signal generator output stage?

[uncle_bob]

Hi

How many layers in your board and how thick is your copper?

Pulling heat out of the area by the slug is highly dependent on layer count and copper thickness. You could easily have a 4:1 delta. I tend to run into this sort of thing a lot. The data sheets are often not as clear as they should be on all the details. They are targeted at a specific market that "always uses at least 8 layer boards" or some such  utter nonsense. After you and the guy who did the data sheet drink a few beers, the real answer of "that's what we guessed" comes out. 

Bob

[FrankBuss]

I planned 2 layers, 1 oz. But maybe would be better anyway to use 4 layers, for some nice big power supply planes and ground plane, and then 2 oz for even better conductivity, thermal and current. I guess this could help with noise and possible ground loop problems. And then under the op amp lots of vias and copper on all 4 layers.

[uncle_bob]

Hi

I'm just suggesting you be cautious when looking at the AD thermal data. Their drawings could be showing a lot of layers. From real world boards, adding layers is a big help. Allowing for a heat sink is not a bad idea ...

Bob

[FrankBuss]

Warning, feature creep starts I was thinking when I use the ADA4870, I have +/-10V @ 1 A. This would be useful for a good power supply as well, and unlike most other power supplies, with push/pull output. With an additional ADC, a simple SMU should be possible with not too much effort (of course, not near as precise and accurate as the commercial ones, but maybe faster).

My idea is to add another 0.1 ohm series resistor to measure the current and a current limit circuit. Would this work? Is it possible to bridge the 50 ohm resistor with a switch, or is this a problem with the 1 A and high frequency (like too much noise because of the cables)? Should I use a realy?

The control voltage of the current limit circuit could be fed by the signal generator. This would make it a 1 MHz programmable current source. So two modes: constant (adjustable, with a slower, simpler DAC) max. voltage, and using the signal generator for the current, and constant (adjustable) max. current, and using the signal generator for the voltage.

[T3sl4co1l]

FYI, using the correct ratio of voltage to current feedback (it would look something like a Howland current pump, with unbalanced resistors), you can have the amp behave as a resistor already.  Downside: above the frequencies where feedback is dominant (i.e., within reach of GBW), the output impedance will go funky as it transitions to whatever the open loop output impedance is (usually an oddball, somewhat reactive value; R2R amps may be reactive and high impedance).

You can switch it between whatever characteristics you like; using separate voltage dividers to set the gain and impedance of each combination, selecting between them with analog switches, is a fine way to go.  All you have to mind is the switch capacitance, in the 10-100pF range, which will need some compensation (loading on the +in will reduce high frequency response and make the output capacitive; loading on the -in will add peaking, or more likely, oscillation outright).

Tim

[uncle_bob]

Warning, feature creep starts I was thinking when I use the ADA4870, I have +/-10V @ 1 A. This would be useful for a good power supply as well, and unlike most other power supplies, with push/pull output. With an additional ADC, a simple SMU should be possible with not too much effort (of course, not near as precise and accurate as the commercial ones, but maybe faster).

My idea is to add another 0.1 ohm series resistor to measure the current and a current limit circuit. Would this work? Is it possible to bridge the 50 ohm resistor with a switch, or is this a problem with the 1 A and high frequency (like too much noise because of the cables)? Should I use a realy?

The control voltage of the current limit circuit could be fed by the signal generator. This would make it a 1 MHz programmable current source. So two modes: constant (adjustable, with a slower, simpler DAC) max. voltage, and using the signal generator for the current, and constant (adjustable) max. current, and using the signal generator for the voltage.

Hi

If you take a look at most power supplies, they really do not do a very good job measuring current. There are a variety of reasons for this. One of the most basic is that it is harder to do than it seems. A power supply should have zero output change when you apply a load. The same thing should be true when you remove the load. No overshoot, no undershot, no odd events. The same is true when you turn it on and off. No spikes on the output.

So far nothing about current ... You need to measure it without messing up any of the things above. You would like to have the current meter read zero when there is no load Having it read correctly when in constant current mode is also handy. Having both the voltage and current displayed at once is very nice. Just getting this much done is a struggle. Making it as accurate as a $30 DVM .... not very common at all.

Now back to the topic.

A power supply (or a voltage reference) with a 50 ohm output impedance is not a really handy item. Closing a servo loop around the 50 ohms is likely to give you spikes. You have a big and unpredictable capacitive load on most power supplies. Adding "stuff" in the signal path of a signal generator is not a good idea. Keep it simple is generally the best way to do it.

Bob

[FrankBuss]

Adding "stuff" in the signal path of a signal generator is not a good idea. Keep it simple is generally the best way to do it.

Right, maybe better to create just a good signal generator. Things like a SMU could then be implemented as extra devices. But I think an additional high speed ADC would be nice (with an extra input). It would not cause problems for the DAC, and it can use the existing USB transfer and FPGA.

[uncle_bob]

Adding "stuff" in the signal path of a signal generator is not a good idea. Keep it simple is generally the best way to do it.

Right, maybe better to create just a good signal generator. Things like a SMU could then be implemented as extra devices. But I think an additional high speed ADC would be nice (with an extra input). It would not cause problems for the DAC, and it can use the existing USB transfer and FPGA.

Hi

There is a wonderful management theory called "cycles of learning". In an engineering sense what it means is that you get much further much faster by going through the entire process multiple times than by trying to optimize the process for a single shot solution. There is a bit more to it than that. I think it applies to a lot of home projects.

I'd build one and accept that you may build another and another after that. Get the basic part working and figured out. If it works fine first time great. If it needs a tweak, redo it. When you go through the next design, add features.

Bob

[FrankBuss]

Yes, good idea. The basic signal generator works already, I'll do some of the try-and-error cycles on a breadboard and test ideas.