Is there still an engineering Usenet group where the vast majority of posts are on topic for electronic design?
In case there are any electronic engineers still watching this group, here is my question:
I need to design a TIA of bandwidth up to 5 MHz. The analog anti aliasing filter is to be set at 5 MHz. I would like the front end to have a transimpedance value of 4.02 kohm. The A to D converter's sample rate is to be 10 Mhz. So that the A to D conversions will alias very little noise in the bandwidth region 5 MHz, to 10 MHz, I need resistors that have self resonance well beyond 10 MHz.
So now I am looking at HF resistors. Because noise is primarily determined by the gain of the first stage, it is desirable for the front end's feedback resistor to be 4.02 kohm.
The Vishay FC series resistors are made in values only up to 1 kohm. They are very expensive, and so I am reluctant to put 4 of these 1 kohm resistors in series. They are, however, available.
The Visay HCHP series is made in values well beyond 4.02 kohm. But I cannot find them on the shelf anywhere, and they are extraordinarily expensive. The minimum buy makes them cost prohibitive to acquire, and the lead times are too long.
I would appreciate advice on any alternatives.
I am not able to find device models for the common thin film metal resistors. I do not know how well they would work up to 10 MHz. I would like to hear from any who are experienced in using them up this frequency.
A common resistor really doesn't have a self-resonant frequency, or one that you'd notice.
Ordinary surface-mount resistors will be fine at 1 GHz, much less 5 MHz. The parallel capacitance of an 0603 is around 0.05 pF. You're probably putting a bigger cap in parallel in the TIA.
Element 14 has Susumu 4.02k 0.1% parts in stock for less than a dollar each
formatting link
As surface mount parts they will be L-trimmed and have very little inductance or parallel capacitance. They will be fine at 5MHz, as John Larkin has pointed out. Back in the mid-1980's we got shocked when Philips put an L-trimmer surface mount roughly 1 Megohm resistor into a video preamp and it worked fine.
Philips - once a upon a time - published frequency response data for their spiral trimmed axial metal film resistors. The L-trimmed sufrace mount part did much better than that. Going to Vishay for you job would be an overkill.
I'd recommend Guy Macon's moderated group. misc.business.product-dev, but it died out 15 or so years back. (Old inhabitants of SED will remember.)
Is it for a photodiode, or something else? The usual issue is input capacitance.
What problem are you trying to solve? As others have mentioned, you can use a 4k resistor up into the hundreds of megahertz with no issues. You shouldn't even notice its shunt capacitance.
The main issue is that the shunt capacitance across a resistor depends on the configuration of nearby conductors and dielectrics.
Last week I was tweaking up a TIA with a 10M feedback resistor that works up beyond 1 MHz. It uses a combination of a bootstrapped pour on one end of the resistor and an adjustable RC lead-lag network to flatten out the gain.
From an RC point of view, that's about 500 times harder than what you're aiming at--the shunt capacitance of a SMT resistor all by itself is about 0.1 pF, and it's more like 0.05 pF when mounted on a ground-plane PC board. (You can measure this yourself, if you pony up $100 for a used Boonton 72 three-terminal capacitance meter.)
At 4k, the RC corner frequency will be around
f_c ~= 1/(2 pi * 4k * 100 fF) = 400 MHz,
and probably well above that when it's board-mounted. Any bandwidth problems you may be having are coming from elsewhere.
It had been a long time since I have been in this group. Just before starting this thread what I saw was spam, off topic postings, and nothing about electronic design. I have since seen several on topic threads.
It will be a photodiode, currently it is the OSI Optoelectronics InGaAs-120L-FC . This might change.
formatting link
The currently selected op amp is the LTC6226
formatting link
the TMUX7462F
formatting link
I would like to be able to have several gain options for the first stage gain from 4K all the way up to 630k. The switch in consideration for this is: MAX14777.
formatting link
formatting link
current idea is to have a high value resistor in the vicinity of
630k in parallel with the switches, and arrange the switches to switch a choice of one of 4 parallel resistors. The would provide 5 gain options. and more if more then one resistor is switched into the feedback path. I know having this option will add capacitance to the virtual ground, even though the analog switch would be between the op amp output, and the resistors.
I am aware of your virtual ground bootstrap techniques for high capacitance photodiodes to raise bandwidth, and reduce noise, by reducing the capacitance seen by the virtual ground. I do not know right now if bootstraping would add more capacitance than would be reduced.
