INA 163 low noise inputs

Last week I mentioned on a thread that I would like to try out new low-noise ICs, and TI were kind enough to send me some samples of their INA163 as being the latest and best?. (I notice that it was actually introduced in November 2000)
I did say that I would report back:??..

Tested strictly in accordance with the spec sheet, it?s certainly very quiet; the ?self noise? was so low that it was almost within the tolerance of my test equipment; measured against my 200 ohm precision resistor, it was certainly within a dB of the physical limit (I was measuring 129dB below input at 50dB gain, 20Hx ? 20KHz RMS.) [My rig could be up to 0.5dB optimistic].

Reconfiguring it as a balanced mixing stage, it did not work at all well; I suspect that the ?block diagram? is very much that? it really doesn?t work trying to use the internal op-amps in a different configuration? so that?s that idea out the window.

Rigging it up as a mic amp, as a comparison with some of my own experimental mic amps, I found that it is horribly sensitive to cable capacitances, and there?s a very good reason that they publish a ?stabilisation network? for the input ?when used with low source impedances?; I only wanted to listen to a ribbon mic! :shock:
Once it was stable, it sounds fine and not significantly different from my ?standard? amp, which is a transformer input balanced class-A design; which, incidentally measures 2dB worse on noise, but sounds about the same.

I had a look at the distortion products on the analyser, which was difficult, because at 50dB gain there?s really not much to see. I couldn?t find any 2nd order, and only the merest hint of 3rd.

While it was an interesting exercise, I?m still looking out for op-amp types that work optimally (as far as noise goes particularly) as current amplifiers.

Ted Fletcher
www.tfpro.com

Thanks for that report. At which gains did you had stability problems?

Samuel

Ted, I thought you designed the VC3? Doesn't that use an instrumentation IC in the mic preamp section?

Hi Samuel,
The inputs seem to be unhappy at any gain (I tried it from 20dB to 55dB) if the source impedance is less than say 500 ohms with a significant capacitative element, ie; with a mic cable attached! padding them out with a 47 ohm resistor in each leg works, but a series inductor in each leg is more elegant.... and that's what TI suggest.

Yes, I did design the VC3 and the earlier ones had SSM2017 chips. We only stopped using them when they got scarce. I changed the design to a balanced MC33078 arrangement which, although technically a little more noisy, was a nicer and more stable input.
Ted

Ted,

Thank you very much for the feedback - i'll copy, paste and forward it on.

As for the SSM2017 problems - INA217 could have stepped into the void.

BTW - for a balanced mixing stage, would the OPA1632 be suitable? You know more than I do at this point, so any feedback you have would be great.

many thanks,

R

Yes, I did design the VC3

Click to expand...

I was quite impressed with that design. The SSM2017 was certainly not what I expected to see when I opened it up.

If I recall, you seem to be fond of the TL08x in EQ sections...

Sorry for the unrelated posts.

The stability problems Ted encountered remind me of the difficulty of using the AD797 at low gains. The front end has very fast and low r sub bb devices (a look at the chip shows how much of the area is devoted to them!) but unfortunately they require some serious damping to prevent oscillation.

My aim was a switched attenuator system that would deliver true 24 bit dynamic range, i.e. ~144dB, and I nearly got there, but it was frustrating to throw away some of the potential performance at the outset. I wound up with some R and some ferrite beads as well. The difficulties were compounded by the use of a couple of AD811's in ~parallel within the loop, but even alone the 797 (as they warn you) was problematic.

Not particularly fond of TLO82/84 they just happen to be useful as low gain chips, and as good as any other. (and not expensive!)
The great attribute of the old VC3 was that it was virtually indestructable; we had one returned once that had been run over by a bus!

Ted Fletcher

the difficulty of using the AD797 at low gains

Click to expand...

That's interesting--I've never noticed problems with them, but then again, haven't had a huge amount of experience with the 797. I've seen an entire Trident console that was rechipped using 797's, including the EQ section, the only thing was that the HF boost/cut would oscillate when cranked all the way up.
ye

Not particularly fond of TLO82/84 they just happen to be useful as low gain chips

Click to expand...

