[jteef]

This is a very high speed current feedback op amp made for dsl applications. I read an article on www.sound.au.com about it and wanted to try it out.

Specs on it are as follows:

ADSL Differential Line Driver
400 mA Minimum Output Current Into 25- Load
High Speed
140 MHz Bandwidth (-3dB) With 25- Load
315 MHz Bandwidth (-3dB) With 100- Load
1300 V/us Slew Rate, G = 5
Low Distortion
-72 dB 3rd Order Harmonic Distortion at f = 1 MHz, 25- Load, and 20 VPP

I plan to use pretty much the same circuit as Mr. Elliott did in the article and will probably try it with and without the OPA134 buffers. I will probably also reduce the gain to a total of about 11dB to reduce the chance of oscillation.

My initial fear in this project was all of the RF considerations and my inability to get a board that worked. I was on the TI page and clicked the link to the evaluation board and was surprised when it was only 50 dollars (my previous experience with eval boards has been $300+) This is somewhat expensive for a DIY project, but a working board is so critical with high speed circuits that it is probably worth it to me.

This is going to be for a pair of Sony MDR 7506's for now (overkill eh?) (does anybody actually know the impedance of these?)

Current draw is probably too high to get any useful life out of batteries, but that is OK.

Any comments or ideas?

jt

 

[Tomo]

Hey,

We discussed about this guy a while ago.

THS6012 is a dual current feedback amplifier. Ppl told me this type of opamps makes alot of IMD. He told me to look for a voltage feedback amplifiers.

I found THS4022 and THS4032. These are low noise high slew rate voltage feedback amplifiers. I am currently building an amp using THS4022.

Unfortunately, all these guys are available in specialized surface mount cases. These are impossible to solder. So you have to build boards for amps using these THS series opamps. You can get DIP cases for these too. So you should use those instead.

RF will not cause much problem as long as you follow the intensive instructions on the specs and the application notes available at TI. (Though Z-Network may not be necessary. I dunno I have not tried yet

)

Tomo

 

[Joobu]

SOIC is actually extremely easy to solder if one uses surfboards. I prefer the surfboards without the SIP pins, then a copper ground plane can be used underneath. Best thing is they're relatively cheap and easy to obtain (digikey).

The chance for oscillation increases as the overall gain goes down, not the other way around. Of course, narrowing the bandwidth prevents oscillation as well.

Other items for high speed circuitry include using a ground plane and bypassing with surface mount capacitors.

An excellent reference on the subject is Linear Technology (www.linear.com) app note 47. Big read but worth it.

 

[Tomo]

joobu-

No no, This particular opamp has very different footprint. There is a spot at the bottom of the chip you need to solder to sink heat. It is possible to solder all pins. But it is impossible to access the soldering pad at the bottom of the chip.

Tomo

P.S. THS series does not have standard SOIC footprint. If you look at the specs(!), you should see that there is a soldering pad at the bottom of the chip. This is there for sinking heat.

 

[ringtheorist]

Soldering SOIC is actually easier (to me) than soldering DIP components. It's not necessarily faster, but here's why:

http://www.seattlerobotics.org/encod...6/oven_art.htm

You can actually get a variant of that stuff at RadioShack. You may try that before ordering the 'real' stuff.

b

 

[jteef]

Tomo: have you read the article?

The IC is designed for ADSL where there are many carriers close in frequency, so IM distortion needs to be very low. I do not think you could use this IC in its intended application if this were not the case. This is good news for our purpose.

The author of the article cites several reasons why current feedback would be superior to the voltage feedback op amps on the market right now. But as this is my first experience with a CFB op amp, I cannot confirm or deny any of them.

I am going to use the evaluation board for my first try. It will save me the headaches of tracking down oscillations and will probably end up being cheaper.

I am fully aware of the heatsinking requirements on the bottom of the chip. My first inclination is to drill a hole(1/8" maybe?) in the PCB and solder it from the bottom. If there is sufficient heat, the solder will be sucked right up in there and make a good bond. I dont know if this will have any adverse side effects though, and may look for another option.

Joobu: are you sure about that? It seems counterintuitive at first glance. Also the author states that a lower gain in the THS6012 device will lead to better stability. It has been a while since I have actually had to deal with the stability criterion for a circuit, and I do not remember the details.


jt

 

[topher]

jteef,

I think you have a good point. ADSL normal use is
much like the way people put in a worst case
signal to make testing for IM easy.

This also means that the AD815 maybe a better
choice. It has pins for through holes like a
normal DIP and a tab to connect a heat sink.

 

[jteef]

ringtheorist: that is indeed very interesting stuff, and looks like it would work great for the heatsink under the chip. Might have to do the last stage of heating a few extra seconds though.

I have begun reading that Linear Technologies app note. 130+ pages! Looks like a good read though.

jt

 

[jteef]

I have considered the AD815, but my main concern about this project was getting a good board layout that minimizes the chance for instability. TI sells an eval kit for 50 dollars. The AD board is 90. If I decide to build more and have some boards made, the AD815 looks like a good canidate.

jt

 

[Joobu]

THS series comes in both D and DGN package. The DGN package is the power pad type with the heat sink.

