Thanks for the suggestion. But my board is already a daughter card on a larger board. The customer has an IP network unit with 6U eurocards as the main boards. He has eight daughter cards on each main board as the I/O interface. So I am stuck with the size, power supplies and a lot of other things including the laws of physics. Like some of the others here, I believe that he is over specing the interface and will need *either* 10 Vpp or to drive 50 ohm loads, but very unlikely both together. At this point I am somewhat concerned about the power issue on the board overall. Some of the posts here got me to take a harder look at that. So I am leaning toward using a single 12 volt rail. But that makes it even harder to meet the 10 Vpp requirement into a 50 ohm load.
I will be talking to him in a few days and we will be discussing the tradeoffs. I just need to know what I can and can't do with available chips and circuits.
Thanks for taking a serious interest in this. I'm not sure I need thermal shutdown in the Fairchild device. It current limits around
120 mA and my circuit is AC coupled. So assuming a max amplitude sine wave the rms current would be 85 mA and the amp would dissipate 0.44 watts. Assuming the SOIC8 package, at 155 C/W the rise would be 70 C. The part is rated for 150C junction so that puts the max environment at 80C well above the 70C requirement.
I have also noticed on some parts that you can't rely on the internal thermal shutdown to fully protect the device. They say that extended temps above the rated 150 (or sometimes 125) will cause damage to the package and change the device characteristics.
The more I think about it, the more I like the single 12 volt power approach. I think that is what I am going to recommend to my customer.
Anyone have experience with Fairchild support? So far the email support seems to take a couple of days and the answers seem rather canned.
On a sunny day (Tue, 22 Jan 2008 08:29:40 -0800 (PST)) it happened rickman wrote in :
My personal experience with 'audio' is, that they drive the cable hard. I once did a 200 Ohm (think it was) audio distribution amp, and had
200 Ohm resistors. They did not terminate the other end, so the levels were 2x too high. So better have no resistors, and then that requires a short circuit proof amp, and some way to dissipate the heat. Maybe you can protect things by foldback current limit in the supply.
Sure, but you will always have a complete 3D envelope available, so that does not preclude the use of piggyback or riser boards on your daughterboard if you can do it. This is precisely one of those times when you should get creative with your 3D envelope design. Remember, pigyback and riser cards can even be 0.8mm or 0.5mm boards if you need to shave an extra 1mm here or there.
Make sure you get a full 3D envelope spec from him too, not just a basic maximum height.
Most amps need only relatively small series R to stabilise against heavy C loads. It's easy to test on the bench, so I'd grab some of the likely opamps and breadboard them up, and test them with nasty loads.
Since the fairchild Opamp has no thermal sense, and you are tightly area/power constrained, I'd look at adding a temperature trip
- devices like AnalogDevices TMP35 are cheap, and small, and could sense the output area, and so reduce the PCB area for ressitors / Opamp and copper cooling
I don't know why you say it is *just* these parts. The point is that I have two channels with differential outputs, so four outputs. Right now I am using three 20 pin TSSOPs with a *bunch* of passives. I don't get why you say there are *only* four resistors... That is pretty far off the track! The current parts are duals, so I get both the A and B output of one channel in each package and a third for the two input amps used as differential to single ended converters. There are filters, feedback resistors, decoupling caps and the list goes on. Then the output resistors are the ones that have to be *HUGE* to handle the power. Do the math... 0.1 Apeak into 50 ohms is 0.25 watts. If the output is shorted, it becomes 0.2 Apeak and 1 watt. A
1 watt resistor is a 2512 size X four outputs equals a lot of board space that you can't put vias into. The opamp has to dissipate 0.35 watt normally, with a shorted output it rises to 0.68 watts. That's a
30 C rise with the optimal PCB layout which includes a large copper area on the bottom where you can't put other parts or vias or run traces... I can show you the gerbers of my half completed layout. After trying to find ways to shove 10 pounds of parts and vias onto a
5 pound board, I realized that it just would not fit and now I am looking for new ideas to make it fit.
I have opamps that will drive the required current *and* the voltage. But the one I picked just will not fit the space available with the current design using those huge resistors (actually four 0805s instead of one 2512). Trying to use four opamps to do what I can do with one opamp is not really a productive step.
But thanks anyway. Going over all this helps to keep it fresh in my mind. When I encounter a difficult design, I tend to want to find a way to make it work. Then I drive myself until I am totally sick of dealing with the problem and often have started to work way too late and am loosing sleep over it. That is part of why I came here. If nothing else, it gives me a chance to think about the problem and not just obsess with trying to make a bad solution fit.
Right now I am pretty happy with two possible solutions. There is a Fairchild part that should either meet the requirements of 10 Vpp into
50 ohms with none or a small series resistor (maybe 5 ohms) for stability. Since the part is not really spec'd in this range, I can't be sure it will drive the full voltage, but the data given sure makes it look good. The part is cheap, the single 12 volt supply saves a lot of power burnt in the opamp. But it has no spice model, so I can't check it before building it. I will have to verify everything on the lab bench where it is much tougher to measure things like currents.
The other solution is an audio amp from TI which *does* have a spice model. It is also cheap and available. But it uses the +- 12 volt supplies and will be pretty warm. It has differential outputs, but the best thermal package is 10 x 10 mm which is large for this socket and the thetaJA is still a wopping 80C/W, so I am not sure about the effect of the power. It is also a fixed gain device, so would require driver opamps to set the gain of the output (that may be a minor issue since the two input amps can share a quad package with the drivers). Finally, this won't drive 10 Vpp, but rather tops out at about 8.5 Vpp.
Yes, the board is really small, 4.5" x 0.85" and most of it is FPGA, connectors and some RS-422 driver/receivers, analog switches for isolation, ESD diodes, two power converters an oh don't forget the codec. Worse, I am constrained by the existing design to put the connectors, one near an end of the board and one a little past the middle. All of the analog stuff should go between the two connectors and the digital on the long overhang.
So there is the current solution that meets all of the requirements other than fitting the board space. Then I have an audio amp that is pretty hot and won't drive the full voltage into 50 ohms, but looks fairly good in other regards. Lastly, there is a opamp from Fairchild that looks like it would do it all and from a 12 volt supply as long as I give up the 50 ohm output resistors, but it has no spice model to simulate.
I do appreciate the comments and just conversation. I helps me get it all straight in my head. I have always known that if someone can't properly explain something, it is because they don't really understand it. So if I can't explain it to you, I don't understand the issues here. By trying to explain it, I learn it that much more clearly.
This is another post that has made me think a bit. I was calculating power in the opamp considering that the current is a sine wave and the RMS value should be used rather than the peak. But when you short the output the current for the most part becomes a square wave of + and -
120 mA. This effectively is like driving a DC current at half the power supply voltage and is significantly larger than the RMS case.
6 volts x 120 mA - 5 ohm x 120 mA**2 = 0.65 watts! The thetaJA of these parts is not so good at 155 C/W. That might be a good reason to go with two of the quads, rather than three of the duals. Still, at
140 C/W that is a 90 degree rise above ambient! Normal operation is only 0.17 watts and is only a 25 degree temperature rise. None of these parts have a thermal pad, too bad.
Then seriously consider vertical riser boards if there is room. A long thin board only say 10mm in height can hold a surprising amount of parts. Likewise, low height profile SMD connectros or whatever can gain you an extra daughterboard to almost double your PCB area.
If you are really out of space for the vias then perhaps consider blind/burried vias as well.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.