Common Centroid Resistor Array

I consider using the common centroid resistor array as shown in figure 14.1 in:

formatting link

to achieve resistor ratios of high stability. My concern is at what frequency crosstalk will be a significant effect. How high a bandwidth is possible?

This would be for a TIA circuit of topology shown in Figure 9 on page:

formatting link

which must have a bandwidth of 40 MHz where the common centroid resistor array would be used for the 2 Mohm resistors in the figure. 40 MHz means I cannot use the same components as Figure 9, but I expect the topology will be useful. High DC output stability is desired.

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

.1 in:

ency crosstalk will be a significant effect.

array would be used for the 2 Mohm resistors in

t I expect the topology will be useful. High DC

The parallel capacitance of a spiral cut thin film resistor tends to be aro und 0.3pF. L-trimmed planar (surface mount) resistors do appreciably better .

I'd expect there to be a problem with cross-talk with 2M resistors before you got to 40MHz.

If you want to use a thin film resistor array, you are going to buy it from Vishay or one of their competitors (if there are any left), The data shee t for the part you buy should offer some information on the subject, and i f it doesn't you can always ask for technical advice.

If you ground both ends of every second resistor in the array, you might be able to do better. An array with a conducting stripe between each resisto r would be even better, but that would add to the stray capacitance to gro und.

Ralph Morrison's book on grounding and shielding is good on the fundamental s

formatting link

The first edition didn't pay any attention to RF, but this is the sixth edi tion, and it does have some discussion of RF problems.

--
Bill Sloman, Sydney
Reply to
Bill Sloman

Resistor matching is a nit in that circuit. The base current of the bootstrap device causes some significant output offset (100-150 mV) anyway, and then there's the beta tempco of Q1 and Q2. (I usually put a diode-connected transistor in series with the upper resistor just to remove the tempco of Q1's V_BE, but that's a fine point.)

You can replace Q2 with a CPH3910 JFET running at about 10 mA, which will be quieter anyway, and whose transconductance is slightly better than a BJT running at 300 uA.

How high is high?

Cheers

Phil Hobbs

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

Thank you for your reply:)

I have abandoned the idea of using a common centroid resistor array for this.

I have seen the diode connected transistor in the version of this front end circuit you have in your book. For this I intend to use the same transistor used for Q1 just for the benefit of part commonality.

In your book you describe single transistor capacitance multiplying filters. I would like to use these on the +/- 15V supply lines which are connected to the two 2 Mohm Rbias resistors (but not the op amp's power supply connections because these are not ohmic, and the noise reducing base to emitter feedback requires the load be ohmic). The positive supply would require a high gain NPN BJT that can go to 40 MHz. And the negative supply would require the same for a PNP.

The higher the gain, the more the capacitance is multiplied. I am somewhat concerned because there is a relationship between bias current, and transistor gain. 2 Mohm, or similar, gives very little current. I had thoughts of and additional emitter to ground resistor just to get the bias current to where gain is optimum for the transistor.

Any suggestions for transistors to use in this capacitance multiplying circuit to be used in a filter that must filter up to 40MHz?

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

In general when you use a capacitance multiplier on a supply rail it is to filter low frequency noise using smaller capacitors than would otherwise be needed. At 40MHz even quite a small capacitor can take over.

piglet

Reply to
piglet

The bandwidth of the cap multiplier should be closer to 4 Hz than 40 MHz. It's just a glorified RC filter, after all. For NPNs, I like the

2SD2704K below 20 mA and the 2SD2114K from 20 to ~100 mA.

For a PNP, you might check out the BC807-40.

You want to put an RC lowpass in series with the collector as well, to prevent feedthrough due to C_CB and Early effect.

Modern op amps tend to be disappointing in that circuit because all the HV ones with low input current have much higher input capacitance than the LF357, so they trip over their own big feet. I've tried quite a few, and they're very disappointing compared with that gizmo from 1975.

I'm actually in the process of ditching that circuit for the third edition, in favour of a JFET bootstrap running at I_DSS, connected directly into a bipolar op amp.

Cheers

Phil Hobbs

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

How about the OPA657? 1.6 GHz, 2pA, 4.8 nV/root(Hz), 0.7pF differential,

4.5pf common mode. The noise corner appears to be about 1 KHz.

I have been studying low noise ripple filters for a GPSDO. I use +15V for Vin, and try to get close to +10V output. The current drain is 50 mA.

The capacitors are a new product by KEMET: Aluminum Organic Polymer,

100uF 25V, ESR=14mOhms ESL=2.5nh Dia = 6.3mm Height = 5.7mm Part # A768EB107M1ELAE036 $0.47 Qty 1 at Mouser and Digi-Key. Both have stock.

A cascode beta multipler works best. The response is -8 dB at 1 Hz, -115 dB at 120 Hz, -140 dB at 1 KHz, and -160 dB from 6 KHz to 3.6 MHz.

A Sziklai pair is almost as good. It follows the Cascode to about 160 Hz, then is 10 to 40 dB worse from 1 KHz to 1 MHz.

The PNP version is hopeless. The upper transistor (Q2) saturates and takes itself out of the picture. This illustrates the futility of trying to stack two current sources in series. You can never get them to match, and one or the other will saturate.

The LTspice files are at

formatting link

--
The best designs are no accident - sw
Reply to
Steve Wilson

The problem with the PNP is that the base labels are the same, so they're shorted. As I've said before, that circuit is not a current source, it's a simulated inductor.

Cheers

Phil Hobbs

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

[...]

Good catch. Thanks. Although still not as good as the others. I had already fixed that dupe once before, but somehow it snuck back in. I used to proofread my articles by reading them upside down. LTspice doesn't allow you to do that, but maybe it would be worthwhile capturing the screen and rotating it in XnView. Or maybe a simple batch program could read the labels and look for duplicates. There would be problems with copies of a circuit that used the same source, but that should be easy to solve.

The labels are of the form

FLAG -480 -352 R2C2 FLAG -272 -368 Q1B FLAG -160 -304 Q1E FLAG -272 -192 Q2B FLAG -480 -128 R4C5 FLAG 240 -352 Q3B FLAG 368 -400 Q4B FLAG 368 -272 Q3E FLAG 240 -48 Q5B FLAG 352 32 Q5E

It would be easy to read the file and store the label in an array. Each time a new label is read, scan the array and see if it already exists.

The array would be short enough that a simple linear search would be fast enough. No need for Quicksort or any of the faster versions.

This would be difficult to do in a batch file, but Python would probably work well. A good reason to start learning Python!

Of course, the program would be useless for people who do not label every node. But they run into different problems every time they change the circuit and LTspice renumbers the nodes.

What about the OPA657? I asked this earlier but got no answer:

--
The best designs are no accident - sw
Reply to
Steve Wilson

Highish Cin, plus I lose 10 dB of dynamic range because it's +-5V.

Cheers

Phil Hobbs

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

Cut 'n' paste gremlins. I only put labels on nodes I'm actually measuring. But then I'm very bad form, as you said awhile back. ;)

Cheers

Phil Hobbs

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

On 2/14/2021 3:08 AM, Phil Hobbs wrote: > Artist wrote: >> On 2/10/2021 11:27 AM, Phil Hobbs wrote: >>> Artist wrote: >>>> I consider using the common centroid resistor array as shown in figure 14.1 in: >>>> >>>>

formatting link
>>>> >>>> to achieve resistor ratios of high stability. My concern is at what frequency crosstalk will be a significant effect. How high a bandwidth is possible? >>>> >>>> This would be for a TIA circuit of topology shown in Figure 9 on page: >>>> >>>>
formatting link
>>> >>> Resistor matching is a nit in that circuit. The base current of the bootstrap device causes some significant output offset (100-150 mV) anyway, and then there's the beta tempco of Q1 and Q2. (I usually put a diode-connected transistor in series with the upper resistor just to remove the tempco of Q1's V_BE, but that's a fine point.) >>> >>> You can replace Q2 with a CPH3910 JFET running at about 10 mA, which will be quieter anyway, and whose transconductance is slightly better than a BJT running at 300 uA. >>>> >>>> which must have a bandwidth of 40 MHz where the common centroid resistor array would be used for the 2 Mohm resistors in the figure. 40 MHz means I cannot use the same components as Figure 9, but I expect the topology will be useful. High DC output stability is desired. >>> >>> How high is high? >>> >>> Cheers >>> >>> Phil Hobbs >>> >> >> Thank you for your reply:) >> >> I have abandoned the idea of using a common centroid resistor array for this. >> >> I have seen the diode connected transistor in the version of this front end circuit you have in your book. For this I intend to use the same transistor used for Q1 just for the benefit of part commonality. >>

supply lines which are connected to the two 2 Mohm Rbias resistors (but not the op amp's power supply connections because these are not ohmic, and the noise reducing base to emitter feedback requires the load be ohmic). The positive supply would require a high gain NPN BJT that can go to 40 MHz. And the negative supply would require the same for a PNP. >> >> The higher the gain, the more the capacitance is multiplied. I am somewhat concerned because there is a relationship between bias current, and transistor gain. 2 Mohm, or similar, gives very little current. I had thoughts of and additional emitter to ground resistor just to get the bias current to where gain is optimum for the transistor. >> >> Any suggestions for transistors to use in this capacitance multiplying circuit to be used in a filter that must filter up to 40MHz? >> > > The bandwidth of the cap multiplier should be closer to 4 Hz than 40 MHz. It's just a glorified RC filter, after all. For NPNs, I like the 2SD2704K below 20 mA and the 2SD2114K from 20 to ~100 mA. > > For a PNP, you might check out the BC807-40. > > You want to put an RC lowpass in series with the collector as well, to prevent feedthrough due to C_CB and Early effect. > > Modern op amps tend to be disappointing in that circuit because all the HV ones with low input current have much higher input capacitance than the LF357, so they trip over their own big feet. I've tried quite a few, and they're very disappointing compared with that gizmo from 1975. > > I'm actually in the process of ditching that circuit for the third edition, in favour of a JFET bootstrap running at I_DSS, connected directly into a bipolar op amp. > > Cheers > > Phil Hobbs >

I do not mean the filter bandwidth of the capacitance multiplier as it filters out a regulator's output ripple, and other noise.

The bandwidth I am referring to is the multiplying transistor output's source impedance at 40 MHz, which has a bearing on ts ability to regulate a load that has 40 MHz signal on it, and how much multiplying it can do at 40 MHz. I suppose the higher the load frequency it must regulate, the less capacitance is needed.

Do you use capacitance multipliers on op amp power supply terminals also? If so do you give them their own capacitance multiplying filters? Or is it the on the same filter as the two 2 Mohm resistors in Figure 9 of:

formatting link
?

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

You don't need the transistor to do anything at 40 MHz except block junk on its collector from getting to its emitter.

The output cap will take care of the output impedance just fine.

Depends on the required BW. Op amps generally have pretty crappy supply rejection up there, but on the other hand your LC decoupling network at the power input will get rid of most high frequency stuff.

In a situation like this I might run the op amp off the first stage's emitter and the photodiode bias off the second stage.

If so do you give them their own capacitance multiplying filters?

What are your actual design requirements? The frontends paper is a discussion of how to go about doing a front end design, in the context of an extended example, but your situation is clearly quite different.

I sure wouldn't just clone that circuit without verifying that it can do what you want. (It's a good circuit and all, but front ends are all horses for courses.)

Cheers

Phil Hobbs

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

This TIA must have a bandwidth from DC to 40 MHz. All signal in the spectrum within that bandwidth must be low in noise. It is not a narrow band pass situation.

The TIA output is to be split into two channels, one DC, and the other AC. The split is to be at 20 kHz.

The need is for it to be usable over a broad range photocurrents. Noise is crucial. A high quality, low noise, output signal is needed over that input range. I expect an adjustable transimpedance value will be necessary by electromechanically switching resistors using HF band relays. The lower the noise, the fewer gain selections wll be needed.

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

Understood. You might do well to build a prototype cap multiplier and see how it works. The f_T of the transistors is a second-order effect.

Cheers

Phil Hobbs

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

How well can the capacitance multiplier output take a capacitive load? Suppose a capacitor is placed in parallel with the load to cover the part of the band above the frequency the multiplying transistor can respond to?

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

You always have a big cap there for that reason.

Cheers

Phil Hobbs

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

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.