Who makes the best highly linear RF amplfier that meets this spec?

Ha, sounds like you need to rip the vertical amp board out of an old Tektronix acope.

Interesting thought, but probably not work well, bcause the accuracy of voltage on the screen is only 3% which makes for a lot of potential distortion.

You might look into CAT amplifiers.

Damn spellcheck. CATV amplifiers. Very linear. Some even DC coupled.

tm

Have you checked Minicircuits?

Cheers

Good idea. Some of the hybrid cable amps are awfully good, and cheap. 75 ohms, but that's easily fixed.

Or maybe a fast opamp.

What are you trying to amplify ?

An antenna signal without pre-filtering in the 0.5 to 100 MHz band ? Depending on location, a long wire connected directly to the amplifier input might cause high distortion values.

Some multitone signal (COFDM as in DVB-T/T2/C2) ? Be careful about the peak to average ratio.

A single sine wave somewhere within that band might be doable.

That's about -147 dBm/Hz output noise and assuming +5 dB gain, input noise density -152 dBm/Hz. As a reference, a resistor at room temperature generates -174 dBm/Hz, so the noise requirements should be easily obtainable.

Such high even order attenuation would suggest a push pull arrangement, possibly with some feedback.

Since you specify harmonic distortion, are really trying to amplify a single sine wave ?

With multiple simultaneous tones, even two audio tones in a SSB signal, the most interesting specification is the intermodulation products, especially the third order distortion, e.g. 2 x f1 - f2. The amplifier capability is specified by the (input or output) intercept point IP3 (or actually IIP3 or OIP3 resp.). From these figures, you can calculate, how much the amplitude of the input tones must be derated to get a specific intermodulation specification.

Yes. I tend to restrict R.F. to > 1 GHz nowadays....

OUTPUT RANGE

Yes, these are, very hard specs to meet.

For reference, a fundamental problem in high speed A/D systems is simply getting a unity gain buffer of low distortion at high speed, e.g. for driving the track and hold. Realistically, you need 50 GHz Ft, bipolar transistors to get -90dBc figures at > 100MHz

An example, the AD9642 14 bit, 250 MSPS ADC is showing typical -95dBc, max -82dBc, at 90MHz.

With that in mind, there are parts like the TI LMH6554 2.8GHz Diff Amp, designed as a an ADC driver, -96dBc @ 75MHz @2Vpp. Doesn't have the output level you want, but illustrates the problems involved. The TI LM6552 has larger output voltage swing, but as not as good distortion.

So, your work is cut out for you. It would certainly be interesting if you found a part to meet your spec :-)

What problem are you actually trying to solve? I would suggest rephrasing it to make it more manageable.

Kevin Aylward B.Sc.

The IP3 point is a theoretical power level at which the 3rd order intermodulation is as strong as the tone(s). By reducing the tone levels by 3 dB, will drop the 3rd order distortion by 9 dB, thus a signal to intermodulation of 6 dB.

By reducing the tones 40 dB below the IP3 point, the 3rd order products fall by 120 dB, thus signal/intermodulation is 80 dB, close to what you wanted.

Since you wanted +16 dBm output, the output IP3 must be +14 dbm +40 dB or +54 dBm (250 W). Usually the amplifier max output power needs to be

You might be able the achieve similar performance using some feedback techniques.

Check for instance "A High-Dynamic-Range MF/HF Receiver Front End" QST February 1993 and corrections in June p 73. It contains a push-pull RF amplifier with "noisless" feedback which should be able to handle your specifications with only 2 x 25 ma at 12 V. However, the two decade frequency range can be a bit much for the transmission line transformers.

LOL! Good call, true, never thought of CATV suppliers. I especially like being able to tap into volume production to keep costs down.

Thank you for the calculations on intercept points, etc. Do you have a URL for this paper? The transformer shouldn't be a problem, I routinely work with tranformers which have 5 decades of operation, but that's more in the 5 Hz to 2MHz ranges.

You raise a good point, how about three amplifiers in a row, each with gain of 2? That should be easy to keep linear, right?

A gain of 2? and put three in a row?

I proabably can live with 1Vpk so TI part may be ok. Thanks for the specific part numbers, will check them out. TI used to have an excellent website to use, bounce around, search, parametric search, models everything, but alas, the SW people updated everything so now TI's website is molasses AND often doesn't even change screens, just sits there, and worse, bogs down PC so can't even do anything else while waiting for the screen to not even change.

Trying to get 18-20 bit ADC 200Msps running WITHOUT spending $20k each. Seems everybody thinks in terms of radio receivers and doesn't much care above 14-16 bits at high speed rates.

Thanks, forgot all about AR. Used their stuff for years and years. Sometime, not taking up space in public forum, have to tell you of incident at RF/EMI Test lab during susceptibility testing of Telecom Product at 30V/m using AR amplifier! Very dramatic failure! Not Ar's fault, clearly mentioned in their manual, but test house didn't read manual.

I meant what is the product that requires such a spec?

Not gonna happen without new technology being invented.

It not that no one cares, it's essentially, impossible to achieve in the realistic world.

Kevin Aylward B.Sc.

Den mandag den 5. maj 2014 17.06.32 UTC+2 skrev Robert Macy:

though I think finding a unicorn is probably easier than the getting 20 bits at 200 Msps

-Lasse

Thanks for the LMH6554 mention, the amp is designed balanced and our system is balanced, not single ended coax but twisted pair, so ideal.

High speed ADC is available if you send in your first born. Agilent makes PCI Express format 2 channel system of ??bits to 4Gbs at $20k each and $80k supporting SW costs. I want to put this into 20-100 per month production, so a bit pricey. So far, I've got the custom design down to around $100 each 18 bits at 200MSps and NO SW system development cost. BUT, require this 'unobtainium' amplifier Should be able to do three amps in a row and get the performance, right? GAIN=8 and meet all the other specs. Gain of two, someone mentioned CATV amplifiers reuqire low distortion, sounds like a good match.

Thanks or the URL will try it. ...at least got the pdf for the search ! AND several are listed on the page! we're talking over 100 dBc down distortion AND 2 channels for $3.80!!! thanks!

It worked better than after google's search for spice model for the LMH6554 Go to TI website and the page sits there with the hour glass on for unlimited time. Clicking on 'view all documents' or anything else does NOTHING! and TI's website used to work so well.

Use the article tittle, "N6NWP" or "noiseless feedback" as search term and you should get much hits, even if the original article might not be on the net.

Really ?

Designing a tube audio amplifier output transformer was a bit problematic, when at 20 Hz there was always the lack of inductance (at least in a reasonable sized transformer) and at 20 kHz the stray capacitances became a problem.

You really should check what transmission line transformers can do and what ferrite materials to use to get even 3 decades 0.1 MHz to 100 MHz range.

In that article MRF586 transistors were used with fT at 3 GHz (for a MF/HF low VHF amplifier). Since you only needed 100 MHz, the hfe would be 30 (about 30 dB), so even after feedback, it would be easy to obtain 5 or 10 dB _closed_ loop gain in a single stage.

Sure, you could use three stages in a row if the frequency response and gain accuracy is really that critical so that the feedback network accuracy really defines the performance (not the transistor temperature, current etc. parameters).

However, putting "hot" (fT > 1 GHz) devices close proximity and you are asking for troubles due to non-intended feedback paths and the mess is going to oscillate, unless the sections are really well shielded from each other.

According to the old wisdom, amplifiers oscillate, but circuits intended to oscillate do not :-)