High speed amp for 20Vpp signal

The THS3491 was intended to upgrade the THS3091 in AWG outputs. It has by far the highest slew rate. 

For Audio? I had done a composite circuit for AudioXpress, here, 

7367.Very Low Distortion Composite with OPA1612 and THS6012 AX#16 Dec 24_2020.docx

Hey Michael, thanks for your interest. I found your composite doc in another thread recently and started reading; I will continue. You know the Groner/Polak composite, right?

Slew rate is not the key factor. This is a pretty unusual application, see below, with errant choice of THS4031. Bandwidth (#1!) and noise and distortion are the relevant criteria. The servos are not critical, but the one on the inverting side should probably be there if the feedback inverter has significant Vos; a typical pro audio line driver can have µV offset at its output...just not mV, unless you're living in the past ;)

One of the original design goals was to make both sides of the 1622 totally unloaded. Ideally, I'd have a buffer before the inverter in the feedback loop, but every attempt I made added group delay and thus peaking ~8MHz. The 1622 can certainly handle a 1K load in addition to driving 600Ω differentially, so I've eliminated the buffer, reattached the feedback inverter directly to the 1622, and am looking for the right part.

FWIW I can put an OPA211 or OPA828 there with no Cf, and peaking is minimal. The LM4562 does a little better, but it has its drawbacks. It's not that I care about the magnitude curve of the LPF up around 8MHz, but by providing a fast signal return back to the inverting 1622's non-inverting input, I can minimize the difference in phase between the pair of outputs down at 20kHz and 200kHz...which is, in essence, the lion's share of the original goal.

I do recognize that I can match phase of a given inverting output simply by putting its non-inverting buddy through two buffers of the same speed. I also recognize that I can match phase by using a fast amp in the inverting position and a half-as-fast amp in the non-inverting position (i.e. an OPA210 and OPA1622, or half an OPA1602 and an OPA210, respectively).

THS3091 looks right, but as with most of the high speed amps its Iq and current output and Vos is high for this application. I see why one might recommend the THS3491, but it is overkill on a few levels. Also, what's AWG?

Well Jonah, that is a pretty complicated circuit. I think of all these multiple amp composite's as adding phase shift in the loop which often causes instability, but your curve looks great. 

AWG - Arbitrary Waveform Generator (your lab signal source). Something has to be behind that output jack and for years it was the THS3091, now the THS3491. 

Almost all AWG outputs are current feedback designs (which is actually where CFA came from originally, with Dave Nelson at Hewlett Packard, the CLC102). But they have poor DC precision, hence the composite design 

The other place you get this kind of performance is in the xDSL line driver world where the THS6012 was intended as an ADSL line driver - again all CFA - those are always differential and the 6012 is dual for that, but I think there might be single channel versions in the THS3xxx series. 

Appreciate the context & history!

There is a recent headphone amp that made waves over at Audio Science Review for its -120dB SINAD, which uses a nested and paralleled TPA6120A2 in each of the stereo loops. I am forgetting the “wrapper,” probably 1611 but don’t quote me. I’d guess John Yang is on here somewhere, credit where credit is due.

When I had this hooked up with an audio amp as the inverting return, it made sense to draw it as a paralleled 2211, both with Rf/Rg at 1K5 — high enough to hopefully look more like the 2K THD+N curve than the 600 ohm THD+N curve, but low enough to avoid the amp’s input capacitance causing much of a peak. CFA or VFA, compensation in the form of a Cf here isn’t an option, so your instinct towards an xDSL driver makes sense.

Yea, John Yang sounded familiar, here was an email I had sent into the TI development group back ion Jan 2021. 

So I was poking around for info out there and stumbled across this – fairly recent and pertinent. I think this guy John Yang may actually have already done the composite that will show up in a week or two in AudioXpress, it is called the Topping A90 headphone preamp and does show the expected <-140dBc HD. The THS6012 noise model is way off, I will fix that in the next article I am thinking.

https://www.audiosciencereview.com/forum/index.php?threads/topping-a90-headphone-amplifier-review.13592/page-5

He updated it in the form of the L30 II: https://www.audiosciencereview.com/forum/index.php?threads/topping-l30-ii-review-headphone-amp.36027/

…notable for its cost. We have reached the point at which outstanding performance can be had for little ducats. But throw a balanced line input in at the front end and forget the record breaking!

Pretty interesting, you know that TPA6120 is just the same die as the THS6012 (a very old ADSL line driver). They were in such a hurry to get that released they did not even bother with a sim model. That is why I was forced to use the THS6012 model in the work I did, its noise model was pitiful, but I fixed that externally in a later article. 

Yes. I wonder about a single version. The 6120 would be handy for stereo, but I'm not seeing a justification for adding more than 20mA to a standardized SE-to-diff driver on a channel-by-channel basis, across many channels and many blocks in (say) an audio console.

Is there any characterization data for the 3120/3121 at 20Vpp, or at much lighter loads than 50Ω? Screencaps from that data sheet below.

Outside of the TI realm, I see that the AD810 is in the ballpark:

Hello Jonah,

I will ask around, but I have not yet found any data on my end characterizing the parts mentioned (THS3120 & THS3121).

If you have any other questions for me to look into, please continue asking away .

Best,

Alec

Well Alec, the original question does remain, which is essentially "What amps that have an ein of ≤3nV/rtHz can put out 20Vpp @ ≥50MHz in a gain of –1 on ±15V supplies,?" and furthermore, if anyone is looking into this in detail: "With what Rf and into what minimum load?"

Sounds like TPA6120A2 is on that list, and THS3491, but otherwise I'm just making educated guesses. 3001? 3121? 3091? I would really appreciate a shortlist, unencumbered by me missing some crucial data point (as I did with the 4031).

Thanks!