1. Home
  2. Electronics Design
  3. Diff amps spec at 500ohms plus, why?

Diff amps spec at 500ohms plus, why?

On a design I am working right now I need to drive differential cables at up to 100MHz and with gusto, several volts swing. I'd use coax instead but the customer is king. Looking at the various diff amps du jour I consistently find specs like this:

formatting link

That example has all the performance data at an even higher load resistance of 800ohms. Considering that regular ribbon is around 100ohms diff and even grandpa's old TV cable was 300ohms, what good does that do?

I guess now I have to not just find a diff amp but also add an analog diff buffer IC (if they still exist) or roll my own transistor stuff. Which will double as a space heater where I really can't use heat. Grumble!

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg
Loading thread data ...

That's a lot of spec sheet.. I didn't read it all. Max current >100mA that should be enough current for ~100 ohms. Maybe some spec is better when driving 800 ohms. (less distortion? see figure 69.)

George H.

Reply to
George Herold

In Aoe III, Table 5.10, page 375, covers full differential amplifiers, with diff IN+OUT, listing about 45 parts. Many are intended to drive ADCs, etc., rather than low-Z cables.

Two differential outputs at +2 and -2 volts, driving a back- terminated 100-ohm differential load, will experience a load current of 4V/200 ohms = 20mA, which isn't too bad.

2 volts amplitude at 100mHz requires an wA = 1260 V/us slew rate. That's within the range of the THS4503 specs. I tabulated the Vpp-out max and the slew rates, and many parts are dramatically faster then the THS4503. I'm sorry I didn't tabulate the maximum output currents. 
--
 Thanks, 
    - Win
Reply to
Winfield Hill

Could you use a fast dual opamp? One amp inverting, one not.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics
Reply to
John Larkin

Expensive! $12 at qty 1.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics
Reply to
John Larkin

That is exactly what has me concerned. Yes, it can drive it but then the distortion goes from not that good to horrid. Possibly I can parallel two or three but that eats real estate.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Most have drivers that are in the single-digit ohms and can easily drive

100ohms. However, the distortion performance drops dramatically under such load and in many of my cases that really matters. Sometime they end the distortion plots in the datasheet below 200ohm, probably because that wouldn't look good. 
--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

I know. There is also the AD813x series for less money. This project is more like yours are where price is secondary and performance is key.

Yes, I could use two opamps but even those often have a hard time driving such loads to +/2.5V swings without losing linearity at RF too much. Many won't come as duals and that makes it tough in my case (no space).

It's weird. Most scenarios I ever came across require driving ribbon, Parlex, some other cable or back planes in the 100-200ohms range yet the mfgs don't offer much there.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Looking a the modem at my Internet connection, one option might be to use DSL or PLC drivers like this:

formatting link

Gets iffy though, the GND thermal pad would need to be connected to a VCC- plane. They are also not intended to work well down to DC but probably will be ok there.

Has anyone used DSL or PLC drivers as differential amps?

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

I've used the ADA4950-1 as a differential ADC driver, but it should be able to drive a cable. Seems well behaved.

Actually, I designed my gadget with an LT6402, but it was nasty so I had to redesign with the ADI part.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com
Reply to
John Larkin

Thanks, that one looks nice though the data for 2Vpp and 200ohms is marginal. I need 5Vpp into 100ohms. Maybe the dual version ADA450-2 could be run in parallel. Possibly that requires 10ohms or so in series with each output, meaning four more resistors.

formatting link

Figure 22 seems to contradict figure 23.

The distortion specs don't look good. High speed or RF isn't really LTC's turf, they are more a switch-mode controller company. Well, now it's all AD anyhow.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

THD increases even more when a part breaks into oscillation.

Some of the LTC stuff is good, but it's all too expensive.

ADI often gets the digital parts of mixed-signal stuff wrong. LTC seems OK about that.

TI does good work.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com
Reply to
John Larkin

Oh yeah. Sometimes oscillation goes unnoticed until a neighbor bangs on the door.

I am sloshing through DSL drivers now. This was seriously better in the olden days where National had their "fast drivers" and "damn fast drivers". No kidding, that's what they wrote in the databook for the LH0063.

I guess nobody sends stuff over cable anymore. It's now all Bluetooth and apps and virtual-whatever.

One of the problems with AD, TI and some other companies is that they have done away with real support at least for smaller companies and consultants (even though this project is for a really big client). They think that a user forum comprises "support".

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Some of the semi people have support folks that prowl the forums and really answer questions; quality varies. I think the idea is that if support is done inn public, more people get to see it all.

But it's impressive how often nobody has the answer. A recent question invoked the response "That's an old Burr-Brown part. Nobody knows anything about it." (dual 14-bit TXDAC.)

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com
Reply to
John Larkin

That has gradually become a business driver for me. Years ago larger companies could still get someone from the semi mfg who almost lived there for a week and helped them with a tricky design. Not anymore, now they are on their own.

Or they might say "You have to go to Golden Sunset Senior Center and ask for resident Joe Smith".

Just came across the THS6214 which looks more capabable than the others but +/-5V would be at the very bottom supply range for this chip:

formatting link

It's also long in the tooth. 2009 is almost ancient for telco chips.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

I designed a couple things around the THS3062, which was claimed as suitable as a VSDL line driver. What I didn't realize is that it's

*only* safe to use as a VSDL line driver. If you use it as a general-purpose opamp, it tends to fry itself. The latest version of the data sheet mentions the problem.

So be careful!

IC data sheets are not universally to be believed. Nor are their Spice models. I usually breadboard a new part.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics
Reply to
John Larkin

That one is a CFB amp. If used as an integrator or even with the slightest capacitive feedback it coan oscillate and unsolder itself. They also don't like capacitive loads but that goes for most high-speed VFB opamps as well.

I will. Sometimes I have the feeling that datasheets are written by marketing types. Why else would anyone spec a diff "driver" at 500, 800 or in some cases even 1000 ohms? It's like specifying a pickup truck's performance with a load of two six-packs in the bed.

I might go with the ADA4950 for now and place 10ohms each in series so it won't go berserk when it sees a few pF of capacitive load. Cable is never ideal. Its drive capabilities aren't much to write home about but this is more a proof-of-concept. Later we'll likely try a DSL driver, depending on high much THD and IM performance we end up needing. Sometimes you have to run a really high voltage supply rail and divide the output back down or clamp it.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Why not 50 ohms in series with each output, to source terminate?

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics
Reply to
John Larkin

Then I don't get enough swing. The amplitude specs (for any decent distortion performance) are just as paltry as the allowed load. Also, I need +/-2.5V on each leg for some of the lines and with 5V a source termination just isn't possible. If I have to use 24V DSL drivers I can supply them with +/12V and then I'd use source termination. Have to, in order to protect gear downstream that is on +/-5V if someone forgets the end termination.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

In that case, you'll probably need a couple of THS-series opamps.

If you need a good receiver for the other end, take a look at AD8130. Really nice chip.

--

John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com
Reply to
John Larkin

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.