I see.

[quote author="TedF"]Not particularly fond of TLO82/84 they just happen to be useful as low gain chips, and as good as any other. (and not expensive!)
The great attribute of the old VC3 was that it was virtually indestructable; we had one returned once that had been run over by a bus!

Ted Fletcher[/quote]

FWIW, my initial & long lasting reaction when I see TL08* in gear is always to heat up the soldering iron to replace them at least with TL07* but I realize that the theoretical noise-improvement will often not be noticable since they'll be used only at less harmfull places.

But is there anything going for an '8* when a '7* might perhaps be as non-costly ?

Interestingly, for audio purposes they are interchangeable.... the 7* is sold as 'lower noise' but actually within the audio band they are identical.
The different chip is the 6* and that has a different architecture. It's actually capable of better output drive than the 7*.... but they are all suspect at high gain and into unknown loads.

[quote author="featherpillow"]

the difficulty of using the AD797 at low gains

Click to expand...

That's interesting--I've never noticed problems with them, but then again, haven't had a huge amount of experience with the 797. I've seen an entire Trident console that was rechipped using 797's, including the EQ section, the only thing was that the HF boost/cut would oscillate when cranked all the way up.
[/quote]

AD tells you to beware at unity gain. I was using them at a gain of +2, and as I say also had the heavy lifting moved off-chip onto a pair of 811's. I was then driving an array of mercury-wetted reed relay switched Vishay bulk metal foils at quite low Z, and then this was followed by a fancy hybrid buffer. The whole shooting match is descibed in an AES preprint from the NY '97 convention papers, which I have some hardcopies of---if anyone wants one PM me with a snailmail address.

The system, alas, was never completed as the client ran out of money---the performance described was based on prototype hand-wired boards. Its application was automatic level matching, using predetermined values of required gain/attenuation, of loudspeaker drive. The gain/atten was of course done at line level before the power amplifiers. This was to be part of the big speaker-mover Harm*n subjective evaluation facility. I am afraid to imquire as to what they wound up using :sad:

> measures 2dB worse on noise, but sounds about the same.

You probably realize this; but in most rooms, listening tests won't show the self-noise level. Room noise usually overwhelms mike/amp noise. The old trick is to throw some wool socks or acousti-foam in a pressure-cooker, set the mike inside, seal the cable, and listen. Mom's old cast-metal pressure-cooker with a new sealing ring and a good cable-seal through the old safety-valve hole will attenuate room noise 20-30dB, which will put many quiet rooms down below the self-noise of good mikes and amps.

I know a few musicians I'd like to pressure-cook, but they don't fit.

> is there anything going for an '8* when a '7* might perhaps be as non-costly?

You may know folks who know. My suspicion is that the current TL07x and TL08x are the same chip, just like the Pentium 100 and Pentium 90 were the same chip.

At least with the TL0xx, at one time they were sorted according to the slight differences in specs, but I suspect current production meets the best of both specs. As processes refine, that becomes too easy to bother with two types. They put numbers on the lids after they see how many orders for each they have in hand.

(The P-90 WAS a P-100, special-order to mega-OEMs so they could advertise a Low Low Price machine without gutting sales of the flagship P-100 and P-120. If you ever again meet a P-90 machine, jumper it for 100Mhz for a blazing 11% increase in CPU speed. It will work fine. But the "P-90" machines also shipped with the slowest drives and RAM, minimal video: lowest price has its costs.)

[quote author="PRR"]>
I know a few musicians I'd like to pressure-cook, but they don't fit. [/quote]

:grin:

> is there anything going for an '8* when a '7* might perhaps be as non-costly?

They put numbers on the lids after they see how many orders for each they have in hand.

Click to expand...

Nice & practical, makes sense.

