Howdy,
I'm doing some opamp switching in my Asus STX Essence soundcard. I want to try the LME49720 chip from National semiconductor. I have read that the TO-99 form-factor version (LME49720HA) is better than the DIP-8 version (LME49720NA). Could someone let me know if it is in fact true that the TO-99 is better and why? It's just that the card has a DIP-8 slot, & I'd like to know if the bother and expense of using the TO-99 is worth it.
I'm doing some opamp switching in my Asus STX Essence soundcard. I want to try the LME49720 chip from National semiconductor. I have read that the TO-99 form-factor version (LME49720HA) is better than the DIP-8 version (LME49720NA). Could someone let me know if it is in fact true that the TO-99 is better and why? It's just that the card has a DIP-8 slot, & I'd like to know if the bother and expense of using the TO-99 is worth it.
The silicon die inside any packaged version of the LME49720 will be the same regardless of package type.
That said, there are reasons why TO-99 cans are still being used. The most important reason is probably that device stress is much, much less with the TO-99 cans. Ceramic DIPs (ceramic with gold lid) probably come in second on the list of low-stress packages. The plastic packages are molded, i.e. formed by squeezing a bunch of plastic granules together around the die and lead frame. Squeezing granules into the die causes localized stress of the silicon where the granules touch the die. As doped silicon is piezoelectric, device characteristics will change very slightly as a result of this package stress. Usually, parameters like voltage offsets (Vos) of op-amps have a tighter distribution when the dies are packaged in TO-99 cans than they do in the molded plastic packages. But note that we are talking a change of maybe a few 10's of uV here... As far as other performance parameters, it's hard to say what the impact of device stress would be without analyzing the circuit of the IC in detail.
I suppose one could argue that the can will provide some shielding. While this is true, the impact of any electrostatic or electromagnetic fields imposed on the die will be minimal -- unless you're talking fields with wavelengths of a few mm (--> frequencies on the order of 100's of GHz) where the wavelength of the incident wave starts approaching the size of the largest metal traces on the die.
The bottom line is still: Will you be able to hear a difference between an LME49720 in DIP-8 vs one in TO-99 in a double-blind (ABX) test? I highly doubt it.
Will you be able to measure it? Maybe.... If you have state of the art test equipment, design a clean test setup, and measure a large sample size of parts in various package types you can probably find differences in the statistical distributions of various performance parameters between the different package types. If National Semiconductor measured a large enough difference between the various package types, they'd spec the TO-99 parts differently. I haven't checked the data sheet, but if there is a significant, marketable advantage of using the TO-99, it would be spelled out in the data sheet in form of a separate table listing the performance of parts packaged in TO-99 cans. I've seen that done on other NSC parts.
I suspect the reason the LME49720 exists in a TO-99 package is that a key customer requested it. But that's just my guess.
Hope this answers your question...
~Tom
That said, there are reasons why TO-99 cans are still being used. The most important reason is probably that device stress is much, much less with the TO-99 cans. Ceramic DIPs (ceramic with gold lid) probably come in second on the list of low-stress packages. The plastic packages are molded, i.e. formed by squeezing a bunch of plastic granules together around the die and lead frame. Squeezing granules into the die causes localized stress of the silicon where the granules touch the die. As doped silicon is piezoelectric, device characteristics will change very slightly as a result of this package stress. Usually, parameters like voltage offsets (Vos) of op-amps have a tighter distribution when the dies are packaged in TO-99 cans than they do in the molded plastic packages. But note that we are talking a change of maybe a few 10's of uV here... As far as other performance parameters, it's hard to say what the impact of device stress would be without analyzing the circuit of the IC in detail.
I suppose one could argue that the can will provide some shielding. While this is true, the impact of any electrostatic or electromagnetic fields imposed on the die will be minimal -- unless you're talking fields with wavelengths of a few mm (--> frequencies on the order of 100's of GHz) where the wavelength of the incident wave starts approaching the size of the largest metal traces on the die.
The bottom line is still: Will you be able to hear a difference between an LME49720 in DIP-8 vs one in TO-99 in a double-blind (ABX) test? I highly doubt it.
Will you be able to measure it? Maybe.... If you have state of the art test equipment, design a clean test setup, and measure a large sample size of parts in various package types you can probably find differences in the statistical distributions of various performance parameters between the different package types. If National Semiconductor measured a large enough difference between the various package types, they'd spec the TO-99 parts differently. I haven't checked the data sheet, but if there is a significant, marketable advantage of using the TO-99, it would be spelled out in the data sheet in form of a separate table listing the performance of parts packaged in TO-99 cans. I've seen that done on other NSC parts.
I suspect the reason the LME49720 exists in a TO-99 package is that a key customer requested it. But that's just my guess.
Hope this answers your question...
~Tom
Myself and others have found a SQ improvement moving from a 8 pin dip LM4562NA to the to99 LM4562HA in a headphone amplifier. The improvement may be due to better heat dissipation, but I also heard that the better performing dies were used in the to-99 cases, which maybe why they command such a premium.
Oh and it's very easy to fit a to99 into a 8 pin dip socket, just use a bit a veroboard.
Oh and it's very easy to fit a to99 into a 8 pin dip socket, just use a bit a veroboard.
Last edited:
This thread indicates, from National engineers themselves, that the TO-99 sounds better. They don't know why.
http://www.diyaudio.com/forums/solid-state/132471-national-opamp-inflation.html
Audioman54 knows these parts from the inside out.
http://www.diyaudio.com/forums/solid-state/132471-national-opamp-inflation.html
Audioman54 knows these parts from the inside out.
Thanks for the fantastic answers, this is an outstanding forum.
I think I'll stick with the DIP-8's and here is why. Being a health science guy (biostats not ee), I've been conditioned to require hard evidence when I make most decisions. As many of the functional imaging studies have shown in recent years, a person's beliefs and expectations directly effect neuronal responses in encoding of the pleasure of an experience. That is not this same thing as saying beliefs and expectations effect how much pleasure one derives, only that it effects our creation of memories of how much pleasure we derived. So I do not put as much stock in subjective recall as I do the data sheets from National which show no differences in the specs between packages.
But I do very much appreciate both points of view, and ymmv yada yada...
I think I'll stick with the DIP-8's and here is why. Being a health science guy (biostats not ee), I've been conditioned to require hard evidence when I make most decisions. As many of the functional imaging studies have shown in recent years, a person's beliefs and expectations directly effect neuronal responses in encoding of the pleasure of an experience. That is not this same thing as saying beliefs and expectations effect how much pleasure one derives, only that it effects our creation of memories of how much pleasure we derived. So I do not put as much stock in subjective recall as I do the data sheets from National which show no differences in the specs between packages.
But I do very much appreciate both points of view, and ymmv yada yada...
For me, when moving from the dip to the to99, the most obvious difference was that some recordings lost the spit, tizz and sibilance that was apparent with the dip, without the loss of detail retrieval. It was a cleaner presentation. Having said that, the basic lm4562 (which was what the LME49720 used to be called) is still an excellent device, when fed a good, clean supply.
Good basis to make your decision though, bburl, and the money saved from not buying the to99 version can go on more music, so you still win!
Good basis to make your decision though, bburl, and the money saved from not buying the to99 version can go on more music, so you still win!
Doing semiconductor design for a living, I have not noticed any significant change in circuit performance in simulations with/without parasitics for circuits operating at audio frequencies. The only noteworthy exception was a crosspoint switch I designed at some point where the layout parasitics played a *huge* role in performance. I ended up balancing every line in the layout in order to minimize disturbances caused by charge injection.
I'm not saying that parasitics will not impact circuit performance, because I know they will. All I'm saying is that when designing op-amps (not "audio grade") I haven't seen any significant difference in circuit performance.
Regarding the LM4562 vs LME497x0. I doubt it's the same die as the specs are quite different (AFAIR).
~Tom
good to know, how thick is the passivation compared to circuit features?
I'm not comfortable with the "fields are contained by a high dielectric layer" heuristic - there is relatively more displacement in the Hi K layer but is there really less outside of the hi K layer?
there'd still be bondwire/lead frame parasitic differences between the packages
I believe I recall Scott Wurcer claiming PDIP dielectric adsorption was noticeable in one AD part - worse packaged measurements than when probing the die
I'm not claiming to "explain" differences in "sound" that I'm highly skeptical of - just tossing out another idea
I'm not comfortable with the "fields are contained by a high dielectric layer" heuristic - there is relatively more displacement in the Hi K layer but is there really less outside of the hi K layer?
there'd still be bondwire/lead frame parasitic differences between the packages
I believe I recall Scott Wurcer claiming PDIP dielectric adsorption was noticeable in one AD part - worse packaged measurements than when probing the die
I'm not claiming to "explain" differences in "sound" that I'm highly skeptical of - just tossing out another idea
Last edited:
You are probably right. I'm pretty sure the SPICE model for the LM4562 at the national site used to be labelled lme49720 (which is why I thought the two were the same), but its now something lme498-something.
Inter-metal dielectric thicknesses depend on lots of things. Obviously, it varies with process technology. I'd say for a generic semiconductor process, thin metal layers are probably 100's of nm thick, whereas the dielectric is maybe 10x that.
You can surf around at MOSIS and look at the processes they offer. For the low-tech processes I believe the specs for the process cross sections are public.
Metal is generally aluminum. The dielectric, generally SiO2.
~Tom
You can surf around at MOSIS and look at the processes they offer. For the low-tech processes I believe the specs for the process cross sections are public.
Metal is generally aluminum. The dielectric, generally SiO2.
~Tom
I think LM4562 (old name) = LME49720 = LME49860 (tested for higher supply).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.