 and had to scramble to order them before they disappeared. Communication from the firm has been what I would describe as 'opaque', hence my question here.
Believe me, I am puzzled! Were there a sign of instability (the 2891 will not suit everyone's PCB layout / decoupling etc.), I'd think I'd done something wrong... I don't think TI are lying on the datasheet. My fear is that the 2891 is simply not as versatile as the ICs it's being suggested to replace and will only meet its paper spec in specific topologies. Were I to find out that my layout of decoupling arrangement is at fault, I'd be quite happy TBH. Vetting the 4562 variants for popcorn is time that I will never get back...
edit - when I tested the 2891, not only did I see a beautiful 20K square, but I left the test jig naked on the bench for a few hours to settle, with the live FFT tracking it. No signs of Radio Moscow. Noise performance was good. Distortion, not so good...
another edit - the datasheet for the 4562 suggests using it as an MM RIAA preamp... That's hilarious, as anyone with half a clue can tell you. It be noisier than a TL072 in this application and possibly unstable. If TI can make an error as glaring as this...
Let's round this up. It is still available and Mouser even has 6000 in stock. Digikey has 1063 in stock. It has quirks and need to be handpicked for specific applications and it is more expensive (same price as a OPA828) and it is made by an unreliable producer but that is all not an issue when looking at its unique qualities not found in any other opamp. 7000 ICs ready to be handpicked!
So just order them in numbers and things are good. There is no problem. You are at the beginning of a LM4562 scarce era and you can be the future dealer distributing that audio jewelry 🙂
So just order them in numbers and things are good. There is no problem. You are at the beginning of a LM4562 scarce era and you can be the future dealer distributing that audio jewelry 🙂
Last edited:
Which caused you to panic. I can understand that if you're using the LM4562/LME49720 in a commercial design at significant volume. But the fact that it disappeared overnight without notice should tell you that it's probably a website or database glitch more than anything else.
Do you have any evidence to support the "withholding stock" part or are you just speculating?
TI makes money by selling ICs. Why would they withhold stock? That makes no sense. If they wanted to do any sort of demand steering they could just fiddle with the prices. And they do fiddle with the prices. Last year I bought a reel of LM4562MAX/NOPB for $0.50/each. Now the same part is $2.556/each at the same quantity. Interpret as you wish.
I was surprised to see the 2891 as a suggested replacement too. As you point out, it doesn't seem to be an "equal or better performance" type part.
Is this a surprise? Do you know anyone who uses DIP parts in production at significant volume? I don't. PTH technology was last century. We're doing surface mount now.
As for Bonsai's comment: Opamps are small devices so there tends to be thousands of them per wafer. And semiconductor manufacturers don't just make one wafer. They make a wafer lot, which is typically around 20 wafers or as many as will fit in the wafer boat. So a single lot of opamps may yield a few years' worth of inventory. So, yeah. Manufacturers do forward build the die stock. It's simply a function of die size.
If the DIP prices are skyrocketing but the SOIC prices aren't, it could be due to the cost of the lead frame. As I mentioned above, PTH technology is going away and has been for a while now, so I would imagine that the production volume of DIP lead frames is dwindling. Manufacturers are probably EOL'ing those. So the remaining manufacturers charge more, which drives up the cost for TI, so they charge more. I don't know this for a fact, but that argument certainly passes the sniff test.
Oh, and a reality check: According to TI's website, the price of the LM4562NA (8-pin DIP) is $2.939 (@ QTY 1k) versus the LM4562MAX (8-pin SOIC, large reel) that's $2.556/each (QTY 1k). That doesn't seem overly out of whack to me.
Yeah. That part is disappointing. I do know that some manufacturers have gotten concessions from TI, but you have to be pretty high up on the food chain to get that.
Tom
We built die stock ahead for 2 yrs worth of sales on some parts because the fab was forced to close (remember the BF862? That part was made on a 4” wafer on an ancient finely tuned process with lots of other RF parts that had the same problem). Sometimes you also build ahead because a process is being moved to another fab. The main point here is it’s not always easy to keep things nice and smooth - this is a hugely complicated business with many moving parts. Big customers are the ones that have to be taken care of first since they drive the revenues. I remember trying explain some of this stuff about 10 yrs ago to Charlie Hansen (RIP), the guy that founded Ayre Acoustics. They were a big name in high-end Audio, but in semiconductors they were a nobody because there were customers doing $50 or $60 million a year in some of the businesses our division was involved in.
If you are doing any kind of volume on PDIP parts, you will run the risk of obsolescence unfortunately - better to move ahead of the curve to SMD rather than be forced to move.
(I may be open to correction but IIRC, the big distributors will provide product status updates but you have to sign up for it. I get change notification quite regularly even though I’m a minuscule customer)
If you are doing any kind of volume on PDIP parts, you will run the risk of obsolescence unfortunately - better to move ahead of the curve to SMD rather than be forced to move.
(I may be open to correction but IIRC, the big distributors will provide product status updates but you have to sign up for it. I get change notification quite regularly even though I’m a minuscule customer)
Last edited:
Yep.
It wouldn't surprise me if single opamps are falling out of favour for similar reasons.
Tom
Good practice is to filter incoming signals/cables properly - then RF doesn't enter or leave uninvited via the cable and choice of opamps isn't critical. Filtering VHF out of an audio signal isn't that difficult as its 4 orders of magnitude higher in frequency. Opamps aren't particularly worse than discrete devices anyway - although any device can have unfortunate resonances that make things worse in practice, and PCB layout has a large effect too - just switching all bipolar for JFET opamps isn't enough I think.
That is known but RF may enter in various ways. Many modern designs even here lack basic input filtering as well. Some use PSUs that create pollution themselves. Using unshielded circuits/wooden casings is a habit too. Add a popular audiophile but non RF immune part to the mix (that may also exhibit popcorn noise).
IME filtering RF even in optimized situations is sometimes not so easy so also using RF resistant parts may just be a good combination of choices. Certainly to opamp rollers and the average BWI.
IME filtering RF even in optimized situations is sometimes not so easy so also using RF resistant parts may just be a good combination of choices. Certainly to opamp rollers and the average BWI.
Last edited:
Ｗhat's the syptom of RF contimination? appear as noise in the output?
When I disassemble consumer grade audio product, they might use pretty crappy op amps, but will have a lot of filters, like common and differential choke on power, and emi filter on singal, almost like building a guard or cage. very eye opening.