The principle of negative feedback is to decrease the overall gain. Hence by applying more feeback, the overall gain decreases. Therefore overall gain is inversely related to feedback. The greater the feedback, the greater the possibility for instability.

A real life example of this is decompensated op-amps. They are generally faster but require that the gain be a minimum of 5 or 10.

 

[topher]

jteef,

Your right. If you haven't prototyped a lot of HF
circuitry, the eval card is a great way to start.
Just a little expensive.

 

[jteef]

Joobu: I have done some calculations (hah, like they need to be done) and come to the conclusion that with modern devices with high open loop gain, amount of feedback is largely inconsequential to the requirements of the stability criterion. I.E. Most devices will operate with very nearly 90 degrees of phase margin. Double what is deemed acceptable by the author of my book.

So assuming Rod Elliott did not lie, what do you think he meant by this statement?

"I configured the amp with a gain of 12 (21.7 dB), and the output level was more than sufficient to cause hearing damage (if you like that sort of thing). Ideally though, the THS6012 should be operated at a lower gain, as this helps to reduce the chance of oscillation. "

I've got sony MDR 7506's and i think the impedance of them is 68 ohms with 106dB/mW. What is a normal output voltage range for a high end CD player or even a portable CD player. talking 0-100mV or upwards of a volt?

thanks

jt

 

[Tomo]

Hello,

I believe our concepts of feedback is rather flawed. I know of some Japanese DIYers who builds tube amp openly use high feedback circuits. They usually refer to them as "----(tube type)HF designs." Though this might only be for tube circuits, non-Feedback is not necessarily the best. I hear some saying that non-feedback designs can have "loose" bass (if not well-designed)**.

Also, I wonder what is it mean to be stable? In terms of oscillation, CFB can oscillate if the feedback resistor has very low resistance. This does not happen with VFB. So be careful when you build G=1/G=0dB circuit with CFB. You cannot just shunt the output to "-" input for CFB. Also, you cannot use a capacitor in the feedback. At high freq, capacitors become simple short circuit. So your circuit highly likely oscillate. This does not happen with VFB.

Note oscillation with high bandwidth circuit can be a serious problem. These circuit can produce RF noise enough to disrupt radio communications and TV communications. This is somewhat similar to microwave oven. In microwave, high bandwidth element is put in resonence to oscillate providing massive power to cook food. Similarly, high bandwidth circuit such as the one with THS6012 may oscillate and cause serious disruptions although you ain't gonna be able to cook food. That is why FCC regulation notice is on the specs for THS6012. You can check for serious oscillation by checking heat output of THS6012. If it is heating badly, it's oscillating. This can happen for VFB or CFB.

Note the difference is behaviors of VFB and CFB. These just indicates that you cannot interchangeably use VFB and CFB haphazardly. These do not necessarily indicate which one is better. In some circuit, one may be more suited than the other but in another circuit, the other is more suited. You must learn about the difference in characteristics and exploit them.

Tomo

** Please note that when I talk about designs of some amps, I assume that they are well-designed. If poorly designed, even most expensive parts in the world would sound crappy.

 

[Joobu]

Ok, what I said applied to voltage feedback. At lower overall gains, where feeback is at its highest a VFB has greater chance to oscillate since the phase margin is smallest. With the addition of capacitive loads and parasitics, VFB op-amps can be tipped into instability.

Current feeback is different. The main determining factor for bandwidth is the feedback resistor, not the closed loop gain. I completely forgot about this fact. My bad

There is usually a recommended value in the data sheet for the feedback resistor. Therefore by raising the feeback resistor slightly, stability is increased.

 

[ppl]

Ok my thoughts on CFbB Opamps. While most opamps of this type have high IMD due to a Non Symectric Slew rate. A select Few of the Newer ones use more than one stage rather than just the Current mirrors that older Designs use and thus have somewhat symmectric Slew rates at a specified output Voltage. None the Less as tomo states you can't just sub a CFB opamp for a VFB type. Also the Bandwidth and Phase margen are determond by the Value of the resistor from the Output to the Inverting input. As was stated low values result in high bandwidth and reduced Phase margin. High values of Rfb result in reduced bandwidth Because of the reduced current on the feedback node. this is the controling factor in a CFB opamp. The most importent drawback of CFB type of opamps is the Nonlinarity at low signal levels resulting in High IMD regardless of the Topology used. as the Signal is reduced the available Current at the Inv input is also reduced and this results in a Changing Transferfuntion with the Musical signal. Most perponents of CFB designs eather are un aware of this or don't think it is importent. I think it is Very importent. Analog devices claim to have a New Topology using Current on demand to reduce this Bad quality but evean thay dont Claim to Eliminate it. in Contrast a VFB opamp has a somewhat Monotonic transfer- funtion and sutch are more Linear over a wide rande of signal levels. it should be noted that A DSL Application operates the Device at a somewhat constant signal level as compaired to the Wide Changes of an Audio Signal. Last but not least the High Input Bias currents of CFB opamps do not allow hanging a Vol Pot off it's inputs with out large Changes in offset voltage requiring a DC servo. Just my 2C worth.