Op-Amp noise figures: Generally >=10dB?

What do you need for decent-looking ultrasound images SNR-wise? I'd guess...

30 to 40dB?

It seems odd to me that taking something like a high-quality CD player's audio output at, say, 90dB SNR and running it through some run-of-the-mill op-amp for, say, a headphone amplifier could easily drop the SNR to 70dB. But I suppose that most people couldn't tell the difference between 40dB SNR and

70dB SNR music anyway?

Until someone drops a hemostat on the stainless steel exam table? :-)

---Joel

A whole lot more. For good Doppler performance you need an instantaneous dynamic range in excess of 80dB or nobody would buy the machine. In a guy who has chain-smoked his whole life and ate bacon every day the (unwanted) vessel wall echoes that are filtered out can reach 40dB beyond the strongest Doppler returns from blood cells. So the front end has to stomach these signals.

Tubes, man, tubes!

Or the lithotriptor upstairs starts firing away.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

If it's a receiver that includes frequencies above 20MHz I'd give it the best noise figure possible at reasonable cost. Where reasonable would be, oh, 50 cents or so ...

Why use an opamp there?

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Hmm... note to self... do not apply for job designing ultrasound amplifiers just yet...

I believe the deal with tubes is that -- instead of just adding white noise like a well-designed op-amp circuit -- they instead produce harmonic distortion in ratios that people "like." :-)

[snip]

What? No TGC ?:-)

...Jim Thompson

--
| James E.Thompson, P.E.                           |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: "skypeanalog"  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
         America: Land of the Free, Because of the Brave

The original idea was based on the problem of taking a very wideband input and splitting it off to, e.g., HF, VHF, and UHF receivers without...

1) Using wideband transformer-based RF power splitters (3dB loss/noise figure per doubling of the number of outputs... ouch...) 2) Building a perfect di/tri/quad-plexer -- the start frequency of one receiver is the stop frequency of another one, so non-brickwall return loss from each output causes amplitude output variations on the other outputs.

Actually, #2 might still be a viable approach (if the amplitude variations can be minimized to be, say, well under 1dB) -- I was just hoping for the cheap & easy "slap down an op-amp with, say, 500 ohm input impedance and the 50 ohm coax won't even know it's there..."

Nah, it ain't rocket science. Basically very similar to what you'd find in ham radio gear except that the collector/drain load has PIN diodes connected so each stage can be adjusted by 20dB or more. This is needed because you must adjust the gain on the fly over the course of a some tens of usec. Tissue is very lossy so signals from 10cm inside are a lot weaker than at 5cm. At 3.5MHz you need about 60dB to 70dB of time-gain-control (TGC). There are also other gain change methods such as the on-chip FET bank used in the AD603. The only thing I never liked and thus never used was the Gilbert multiplier.

My first job after I got the degree was to finish up a design of such an amplifier. Then came the test, on myself. Huge disappointment. Turned out it worked very well on everybody else and that my body is quite impenetrable to ultrasound. Yet designing that stuff is my bread and butter. The gal from Marketing who was tasked with testing all this back then is now my wife :-)

Somehow they do. The Hammond organ in the living room uses the 6L6. Not a whole lot of power but with the proper register setting my wife can play a church tune and it really sound like you are sitting in a sanctuary. I believe a lot has to do with the design of the (huge) speaker in there.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

guess...

Sure, but it has to work without much loss in dynamic range. Which is why I dislike Gilbert cells.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Why not do it like in TV distribution amps? The only reason why they roll off below 30MHz is to avoid intermod from AM stations. I wish they didn't do that.

So you place a nice low noise front-end amp and follow that with three buffers. Those would individually feed whatever you have to connect. If you want to keep the design really simple use MMIC but then the whole thing will cost a couple of bucks more.

If it's emitter-follower buffers there won't even be 1dB of difference.

That'll be quite noisy.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Hi Joerg,

Ah, power handling, I guess: I should have mentioned that the inputs can see up to +5dBm, so the LNA specs boil down to something like:

Frequency: 1-500MHz Gain: >=+20dB (so that noise figure of the LNA here determines almost all of the system noise figure... 10dB is perhaps enough if I'm careful with later stages, though) P1dB: >=+25dB (based on +5dBm input and 20dB gain) Noise figure:

+5dBm is a mere itch but the compression point requires a stiff quiescent current. Whether you use MMIC or a transistor doesn't make a difference, either way it means lots of current and would have to be cooled.

Back when I was at the university we used to do such such stuff with the P8000 and similar FETs but I never used it above 150MHz. Plus there was always that battle between spending the money for another big VMOS device or a crate of beer. When the last one was obsoleted we all stood there with moist eyes and reverted to using transistors in parallel to achieve the desired compression point. Tougher to heat sink but it had to be done.