Is the next edition of your book in which you describe bootstraping out? I already have the one before it.
On a sunny day (Wed, 27 Jul 2022 02:39:02 -0700 (PDT)) it happened Phil Allison snipped-for-privacy@gmail.com wrote in snipped-for-privacy@googlegroups.com:
Jo La is just an ego tripper his business must run bad with all that time on his hands to post here and insult people.
People like Artist come and go, thought he was an alter-ego of Ph Ho to sell his book.
That's a really small one, about 2 pF. e_N*C noise current will start to dominate the Johnson noise current when
f > sqrt(4 k T / R_F) / (2 pi * e_N * C_in),
where C_in is the total summing junction capacitance, which includes board strays and diode capacitance as well as the amplifier's C_in.
Some good things to remember:
1 nV is the 1-Hz voltage noise of 60.4 ohms,
1 pA is the noise of 16.3k ohms, and
1 pA is also the shot noise of 3 uA.
Figuring 10 pF for all the gingerbread you're planning to hang on the summing junction, a decent (5 nV) CMOS or JFET amp with a 4k feedback resistor will be in the e_N*C region above
f = 1 pA*sqrt(16.3k / 4k)/(2 pi * 10 pF * 5 nV) = 6.5 MHz.
Which isn't horrible, but could be better--you'll lose a dB or so of SNR at low light.
However, at 600k ohms, things start falling apart at only 43 kHz, and it gets _cubically_ worse above there--since the e_N*C current spectral density goes as f, the noise power density goes as f**2, so the integral goes as BW**3.
Marginally okay at 4k, but not a good choice for higher resistances--its input current noise is 2.7 pA in 1 Hz.
From the rules above, that 2.7 pA equals the shot noise current of
3 uA * 2.7**2 = 22 uA
and the Johnson noise current of
16.3k / 2.7**2 = 2.2k.
On the other hand, that 1 nV typical noise number helps the e_N*C problem a lot--it goes away at 4k, and is 14 dB better at the higher resistances.
Completely unsuitable--its OFF capacitance is over 10 pF. At 600k ohms, you'll have 4k + 10 pF from the SJ either to the output or to ground.
If it's to the output, it appears in parallel with the 600k, and so will reduce the bandwidth to about 25 kHz.
If it's to ground, that will give you an AC noise gain of 150, besides dumping almost the full Johnson noise of the 4k resistor into the summing junction. (You'll also need two switches per resistor.)
Relays are the ticket for that job. Sometimes you can get away with a lower-capacitance MUX such as a TMUX1511.
Sounds like you're going to need to use more than one complete TIA, and use relays to switch them in and out. If you're not bootstrapping, you can use one TIA at each end of the photodiode, with suitable bias arrangements. That does cost some supply headroom, because the PD always wants to make the TIA move in the inconvenient direction.
Sounds like a candidate for what I call a type-2 bootstrap, where the op amp's inverting input is driven by the bootstrap output, so that the bootstrap is inside the TIA's feedback loop instead of being off to one side.
That gets rid of the op amp's C_in, which is a big win in this instance, and allows you to use low-e_N bipolar amps such as the LT6226. It does add the JFET's offset and noise voltage, but that's an excellent trade.
It works much better than the old-fashioned JFET-pair + op amp approach.
Yup, it came out in January. Fully revised, and with a whole lot of new material (the LaTeX source is about 37% longer, but some of it is commented out in both editions).
Wiley had to issue it in coffee-table size (8.5x11) to avoid going over their mystical 900-page limit. New stuff is now mostly going into the new book manuscript, tentatively titled "Designing Electro-Optical Systems: 30 Instruments".
No, really, I have function. Crazy ideas. Inventing things.
Nowadays, I can't do everything. FPGAs, Linux, Python test set software, PCB layout, mechanical design, all that can be mostly delegated to saner people.
"Oh, but it just may be a lunatic you're looking for..."
Who have had a lot of practice at flattering the boss until he moves on to pester somebody else.
He can talk to customers for ages, and the more time he spends doing that, the less time he spends distracting people who can still design circuits that work.
Even the most lunatic idea can be repackaged into something that can be made to work, as long as you can keep the geriatric clown out of the detailed design.
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.