[quote author="TedF"]The different chip is the 6* and that has a different architecture. It's actually capable of better output drive than the 7*....[/quote]
I always thought the '6* was a low-power version of the '7* or '8* and having lesser performance. But I can't compare so the latter might not be true, since the '7* quotes 0.003% THD typ and the '6*... doesn't even dare to mention it :grin: (at least in the datasheet I have).

but they are all suspect at high gain and into unknown loads.

Click to expand...

I wasn't that happy to see a TL072 DI-section (DC-coupled even) in an expensive bass-amp...

Bye,

Peter

I mentioned a war story in a post a while back, but it bears repeating: beware of variable input capacitance effects and the resulting distortion with high-Z sources and large common-mode swings. It renders the stated distortion numbers for BiFET op amps essentially worthless.

Hi PRR, I think you were talking to me... about mic amp noise. Yes, I always used to use a 'blanket box' back in the 60s and 70s. It was necessary to check the character of the self-noise of a mic amp in broadcasting mixers, and using a dummy load is never good enough.
We used to bury a dynamic mic in the middle of blankets, seal the box and put it in the cupboard under the stairs!
I remember one test we did with a prototype mic amp using a descrete FET input stage and a 1:10 transformer with no resistive components in the input stage at all. The overload margin was not good, but the self noise was as good as nil! I reckon you could hear the molecules hitting the mic diaphragm.
Another trick or course is to short circuit pins 2 and 3 on the XLR input; that will give another indication of self-noise with low impedance sources.

Self-noise can be a tricky one; it's only recently that 1/f noise ceased to be such a big problem; many earlier ICs and descrete amps had problems with low frequency 'bubbling' noises.
Getting back to the topic, the noise characteristic of the sample TI chip that I tested was extremely smooth.

Ted, have you tried the new THAT preamp chips? The 1512 in particular looks very promising.

The front end has very fast and low r sub bb devices (a look at the chip shows how much of the area is devoted to them!) but unfortunately they require some serious damping to prevent oscillation.

Click to expand...

I like the 990/ADI 121 approach with the inductor-bypassed emitter resistors to get rid of that problem, as it kicks up o/l gain in the audio band as well. No way for an IC, though...

I've seen the statement that low source impedances cause "the input stage to oscillate" - implying parasitic stuff in the input pair. That's BS, isn't it? It's simply the transconductance going up and we get nyquist instability, right?

Reconfiguring it as a balanced mixing stage, it did not work at all well; I suspect that the ?block diagram? is very much that? it really doesn?t work trying to use the internal op-amps in a different configuration.

Click to expand...

So did you get higher noise or was something else wrong? I remember an AD app note where they showed their equivalent chip this way.

I guess these single-ended inputs have far higher current noise at the inverting input as do many CF op amps.

Samuel

"Self-noise can be a tricky one; it's only recently that 1/f noise ceased to be such a big problem; many earlier ICs and descrete amps had problems with low frequency 'bubbling' noises. "

Yes, besides the more-or-less smooth rise with a 1/(f^n) slope, popcorn noise, which is really a current gain random modulation among fairly well-defined levels was quite a problem in some old processes. The CA3086 transistor array was particularly bad. ST made a substitute that simply re-used the datasheet from the original manufacturer but was way way better. Unfortunately they discontinued it a while ago. That (no pun intended) part was a cheap means to making not-too-bad variable transconductance amps.

"I've seen the statement that low source impedances cause "the input stage to oscillate" - implying parasitic stuff in the input pair. That's BS, isn't it? It's simply the transconductance going up and we get nyquist instability, right?"

Not necessarily, in fact most probably not. It's more akin to the parasitic inductances interacting with the input device(s) at high frequencies, and the function of the R's in series is similar to those "grid stoppers" required in many tube circuits.

If it were a global loop compensation problem then typicaly lower Z is better, at least for a voltage feedback amp, since it moves the parasitic input C/feedback network resistance pole higher.

[quote author="Rossi"] ... have you tried the new ... THAT preamp chips ? The 1512 in particular looks very promising.[/quote]

I just received my samples ... finally
must do some tests
I'm excited
:green: