Separating high pass and low pass TIA output bands.

I have an assignment to design a transimpedance amplifier for a photodiode. The bandwidth is to be 40MHz, and there are to be two outputs, a high pass output above 20kHz and a low pass output below

20kHz all the way to DC. It is permissible that the two bands overlap a little bit.

The noise performance of an amplifier is primarily determined by the noise in the first stage where it is desirable to have its gain as high as possible. I realize the first stage's transimpedance will be limited by the input current's DC output level, and if this were absent the transimpedance value could be much higher to take full advantage of the op amp's output range, and thereby have much better noise performance.

So I am wondering if it is practical to have two first stage transimpdance amplifiers, one for each output band, and between these two, and the photodiode, separate the two output bands by means of a passive LC network.

At this time I know very little about the photodiode's source impedance, and what current levels to expect.

I am aware of inductor and capacitor self resonances, and would select these passives accordingly. I have studied high frequency board layout techniques. And I know about opamp gain bandwidth constants and how to figure gain and bandwidth compromises in a op amp voltage gain circuit.

--
To email me directly remove sj. from my email address's domain name.  
This is a spam jammer.
Reply to
Artist
Loading thread data ...

The frequencies don't seem that high in the grand scheme of things, once the signal is thru the first stage it seems like the SNR would be good enough to deploy a gyrator-based second order filter rather than use real inductors, you can control the response better

Reply to
bitrex

Good TIAs at that bandwidth are not that easy to design. I suggest spending most of your time on a fast, quiet DC-coupled TIA and filter afterwards.

You might find this article helpful.

Cheers

Phil Hobbs

Reply to
pcdhobbs

I have played with the idea of putting a TIA on both ends of a photodiode. It's probably dumb. The only use might be to do precise DC zeroing of a fast but DC-inaccurate amp.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

I have done this once. The bands did not overlap (probably could have) and there was one TIA optimized for RF, low noise and such but that had high offset and drift. The other was optimized for low drift and low offset errors because that was used to servo something. I can't go into details but it all worked fine. RF guys call this sort of splitter a diplexer.

As Phil said in your case this might not make much sense because 20kHz is already audio range and 40MHz is RF but not very high. In my case it was different, an audio range servo loop and then the RF was a certain band in the low VHF range. If you need professional support on this Phil is the guru in that field.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Look at figure 12 in this document.

formatting link

It's the DC restoration circuit. There will be noise from both A1 and R3 in the feedback loop, so there are trade offs here you need to make.

Reply to
Wanderer

I do not have choice regarding the photodiode. That will be chosen by someone else. An integrated photodiode tia package is not an option.

--
To respond to me directly remove sj. from the my email address's domain  
name. This is a spam jammer.
Reply to
Pflanze, Stephen

I'm not suggesting that you use the OPT301, just the circuit. It's just a TIA with another opamp in the feedback. The OpAmp in the feedback is the low frequency stuff. The output of the forward TIA is the high frequency stuff.

Reply to
Wanderer

One good reason to do that is to avoid range switching. You bias the noninverting inputs far apart, to get some bias on the PD, and use diode switching to select between the high- and low-gain paths by shorting out the one you don't want to use.

Of course you have to invert and level-shift one side, but it's sometimes useful.

Cheers

Phil Hobbs

Reply to
pcdhobbs

I see that schematic now, and it does look to me to be a better alternative to an LC network between the photodiode and two TIAs. This has an error integrator in the feedback loop to adjust the TIA's output to zero offset.

That method is compatible with the circuit in the article Phil Hobbs linked to. I will try it.

Thanks for this.

--
To respond to me directly remove sj. from the my email address's domain  
name. This is a spam jammer.
Reply to
Pflanze, Stephen

Thanks for your suggestion. I have your book, and I am studying that circuit in its chapter 18. The candidate opamp is:

Analog Devices ADA4817:

formatting link

It has a very high GBW and very low differential and common mode input capacitances. It also has a very convenient pin out for the feedback resistor and guard ring traces. Its offset error is a bit high, and it is supply voltage is limited to +/-5V.

There is a version that has two opamps in one package. I am going to avoid using it because I do not have a way to predict if cross talk within the package will be a problem.

Also considered are: TI THS4631:

formatting link
TI OPA656:
formatting link

I would use your capacitance buffering circuit in conjunction with the error integrator in the feedback loop in Fig 12, page 11, of this PDF:

formatting link
to eliminate DC while at the same time making the transimpedance value as high as possible. For this integrator I plan to use the same opamp I use for the TIA.

I am looking using metal foil resistors for the noise ctrical parts.

--
To respond to me directly remove sj. from the my email address's domain  
name. This is a spam jammer.
Reply to
Artist

I don't think they help. The Johnson noise is the same for all resistors, and reasonable-value cermets don't seem to have shot noise. I think I have seen shot noise in high-value cermets, like 100Mohms, but it's a difficult measurement, and I may have just been seeing tempco.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

ding most of your time on a fast, quiet DC-coupled TIA and filter afterward s.

7-1_4817-2.pdf

If you're doing what I think you're doing you are going to be shot noise li mited by the high dc current in the photodiode but you need to do the noise analysis. You should be able to find how to do that in a number of books o r app notes.

I've used the opa656 for similar circuits. Be careful of that split resisto r feedback circuit shown in the datasheet. It doesn't give the best roll of f of the noise bandwidth. I ended up sticking a 0.5 pF cap across the feedb ack resistor without a cap in order to meet my requirements.

Reply to
Wanderer

The point of the bootstrap is that it needs to be quieter than the op amp, so that the eN*C mechanism multiplies the 0.8 nV noise of the FET rather than the 4 nV noise of the op amp. That's 14 dB right there, assuming the FET doesn't dominate the PD's capacitance and that the PD's series resistance isn't too large.

What's your photodiode capacitance?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

The main issue with thick film resistors is 1/f noise, due apparently to conductance fluctuations at boundaries between conducting grains.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Yep. Back when I made hybrids, we used thin-film resistor chips for low-noise applications. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson

On Friday, July 21, 2017 at 8:37:16 AM UTC-7, Jim Thompson wrote: ...

...

They can take up a lot of space on-chip if you need high resistance though.

A mult-ichannel high-voltage amplifier chip I architected with Supertex had 16 12 Megohm SiCr resistors to meet noise requirements - they dominated the chip area.

kevin

Reply to
kevin93

I was referring to hybrids, Alumina substrates (1" x 2"), separate chips for resistors.

On today's microchips, 1K/?sq is commonplace, and several of the processes I use have 10K/sq.

Yep. But that's a bit extreme. Usually I have only a few meg-Ohms total... mostly in the bandgaps and critical mirrors. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson

On Friday, July 21, 2017 at 10:27:27 AM UTC-7, Jim Thompson wrote: ..

I realize that - sorry for being unclear what I meant by 'chip'.

I also did a hybrid implementation but we only had thick-film and we didn't go into production with that version.

I think we used ~5k/sq.

This was for a system where I had ~3,000 HV amplifiers for driving MEMS mirrors at up to 170v in an optical switch.

I had to keep the power down - hence the high value resistors. The stability and noise requirements forced us to thin-film rather than polysilicon.

A competitors amplifier had 100's of millivolts of noise when the output was at 150v - noise went up with output voltage - I guess they used poly.

kevin

Reply to
kevin93

Sounds like a project I worked at, IIRC, SpatiaLight.

...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.