Back then we used ham radio for this. You could almost always find someone older who already dunnit. They may have been several countries away but that never mattered. Having a shortwave license endorsement was key.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Hi Joerg,

That's OK, the box already has plenty of fans, and we have a good mechanical engineer who honestly enjoys a challenge since most of the time he's doing drop-dead easy bent sheet-metal designs.

If you know of manufacturers for off-the-shelf MMICs that meet those specs, I'd love to know who they are. Otherwise, you're suggesting that building an amplifier out of discrete components to meet these specs isn't particularly difficult if I don't have to worry too much about power (or price :-) )?

Hmm... I'll have to get that HF antenna up in the near future and see if there's much still going on there (I do have the license for it). Right now I just have a 2m/440 antenna up and the local repeaters are pretty much dead and (being a small town) the few people who do use them are generally not technically oriented.

My wife and I are in the process of turning our "sun room" into a combined office (for her)/lab (for me) at home... there's a giant built-in (sort of) bookshelf left by the previous owner that's going away, to be replaced by a long, deep counter top and a shelf for holding test gear and some computers. I've enjoyed seeing photos of other people's benches from time to time around here -- seeing what good ideas they had --... perhaps you'd be willing to shoot a few of yours at some point? :-)

---Joel

1nV/2.6pA is about 400 ohms. Use a 1:3 turns ratio transformer on the input. Since you don't need to go to DC, that's easy.

The OPA847 is a bit lower noise than that, and a lot lower distortion as you get into HF. The ADA4899 should be really good for MW AM broadcast work, though the sample circuits in the OPA847 data sheet give some ideas how to take advantage of its unique properties.

Cheers, Tom

Cool, I'll go take a look.

---Joel

I don't know much about MMIC since I prefer discrete designs. There are some good ones but your required IP3 above +35dbm is tough. Check the usual suspects like Sirenza and Hittite, maybe they carry a better one than this example:

There isn't much of a market for such high IP3. Maybe John Larkin knows some. Personally I'd try a discrete design and you may need higher supply voltages than the usual 12-15V. If you go that route consider power FETs in grounded gate fashion (regardless of whether the professor said that's stupid...) and with >20V supply. I've often used switcher FETs in linear fashion, also against what da professor said. 500MHz might be a stretch for a FET though but there'd also be nothing wrong with paralleling and cooling BJTs. What you'll end up with is more of a little power amp.

Anyhow, do the mixers of the connected receivers sport such good numbers? An amp with 1dB compression around +25dBm can deal a fatal blow to a receiver input stage that isn't up to par. All it takes is someone in the area keying his mike.

It's pretty boring. A large corner bench with stacks of the usual RF gear.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Hi Joerg,

Yeah, I've taken a look at them. So far the Sirenza SBF-4089 looks like a winner with the major concern being that the noise figure isn't documented under 60MHz -- the FAE at the factory is working on getting ahold of that data at the moment.

We've got plenty of 28V running around anyway, which is certainly prime "power amplifier" territory. If I get a chance, I'll take a look at doing a discrete design and see what happens.

Supposedly they do -- at least at HF -- but I don't have access to their detailed technical specs at this point.

We have to deal with +30dBm without damage, and I imagine the receivers have a similar spec.

I'm sure other would find it far less boring that you do? :-)

---Joel

Just FYI, figure 7 shows an example of a preamp that goes a whole lot farther in compression point:

(attention, very large file)

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

The 1dB compression data looks like it's not holding your spec. You only get 20dBm or 18dBm min. I am a bit surprised by the 20dB+ spread between IP3 and 2dB. Strange. The dog leg in the S11 at 100MHz is also a bit strange, maybe it won't work for you.

They better. Else ... phssst ... poof. I've fried more receiver inputs than I cared for. Comes with the territory in pulsed ultrasound. BJTs have a tendency to "sort of" work after a hit but with degraded performance.

Well, boring to all RF guys and weird to all non-RF guys ;-)

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Hi Joerg,

Actually it's OK -- the real spec (for me) is the 5dBm input power, so with the SBF-4089's gain of 14.8dB, you get an output power of 19.8dBm -- just a skosh under the spec'd P1dB of 20.1dBm.

Yes, agreed -- I'll be quite interested to see what happens all the way down to 1MHz.

You don't give yourself enough credit, Joerg! I'm sure many would find it a work of art, in a Frank Lloyd Wright Great Workroom

sort of way! ;-)

Thanks for the QEX magazine link -- I'l read it over lunch here.

---Joel