Aren't those chip guys awful, when ADSL jumped the shark they pawned off the excess as cell phone headphone drivers. Hint, nothing was changed.
LM4562 was released just before National changed their product naming convention. My understanding is the rest of the family was due to get new names, so they created LME49720 to fit with the new convention, but did not drop the LM4562 named part since it had customers already. LME49720 is indeed identical to LM4562. It appears your ears are not as trustworthy as you might think. It probably wasn't another LME49xxx because there are not many popular duals in that family...
Despite spec sheet similarities, I'm not so sure they're "identical".
Same silicon design? Perhaps.
But my lots of both suggest the pieces with higher end of the range quiescent current get 4562 slapped on them and sold for roughly half price. I have a batch of 4562 sourced thru Mouser that all have 14ma QC while bogey is 10ma. Another batch are all around 12ma.
Its a real issue in layouts with densely packed opamps (mixing boards, sound cards with multiple DACs and ADCs, etc.) since its a 20%-40% increase over a high value to begin with. Power supply and thermal budgets become big issues when counts/density are high.
Subjectively they sound the same to me.
All parts will have batch to batch and part to part variations even if the masks are identical. I put a test socket on my CS4398 DAC and started to measure a large batch of OPA2134 and found quite a large distribution in the 2nd and 3rd harmonics (tested with 6 kHz sine generator in REW). All pass the specifications however some parts are 10 dB better than other parts. Currents, offset voltages, gains, etc will all have a distribution. Parts are not identical.
Think of it as a bit like the part to part and batch to batch variations in transistor current gain, Vbe, etc.
Think of it as a bit like the part to part and batch to batch variations in transistor current gain, Vbe, etc.
Interesting observation that reminds me of cartridge pickup systems for record players, where an individual measurement diagram is often attached (at least, if it is a top class version like Audio Technica OC-9).
I observe, that cartridges from Ortofon and Benz often exhibited these undesirable scatter, that various user's noted on listening tests (please note 15-20 years ago).
Do you think, that such undesirable scatter are only on OP-Amps which were produced in very large numbers and also in different manufacturing facilities (like OPA2134, NE5532 and so on) ?
Or is that also to be expected even with types like AD825, AD797, AD844/846 and LT-series from Linear Technology ?
An other reason could be, that a circuit topology with more than one voltage gain stage in the NFB loop especially tend to scatter (in opposite to versions with only one voltage gain stage realized mostly as folded cascode for the input stage).
Maybe the complexity in general and some special features of the internal topology also has a major impact on the extent of the unwanted spread from device to device - maybe a reason for several companies using discrete solutions like mentioned under
Commercial available discrete OP-AMPs - Overview wanted
As I note on the begin of this thread, variation on H2 and H3 are not critical on listening tests.
What about such undesirable scatter in relation to the very audible high order distortion ?
Thank you for further information.
Last edited:
The datasheets inform that semiconductors are variable devices, whether they are discreet devices or in IC. Many parts do not state min/max parameters. Consistency is down to testing for which there is an associated cost since a percentage of parts must be rejected and additional time for the test.
The part to part and batch to batch variations in integrated circuits come from quite a large number of different sources. All processes from all companies should be expected to exhibit significant variation. However the variation should not be expected to be the same for different process technologies, different quality/capability fabrication equipment, etc.
One example variation source would be variations in the epitaxial processes that grow the semiconductor layers on the substrate. (Or in some processes the ion implantation and activation processes, or in yet other different processes the doping/diffusion processes.)
Another example (depending on the technology used) would be variations in the values of the resistors fabricated due to variations in line width. The resistors can also vary in value due to variation in the deposited thickness. (Such as in processes that use deposited layers of NiCr, as just one example material that is used.) Other processes use resistors formed in the semiconductor layers.
There can also be variations causes by variations in etch processes.
There are quite a large number of variation sources to consider.
Also the capacitors on the chip will vary. For example a Metal-Insulator-Metal (MIM) capacitor will have variations due to the variations in the thickness and composition of the deposited dielectric layer.
There can also be stress issues and in some processes and products the product characteristics will shift between the wafer probe and final test stages due to stress changes.
So what I am saying is that there are a large number of sources of variation and all processes (regardless of manufacturer) will have a large number of sources of variation. The processes are different between different manufacturers and also the equipment used is different. Some companies have much newer and more capable equipment. There are also differences in process engineering and operations "talent" between companies. Some will be better here and there and others will be worse.
But it should be expected that all would exhibit such variations. However the cause and magnitude might vary significantly from product to product, process to process and vendor to vendor. To make things even more complicated the same product from the same vendor from the same fabrication plant can be different when they purchase the starting wafers and/or epitaxial process from a different sub-con. They might use two or even three vendors at the same time.
In certain products the bias current is particularly critical and a company that does not trim will have much larger variations than a company that does assuming that the base resistor technology is comparable. However not all trim processes and equipment are as capable as others. (It is complicated.) Some base resistors processes have much tighter distributions than others.
On RF amplifiers for cellular phones the variation of IMD products due to manufacturing variation is a very big deal, for example.
Last edited:
If you believe they don't bin/label them according to measured spec, then by all means buy 4562 for about half the cost as per current pricing, and be happy!
Its well known that parts that fall into various places on the distributions will get binned and labeled as different part numbers. Its very common in cpus. You also see it in transistors, such as BC550 which has A, b, C suffix for where they fall in the distribution of hfe for BC550 (all). So giving the oddity that all of the 100+ 4562 I've purchased as several lots all being above bogey (and sold for about half price), I suspect this is what is going on.
On some op-amps I think I read somewhere that while the designs can be the same the different part numbers might be actually trimmed (and tested and binned out) to different trim targets and test specifications depending on the part number and application. I don't know how accurate that information is.
For example some part numbers for precision applications are said to have tighter trim targets and test specifications for offset voltage while the part number (associated with the same base design) for audio has different trim targets and test specifications.
Not saying that has anything to do with the LM4562. I thought I read that that was just an old (original National) part number and later the LME numbering scheme and marketing came along. I don't know if there is or is not any difference at all between the LM and LME part numbers. I don't remember where but I thought someone from National answered that question some time ago.
For example some part numbers for precision applications are said to have tighter trim targets and test specifications for offset voltage while the part number (associated with the same base design) for audio has different trim targets and test specifications.
Not saying that has anything to do with the LM4562. I thought I read that that was just an old (original National) part number and later the LME numbering scheme and marketing came along. I don't know if there is or is not any difference at all between the LM and LME part numbers. I don't remember where but I thought someone from National answered that question some time ago.
Last edited:
BTW, no go on 1212m and 1820m at the same time. I tried it several weeks ago and it only recognized the 1820m. It also jacked up the driver and I had to reinstall it. Fortunately no after effects...
Have my 1212m 100% recapped and retrofit with 1656, one 1820m with same, second 1820m about 70% complete. I decided to go ahead and pull the 4562 out of the old one after developing a technique to remove the opamps that is safer for the circuit board pads. Fantastic sounding pieces with the restoration and opamps. The subjective difference between 4562 and 1656 is very slight and both sound great in this application. The heat difference is pretty big though, lol.
I also need to measure the 1/4" preamp input impedance on the 1820m after the 1656. The stock 33078 used in the 1820m preamps have bipolar inputs and the 1656 has FET, so hopefully this takes the 1/4" input preamp impedance up to the megohm range so no more need for a buffer in order to plug a guitar in....will post when I have a chance to measure it.
Last edited:
There was an interesting (old) exchange on the TI forum which has been brought up before.
"The wafer process of LM4562/LME49720 is Bip process and sometimes the Bip process causes the pop-corn noise or burst noise problem. Any Bip process op-amp has the possibility of pop-corn or burst noise sometimes. It happens in wafer processing and it is very difficult to screen the devices with pop-corn noise within spec in testing process."
"In my opinion, if you really care about the "excellent performance in all audio range", I suggest you do the screening by yourself. Sorry, it is not a very good answer, but I cannot offer you a better one."
"The wafer process of LM4562/LME49720 is Bip process and sometimes the Bip process causes the pop-corn noise or burst noise problem. Any Bip process op-amp has the possibility of pop-corn or burst noise sometimes. It happens in wafer processing and it is very difficult to screen the devices with pop-corn noise within spec in testing process."
"In my opinion, if you really care about the "excellent performance in all audio range", I suggest you do the screening by yourself. Sorry, it is not a very good answer, but I cannot offer you a better one."
Would be a waste of time, and I suspect you already know it. If you and your posse believe you can hear batch to batch device variation, within the data sheet values, then there’s nothing else to say.
What is the best removal technique you have found? I last used two Hakko "K" knife tips at the same time.
What do manufacturers do with parts subject to process error? Do the parts become landfill, or do they sell them off cheap?
Most process errors result in pure garbage. There are some fabs that are highly automated with advanced equipment, recipe download, etc. There are some fabs that are still highly manual.
Process errors are quite different from process variation which leads to marginal parts.
Depending on the design, specs and process capability there might be a large percentage or a small percentage of marginal fails from process variation. In cases of hard system specifications marginal fails are scrap. As a different example consider the binning of a desk CPU or GPU. The manufacturer often keeps making more speed, TPD and core count bins to minimize scrap and maximize profits. [Just simplistically speaking for explanation/discussion.]
If the process uses precious metals the scrap is often collected in drums and sent for reclaim.
Of course there can be deviation between what is supposed to be done and what is actually done. There are stories and rumors that some factories/sub-cons have problems with scrap ending up being sold. I once bought a batch of D1047 & B817 that must have been all rejects.
Operations groups are very careful but occasionally trays of tested parts and rejects have been mixed up. Sometimes it is caught sometimes it is not. I have seen companies move to color coded trays to reduce that sort of mistake. Various procedures usually exist to safeguard outgoing quality levels.
Everything will vary depending on the particular sub type of product and each particular company. Not all companies do the same thing. Experience from one company will often not match what is done at other companies. The IC industry is extremely diverse.
Process errors are quite different from process variation which leads to marginal parts.
Depending on the design, specs and process capability there might be a large percentage or a small percentage of marginal fails from process variation. In cases of hard system specifications marginal fails are scrap. As a different example consider the binning of a desk CPU or GPU. The manufacturer often keeps making more speed, TPD and core count bins to minimize scrap and maximize profits. [Just simplistically speaking for explanation/discussion.]
If the process uses precious metals the scrap is often collected in drums and sent for reclaim.
Of course there can be deviation between what is supposed to be done and what is actually done. There are stories and rumors that some factories/sub-cons have problems with scrap ending up being sold. I once bought a batch of D1047 & B817 that must have been all rejects.
Operations groups are very careful but occasionally trays of tested parts and rejects have been mixed up. Sometimes it is caught sometimes it is not. I have seen companies move to color coded trays to reduce that sort of mistake. Various procedures usually exist to safeguard outgoing quality levels.
Everything will vary depending on the particular sub type of product and each particular company. Not all companies do the same thing. Experience from one company will often not match what is done at other companies. The IC industry is extremely diverse.
Last edited: