No es posible, señor.
Unlike NE5532 / LM4562 / LT1124 / most existing dual opamps, the TI OPA1622 explicitly includes a GND pin, and connects its compensation capacitor to GND (for excellent technical reasons). To make OPA1622 work as a dual opamp you need 9 I/O pins. The tenth pin, Amplifier Enable/Disable, might be considered optional in mains powered gear. Battery powered devices, on the other hand, really do benefit from the Enable pin.
_
Attachments
The GND pin can be connected to the negative supply voltage and the part will still operate, albeit with a reduction in PSRR. On the bench this typically degrades the high frequency THD in headphone applications (high load current causing supply ripple) where even harmonics from the supplies starts to creep in. But the performance is still very good. During the development phase we planned for some users to attach the enable pin to VCC and GND to VEE and treat it like a regular op amp.
I bet that would dissipate under 850 milliwatts, and the junction temperature increase above ambient air would be less than 55 degrees C. If the air is 50C or less, it looks fine. Especially since the datasheet permits junction temperatures of 200C (!!).
Wow, that's very good news. Thank you John! I thought I read the documentation carefully but obviously I missed this little nugget. Would you please tell us which document (and which page#) discloses this fact? I'd like to re-read that part and maybe figure out why I didn't remember this little gift of design flexibility. Thanks!
How much heat sink?
johnc124 - So per the discussions above, how much heat sink foil would you recommend - if any (high allowable die temp as Mark noted) - on the chip for driving the average headphone? I know you tried many cans in your lab. This would be excluding any hard-to-drive cans that are 2 sigmas out on drive power. Also what mix of top and bottom (the thermal vias) foil would you recommend? Lets say this would be for up to +/-10Vdc rails.
And a related question, are the 4 small tabs on the side of the chip intended for connection to top heat sink foil (what I've been assuming) or are those primarily just intended for the single necessary electrical connection to Vee? The more I study the datasheet I'm thinking the primary way the chip is intended to be heat-sinked is via those thermal vias to the bottom foil.
I'm going to use the chip in an amp and just want to get a starting point on foil square cm's.
Thanks! 🙂
johnc124 - So per the discussions above, how much heat sink foil would you recommend - if any (high allowable die temp as Mark noted) - on the chip for driving the average headphone? I know you tried many cans in your lab. This would be excluding any hard-to-drive cans that are 2 sigmas out on drive power. Also what mix of top and bottom (the thermal vias) foil would you recommend? Lets say this would be for up to +/-10Vdc rails.
And a related question, are the 4 small tabs on the side of the chip intended for connection to top heat sink foil (what I've been assuming) or are those primarily just intended for the single necessary electrical connection to Vee? The more I study the datasheet I'm thinking the primary way the chip is intended to be heat-sinked is via those thermal vias to the bottom foil.
I'm going to use the chip in an amp and just want to get a starting point on foil square cm's.
Thanks! 🙂
Last edited:
Looking back now it's probably not as explicit as it could be. If you look at the absolute maximums table (page 4) , it states that those pins can withstand voltages up to 0.5V outside the supply range of the device. In the enable pin section I also say "The enable pin can be driven by a GPIO pin from the system controller, discrete logic gates, or can be connected directly to the V+ supply" (page 16).
If you look at the internal schematic of the pins (page 16, figure 43) you can see that there is a bunch of protection on the pins specifically for this reason. But hindsight is 20/20, I should have more clearly stated that these pins can be tied to the supplies, with a small degradation in performance.
I will say that if you tie the enable pin to V+ and GND to V-, don't expect the part to power-up pop-free. There will be a period where the V+ voltage is above the logic high threshold, but still below the minimum supply voltage of the amplifier. For that brief period the output can have some unusual behavior. This was a compromise made so that the enable pin logic levels could accommodate 0.9V logic. This is also why I show the voltage divider circuit on the enable pin in Figure 48.
Ignore the 4 little exposed tabs at the ends of the chip, this is just a small exposed portion of the leadframe paddle (what the die sits on) and really isn't meant to have any functional use. The primary thermal connection is directly through the exposed thermal pad on the bottom of the package.
I wish I could say I have done a quantitative thermal analysis with tons of awesome pictures but to-date I haven't. My suggestion would be to calculate your maximum power dissipation which will depend on your supplies and the minimum load impedance and then use one of the many online calculators to estimate required copper area. This will be a (very) pessimistic analysis though, unless you just listen to sine waves. If you do, it'll be spot-on 😉 I think you'll likely find that the copper area required is pretty minimal and the simple "dog bone" approaches shown elsewhere on this thread are more than adequate.
The good news is that the chip has thermal shutdown that kicks in before the die hits 200C. Well before! It was one of the first things I looked at when we got our silicon back, along with trying to break the part with tons of thermal cycling.
johnc124- thanks, that helps! I'll just cover the whole top thermal pad area with top foil (not just to the 4 little side leads) and use the vias and rear-foil for additional heat-sinking then. I'll run through those thermal calculations for what I would consider typical these days, something in the 100dB/V sensitivity range at 32R with the +/-10Vdc rails
Good point about music power vs. sine waves! That really does shrink down the need for heat-sinking. Also good to know that the thermal shutdown kicks in before that max die temperature.
Interesting that you tried killing it with thermal cycling. I'm glad you did. I've run into that first hand with some chips in the past, that the thermal cutoff only works for a short period (in abuse situations with near-shorts on the output, etc), then the chip dies anyway. Sounds like it will be one hardy little chip.
Good point about music power vs. sine waves! That really does shrink down the need for heat-sinking. Also good to know that the thermal shutdown kicks in before that max die temperature.
Interesting that you tried killing it with thermal cycling. I'm glad you did. I've run into that first hand with some chips in the past, that the thermal cutoff only works for a short period (in abuse situations with near-shorts on the output, etc), then the chip dies anyway. Sounds like it will be one hardy little chip.
Last edited:
You could use a mixture of both plated-through vias and component pad thru-holes to connect the topside and bottomside copper radiator plates. If your drill holes are 35 mil (0.9mm) you can insert the leads of 2 watt resistors (diameter=0.8mm) and solder on both sides. Voila, a thick metal post that solidly connects top with bottom, electrically and thermally. One sacrificial 2 watt resistor provides enough lead-length to create at least ten thermal posts. Naturally you could kick this idea up a notch and instead use a sacrificial 3 amp diode in the DO201AD package (lead diameter = 50 mils) for thicker thermal posts with greater machismo. 1N5404 is an inexpensive example.
I had lots of little torture tests for the OPA1622 when it first arrived in Tucson...
To test the thermal shutdown I left it with its output shorted to ground, +/-18V supplies, and some DC voltage on the input. The part heats up until it hits the thermal shutdown trip point, then the output goes high impedance and the die temperature falls until it crosses below the hysteresis point (about 20 degrees below the upper threshold if I remember correctly). If the output fault is not removed, the part will just oscillate between the shutdown and non-shutdown states. I've left the part in that oscillating state for very long periods without a problem.
I also wanted to see if thermal cycling the die a large number of times would create mechanical stresses that could at worst crack the die/package, or at a minimum degrade the performance. I put a low-frequency sinusoid through the amplifier into a low impedance load, and watched the die with a thermal camera to ensure that the entire die would heat-up and cool-down during 1/2 period of the waveform. Then I left it like this for over a day, taking before and after measurements of THD and IMD. I saw no change in performance in the units I tested.
Great idea! I had never thought of that. Like you say, the solid metal wire should transfer a lot more heat than just the side plating on through-holes.
Mark - any chance of talking you into making a V2.0 of your board that is a little longer with the end-fire dog bone radiator foil on the top? I was looking through the datasheet for the preci-dip socket you specified. Looks like the 12 - 16 pin are all the same width, just longer. One thought is maybe use a 12 or 14 pin socket, not for more electrical contacts but just mechanical stability for the long(er) board. My amp is going to power headphones, so it sounds like some small amount of heat sink foil is going to be needed on the chip. Or maybe the 10 pin socket is mechanically secure enough to handle some more board length on each end.
johnc124 - I have to say I'm impressed! 🙂 Wow, talk about thorough. Even an IR camera scan for die hot spots. Sounds like there are going to be no worries about the 1622 dying in service! Especially that issue about the headphone TRS plug shorting on the way in and out of the jack. That has been known to quickly kill some high current chips in headamps that don't have current limit (*cough* AD8397 *cough*) unless the user remembers to turn the volume all the way down during plug in/outs.
Last edited:
Years of trying to get along in the world have gradually taught me to ask the question "how important is this to you?"
If it's important to you, and I mean really important, you won't eat, you won't sleep, you won't let anything prevent you from making this happen. If you have to take six days off from work, you will. If you have to miss your wife's birthday dinner and your grandmother's funeral, you will. It's really important to you. Nothing can prevent you from getting it done.
If it's not THAT important to you, if you won't exert yourself at all, except maybe to ask somebody (anybody!) else to do it, so that you don't have to, then: it's not really important.
Maybe you could ask yourself: how important is this?
Let me suggest a basic reality check on headphone power requirements. The EU has established a standard for headphone sensitivity that translates to a maximum of 75 mV for 94 dB SPL. its more involved and thmeasures with a specific noise signal but it gives a set of expectations. There are headphones less sensitive, mostly in the esoteric audiophile area and a few that are more sensitive but those will probably cease to exist if they can't be sold in a major market.
This would translate to 750 mV for 114 dB (ouch!) or 35 mW into 16 Ohms (100 mA peak). With the exception of the Planar headphones almost all the headphones I have measured (a lot) are around this sensitivity. A few are less sensitive but not many. Like speakers, in headphones louder means sells faster. (I am constantly battling the marketing and sales guys desire to exceed the EU sensitivity limit.)
This would translate to 750 mV for 114 dB (ouch!) or 35 mW into 16 Ohms (100 mA peak). With the exception of the Planar headphones almost all the headphones I have measured (a lot) are around this sensitivity. A few are less sensitive but not many. Like speakers, in headphones louder means sells faster. (I am constantly battling the marketing and sales guys desire to exceed the EU sensitivity limit.)
Mark - I'll take that as "not interested in doing a version with heatsink tabs". 🙂 Not a problem, I'll whip something up for what I'm building. Probably using a 14 pin socket with 4 of the end pins cut off, since 14 pin looks like the most common next step up.
1audio - thanks for the analysis! Yeah I had a pair of AKG K550s once that were blastingly loud at about 90mV (rms). I agree that is the trend going forward, high-sensitivity low-Z cans that work well with cell phones / ipods / tablets / laptops. So yeah, not much heat-sinking needed. Mark's board might drive quite a bit just as it sits. Also goes to show that TI was probably right on the money with their design targets for the 1622.
1audio - thanks for the analysis! Yeah I had a pair of AKG K550s once that were blastingly loud at about 90mV (rms). I agree that is the trend going forward, high-sensitivity low-Z cans that work well with cell phones / ipods / tablets / laptops. So yeah, not much heat-sinking needed. Mark's board might drive quite a bit just as it sits. Also goes to show that TI was probably right on the money with their design targets for the 1622.
grounded caps on the input
johnc124 - I have another question. I'm going to use a pot-in-front design with a likely 2.2uF film coupling cap between the wiper and the 1622 input. The cap is to keep the chips's DC input bias current out of the wiper (scratchies when pot adjusting) and the user's source DC out of the headphones. 🙂 Then the typical resistor >> than the pot value from chip input to ground for the input bias current return.
The input jack is going to be wired so that it is grounded when nothing is plugged in, similar to what NwAvGuy did with the O2 headamp. That means with no source plugged in, a grounded capacitor will be hanging off each 1622 input to ground, in parallel with a resistor. I know some op-amps have stability issues with caps to ground on the input. How does the 1622 do here? If I've missed the answer in the datasheet or one of the posts so far just point me in that direction. 🙂 Thanks.
johnc124 - I have another question. I'm going to use a pot-in-front design with a likely 2.2uF film coupling cap between the wiper and the 1622 input. The cap is to keep the chips's DC input bias current out of the wiper (scratchies when pot adjusting) and the user's source DC out of the headphones. 🙂 Then the typical resistor >> than the pot value from chip input to ground for the input bias current return.
The input jack is going to be wired so that it is grounded when nothing is plugged in, similar to what NwAvGuy did with the O2 headamp. That means with no source plugged in, a grounded capacitor will be hanging off each 1622 input to ground, in parallel with a resistor. I know some op-amps have stability issues with caps to ground on the input. How does the 1622 do here? If I've missed the answer in the datasheet or one of the posts so far just point me in that direction. 🙂 Thanks.
Last edited:
I don't think the cap to ground on the input is going to cause stability issues with the OPA1622. The bigger issue I see with this arrangement is that the input bias current of the OPA1622 is fairly large (no input bias cancellation). This will produce a significant DC offset when it flows through the large resistor after the AC coupling capacitor, which will cause clicks and pops on power-up and hot-plugging. Thankfully the input offset current is extremely low (due to rather ridiculous amount of pain we went through to match the two input devices, it also shows up in the DC offset spec). Matching the DC resistance presented to each input of the amplifier will present this issue. The key here is to not add large amounts of thermal noise from the resistors.
Whoops - I missed that IB number on the datasheet. Yep that would result in somewhere around 60mV out, with a 10K pot and a 49.9K ground return resistor after the coupling cap.
So.... I think what I'm going to do is use a OPA2140, or your new OPA1688, wired up as a voltage follower between the pot and the input of the 1622. That will let the 1622 look back into zero ohms and let a FET input deal with the IR drop from the input ground return resistor. I also see the 1688 is available in WSON. Maybe I could get it on the same small board with the 1622, then I could have an assembly house stuff both at the same time. If the FET chip is on the same board as the 1622 I could even look at looping it around, like you mentioned in a past post. Do you guys have any quick methods there for checking phase margin of that 1688/1622 looped pair?
So that brings up another 1622 question. 🙂 With the LME49600, and I believe the BUF634, they have protection diodes on the input that apparently get reversed biased if the output is shorted (TRS in/out) and the input goes 3V higher, which can cause large input currents. I've always followed datasheet advice ("overvoltage protection" header) and put 200R or so between the input of the buffer and the output of the looped op-amp to solve that, unless I was really sure the op-amp involved could handle repeated output current limiting. Does the 1622 need any input current protection in the event of an output short?
Thanks! 🙂
Last edited: