Gain tweak--needs to be medium fast and very very quiet

ut.

I don't suppose it would help to use two pots as variable resistors top and bottom. One could do the coarse tweak, and the other fine.

Hmm, other cofigurations possible too... If I'm allowed two pots.

George H.

I've thought about just putting four in parallel (one AD8403). The zipper noise is a problem though.

If LM13700s were faster, I could use one of those...2 MHz just isn't quite there. I think it uses lateral PNPs, so even with a 5 ohm load it wouldn't speed up much. The OPA861 is very fast, but not very linear. It sure would be a useful part if it were as linear as an LM13700 with the diodes on the input. The noise is highish, but once it's divided down it would probably be OK.

VTC used to make some nice OTAs. RIP.

Cheers

Phil Hobbs

What? No solution in AoE ?:-) ...Jim Thompson

ut.

put the gain tweak after the first stage of gain where noise is less critical..

Mark

There are some MDACs that have that sort of frequency response from the Vref input, and 12-16 bits of resolution. Of course, you'll still get small glitches as you trim gain, but they aren't bad.

You could do really coarse gain trims with mosfets or even relays, and fine trim with an mdac or dpot, to keep the noise down. How about a

Or a dip switch for the coarse, mdac or dpot for fine. Cheap.

John

You can even get changeover opto-MOS relays.. the isolation wouldn't hurt.

On second thought, SSRs would probably have too much capacitance.

John

There's no such place, unfortunately. This network has to go between a predistortion circuit and a BJT diff pair--there's no more than 250 mV total swing available. The gain tweak is to take of the difference between the kT/e factors in the two different transistor arrays, so the predistortion works correctly. You can think of it as being like the linearizing diodes in an LM13700, but unfortunately it's spread out over three packages altogether.

Cheers

Phil Hobbs

How about the LT1228 as a substitute LM13700. It's faster - 75Mhz, and with good specs in other areas.

T24gSmFuIDI4LCA4OjM3oHBtLCBQaGlsIEhvYmJzCjxwY2RoU3BhbU1lU2Vuc2VsLi4uQGVsZWN0 cm9vcHRpY2FsLm5ldD4gd3JvdGU6Cj4gSSdtIHdvcmtpbmcgb24gYSBmcm9udCBlbmQgZ2l6bW8g dGhhdCBuZWVkcyBhIHNtYWxsIGdhaW4gdHdlYWssIGFib3V0ICstCj4gOCUsIHdpdGggYSBiYW5k d2lkdGggb2YgYXQgbGVhc3QgMTAgTUh6LiCgVGhlIGtpY2tlciBpcyB0aGF0IGl0IGhhcyB0bwo+ IGhhdmUgc3ViLW5hbm92b2x0IG5vaXNlLiCgVGhlIG91dHB1dCB3aWxsIGdvIHRocm91Z2ggYSAx MCBvaG06MTAgb2htCj4gdm9sdGFnZSBkaXZpZGVyIHRvIGRyaXZlIHRoZSBiYXNlcyBvZiBhIEJK VCBkaWZmZXJlbnRpYWwgcGFpci4KPgo+IEkndmUgdGhvdWdodCBhYm91dCBwdXR0aW5nIGEgZHBv dCBhY3Jvc3MgdGhlIHZvbHRhZ2UgZGl2aWRlciwgbGlrZSBzbzoKPgo+IFZpbiAwLS0tLS0tLS0t LS0qLS0tLS0tLS0tLS0tKgo+IKAgoCCgIKAgoCCgIKAgoCCgfCCgIKAgoCCgIKAgoHwKPiCgIKAg oCCgIKAgoCCgIKAgoHwgoCCgIKAgoCCgIKB8Cj4goCCgIKAgoCCgIKAgoCCgIKB8IKAgoCCgIKAg oCCgUgo+IKAgoCCgIKAgoCCgIKAgoCCgfCCgIKAgoCCgIKAgoFIgoDEwIG9obXMKPiCgIKAgoCCg IKAgoCCgIKAgoHwgoCCgIKAgoCCgIKBSCj4goCCgIKAgoCCgIKAgoCCgIKBSIKAgoCCgIKAgoCCg Ugo+IKAgoCCgIKAgoCCgIKAgoCCgUiCgIKAgoCCgIKAgoHwKPiCgIKAgoCCgIKAgoCCgIKAgoFIg oCCgIKAgoCCgIKB8Cj4goCCgIKAgoCCgZHBvdCCgIKBSPC0tLS0tLS0tLS0tKi0tLTAgVm91dAo+ IKAgoCCgIKAgoCCgIKAgoCCgUiCgIKAgoCCgIKAgoHwKPiCgIKAgoCCgIKAgoCCgIKAgoFIgoCCg IKAgoCCgIKB8Cj4goCCgIKAgoCCgIKAgoCCgIKBSIKAgoCCgIKAgoCCgUgo+IKAgoCCgIKAgoCCg IKAgoCCgUiCgIKAgoCCgIKAgoFIgoDEwIG9obXMKPiCgIKAgoCCgIKAgoCCgIKAgoHwgoCCgIKAg oCCgIKBSCj4goCCgIKAgoCCgIKAgoCCgIKB8IKAgoCCgIKAgoCCgUgo+IKAgoCCgIKAgoCCgIKAg oCCgfCCgIKAgoCCgIKAgoHwKPiCgIKAgoCCgIKAgoCCgIKAgoHwgoCCgIKAgoCCgIKB8Cj4goCCg IKAgoCCgIKAgoCCgIEdORCCgIKAgoCCgIKBHTkQKPgo+IGJ1dCBldmVuIHdpdGggYSAxayBkcG90 LCBJIGNhbid0IGdldCBlbm91Z2ggcmFuZ2Ugd2l0aG91dCBzYWNyaWZpY2luZwo+IHJlc29sdXRp b24uIKBCZXNpZGVzLCBnYWluIGFkanVzdG1lbnRzIHdpbGwgY2F1c2UgYXJ0aWZhY3RzIGF0IHRo ZSBvdXRwdXQuCj4KPiBBbnkgYnJpZ2h0IGlkZWFzPwoKQSBjb3VwbGUgb2YgcGFpcnMgb2YgSDEx RjEgcGhvdG8tRkVUcz8KCmh0dHA6Ly93d3cuZmFpcmNoaWxkc2VtaS5jb20vZHMvSDEvSDExRjJN LnBkZgoKLS0KQmlsbCBTbG9tYW4sIE5pam1lZ2VuCgo=

Xc ~ 1K at 10MHz.

Best regards, Spehro Pefhany

Interesting part, thanks. It seems to be very noisy though--700 pA/sqrt(Hz) at 10 Hz output noise. That's 3.5 nV even with my 5 ohm load!

Cheers

Phil Hobbs

to

hms

hms

utput.

.net

Can you change the tail current in the diff-pair to change its GM?

kevin

Thanks. It isn't the g_M that's the problem--it's the scale factor. If you run one diff pair at temperature T1 to make a predistorter for another one at T2, the output of the second one goes like

I_out ~ I_in**(T2/T1).

It's really

I_out = I_E2*tanh(T2/T1*atanh(Iin/I_E1)).

That isn't anything pretty unless T2 is very close to T1.

The whole thing has to work over a fairly wide range of g_M in the second diff pair--probably 1 uA to 5 mA per side. There's compensation for extrinsic emitter resistance as well--the circuit is a glorified laser noise canceller, using some of the ideas in

If I have to, I can temperature-control the diff pairs, but it'll be quieter and need less power if I can just dork the gain a bit.

Cheers

Phil Hobbs

AoE is a pretty good book. It's not for everyone, and I certainly didn't learn circuit design out of it...I didn't run across it till I was a (physics) grad student, after having designed circuits for a living for a couple of years. However, at that point it was a very good read, and I learned a lot from it.

Cheers

Phil Hobbs

Aha! In the manner of our Atlanta buddies ;-)

I wasn't quite sure what you were getting at. Let me think on it. ...Jim Thompson

Thanks, John. I could probably use an amplifier plus an MDAC, and do the gain tweaking in very small steps...1/2048 of 8% is 40 ppm per step, which would probably be quite visible. With 16 bits, it would be pretty well down in the mud, but would need a lot of stepping to get anywhere.

BJTs have really predictable characteristics, which is one reason I like them so much. In particular, once you compensate for r_EE', everything else is temperature, with a slight tweak from beta nonlinearity. So hopefully there won't be a lot of unit-to-unit variability of the sort that DIP switches would help. (They'd also need a bunch of technician time and a fairly swoopy test jig, which I'm hoping to avoid.)

Most of the temperature variation is probably going to be due to the power the box has to dissipate to maintain dynamic range through all that circuitry. That would make it reasonable to use some combination of relays or MOSFETs for the coarse tweaks plus a dpot for the finer stuff. The dynamic dissipation will probably cause less than 1 degree of change in reasonable circumstances.

Cheers

Phil Hobbs

Could be handy, especially if the capacitance is independent of the drive current, so that the voltage divider ratio is reasonably accurate even though the impedance will vary a bit.

There's nothing by way of a noise spec, though, and the nonlinearity isn't what I'd like it to be...but if I get stuck, I'll try a couple.

Thanks.

Cheers

Phil Hobbs

t +-

s to

hm

o:

=A010 ohms

=A010 ohms

output.

al.net

n a

er

.net

=A0If

Maybe that's easier? How big a temperature difference?

But the 'dork' is linear and low noise, to 10MHz,?

If you said it also had to cost less than a buck, you could call in Joerg. :)

George H.

t- Hide quoted text -