super precision Volt ref.

[MrAl]

Hi,

spec:
Dont jump the gun just yet

A spec without defining the source impedance and load impedance for each of multiple outputs is not a spec.

Nor is max load 50 Ohms a spec. nor is it a pulse with 1ns rise time.

If you model an ideal voltage source, load does not matter with tolerance.
But a true voltage source contradicts a 50 Ohm source with <1ns rise time.

Hence what spec!

Tony:

This brings up the question, what spec's are there in the original requirement? That really tells us how critical we have to be.
For example, we know from the general diode curve that when D3 is conducting D2 will not be conducting much at all, so that means the voltage drop across D3 will be higher than the drop across D2. The saturation voltage of the output transistor of the 555 is non zero. These two taken together means that the output will not reach exactly to zero when the output state is logically zero.
But does it have to go all the way to zero?
We also know that op amps do not respond as fast as we would like sometimes, but do we have an original spec that tells us what rise and fall times have to be met?
So it depends a lot on what we really need. I think it would be wrong to compare to a commercial unit unless we really were trying to duplicate that kind of device perfectly.

 

[Tony Stewart]

Super precision and super fast is not super easy nor cheap.
I pasted original specs in previous post link just below...

View attachment 95129

Before starting any design ( my frequent rant) you must write your own Design Spec.
If duplicating this instrument, ensure this is what you need, and we don't start chasing skinny little rats late into the design....

- for example this product has a rise time < 1ns with differential outputs +,- Is that what you will do?

if the signal risetime is at least seven times faster than the risetime of the oscilloscope vertical system, the displayed (observed) waveform will have a
risetime that is very close to the risetime of the vertical system. ...

If a fast-step signal produces a crt display with little or no overshoot or ringing, the product of oscilloscope risetime and oscilloscope bandwidth should result in a factor whose value lies between 0.329 and 0.350.


View attachment 95129

These are the specs he wishes to duplicate, because the instrument is very expensive.

Super precision and super fast is not super easy nor cheap.


The open circuit voltage will be double 50 Ohm load and that source impedance must be as precise as the load and tolerances expected in addition to source voltage error.

 

[Mikebits]

Apologies Tony I have obviously done bad. The trouble is being a newbee I dont understand the form: I just focus on the electronics and forget the wider issues. Perhaps ask admin to delete my posts and as they are might be misleading.

I did not mean to cause you any problems Spec, I hope this is not my fault. My confusion on the circuit should not be your problem, it is mine. Having members who are willing to devote their personal time to help other people solve problems is a good thing, and should be encouraged, so I apologize for any trouble or hard feelings I may have caused.

 

[MrAl]

I did not mean to cause you any problems Spec, I hope this is not my fault. My confusion on the circuit should not be your problem, it is mine. Having members who are willing to devote their personal time to help other people solve problems is a good thing, and should be encouraged, so I apologize for any trouble or hard feelings I may have caused.

Hi Mike,

I dont think you should feel bad in any way. Me for one, i was happy to see the questions come up because it gave me a chance to review these aspects of design too. I have to review stuff like this from time to time so it is fresh in memory. It is good to remember stuff like this so i appreciated your question.

 

[Mosaic]

Just to clarify here...

I am duplicating the functionality needed to calibrate the 24XX series of scopes. The PG506 provides precision amplitude (vertical) calibration precision in this case as opposed to rise time precision. Square wave to a few Khz would be ok. The PG506 rise times are woefully inadequate for 400Mhz scope bandwidth evaluation.
The 50 ohm requirement is good enough to 5V.

For timing calibration I would need a sub 100 picosecond rise time square wave for which 50 ohm impedance matching is required to handle transmission line issues. I am considering a comparator with hysteresis control to sharpen the pulse rise times and effectively permit both bandwidth evaluation and time mark generator functionality. I am considering using this device, driven by a uC:
https://www.farnell.com/datasheets/1903291.pdf

I haven't examined doing a leveled sine wave gen. optimally yet. It's probably going to be the hardest thing to do. I have a DSA815-TG Spect. Analyzer to use when developing that leveled sine wave signal.

I may just buy one if I get a ok used deal.

 

[Tony Stewart]

You will probably want to use miniature semi-rigid coax for routing to SMA connectors on board, if not front panel mounted., and use GETEK or quality Polyamid PCB material rather than cheap FR4.
This is pretty standard for high quality instruments.

Return loss is far better for precise signal quality. It's not hard to make. When I made a calibration jig for testing up to 6GHz, an RF design Engineer at Motorola in Arizona showed me how to test their LNA and difference in quality of connectors and cable is not visible to the naked eye can be significant even at 1 GHz.

 

[Tony Stewart]

I did not mean to cause you any problems Spec, I hope this is not my fault. My confusion on the circuit should not be your problem, it is mine. Having members who are willing to devote their personal time to help other people solve problems is a good thing, and should be encouraged, so I apologize for any trouble or hard feelings I may have caused.

No Worries spec welcome to the group. Your depth of experience and insight is always welcome. I have always tried to think outside the box, which helped me do the impossible in my career and make anything work. So I like to share my experiences from that perspective.

 

[Mosaic]

You will probably want to use miniature semi-rigid coax for routing to SMA connectors on board, if not front panel mounted., and use GETEK or quality Polyamid PCB material rather than cheap FR4.
This is pretty standard for high quality instruments.

Return loss is far better for precise signal quality. It's not hard to make. When I made a calibration jig for testing up to 6GHz, an RF design Engineer at Motorola in Arizona showed me how to test their LNA and difference in quality of connectors and cable is not visible to the naked eye can be significant even at 1 GHz.

Well, it's not production and about 800Mhz max, so I can prob get by with vanilla FR-4, I'll try to get some 1mm or 1/32" FR4 to do stripline.

 

[Tony Stewart]

Well, it's not production and about 800Mhz max, so I can prob get by with vanilla FR-4, I'll try to get some 1mm or 1/32" FR4 to do stripline.

Keep in mind dielectric constant tolerance of 20% is typical which affects transmission line impedances, ringing etc.

 

[Mosaic]

How bad is FR1 board for this?

 

[Tony Stewart]

FR1 is just a lower flame retardant spec because it is cellulose based with epoxy and not very strong so no PTH and poor dimensional stability needed for stripline.

Materials with lower values of permittivity are increasingly needed for high-frequency applications.
Figure 1 plots both permittivity and loss characteristics for a range of materials, from which it can be
seen that lower values and losses than conventional epoxy-glass boards are really only available by
using laminates based on cyanate esters or PTFE.

If you have some material etched, try seeing if it gets hot in a microwave oven ( no copper) or if it bursts into flames it has a poor loss tangent or dissipation factor and will give poor results for controlled impedance fast rise time signals.

 

[Mikebits]

Well, it's not production and about 800Mhz max, so I can prob get by with vanilla FR-4, I'll try to get some 1mm or 1/32" FR4 to do stripline.

I don't think you would want to go stripline as that would force you to buy a 4 layer board since the stripline is sandwiched between two layers. Not only that, but most of the times the transmission line impedance will not be matched to the load so some sort of tuning is required usually adding a matching circuit at the load like series L, shunt C. You really have a tough time tuning Stripline as it is a inner layer.

Also if you look at a standard 4 layer board the stack up will make the stripline calculations more difficult.

MicroStrip is done on the top layer and you can get by with a 2 layer board, not to mention you can tune the trace to match the load if need be (assuming you can get your hands on a network analyzer).

Well anyways, that is my thinking on the topic, and is fresh in my mind as I been doing a lot of reading on RF layout this past week, but I am in no way a Jedi RF master like Radio Ron or Toni Stew, so they would be much better resource on this subject and I.

I am in the process of building a reflectometer to measure and tune magnitude and phase of a 868 MHz trace for a SMT chip antenna.I am doing this since I can't get a VNA. I am still doing the layout for the reflectometer so I don't know how well it will work.

 

[Mikebits]

I almost forgot, Grounded Coplaner Waveguide (CPWG) is another layout option you have, and the nearby ground makes soldering tuning parts easy. I have avoided this option as the math looks complicated and I am slightly a dunce when it comes to math (I had to take calculus 3 times before I got a C), so I just stuck with microstrip. Take a look at the CPWG math equations below:


Just to give you an example of what I mean about the tuning, below is a snip from a tuning recommendation from an Appnote provided by a chip antenna manufacturer. You can see where they provided pads on the transmission line for tuning.


I don't mean to derail the topic main thread, and perhaps if warranted, this topic would be better left to a new thread, so I am done and return you to what it was which was the main topic.

 

[Mikebits]

No Worries spec welcome to the group. Your depth of experience and insight is always welcome. I have always tried to think outside the box, which helped me do the impossible in my career and make anything work. So I like to share my experiences from that perspective.

Tony, I think you got me confused with Spec since you replied to my post and not Specs. Or am I once again confused?

 

[Tony Stewart]

It was directed to Spec and also to add to your welcome to him.

 

[Tony Stewart]

FR1 is just a lower flame retardant spec because it is cellulose based with epoxy and not very strong so no PTH and poor dimensional stability needed for stripline.

Materials with lower values of permittivity are increasingly needed for high-frequency applications.
Figure 1 plots both permittivity and loss characteristics for a range of materials, from which it can be
seen that lower values and losses than conventional epoxy-glass boards are really only available by
using laminates based on cyanate esters or PTFE.

View attachment 95141



If you have some material etched, try seeing if it gets hot in a microwave oven ( no copper) or if it bursts into flames it has a poor loss tangent or dissipation factor and will give poor results for controlled impedance fast rise time signals.

Note that FR-1 and CEM-1 are never considered for broadband UHF design for a reason. Loss tangent, dissipation factor and change in dielectric constant over the band being used. Although narrow band tuners can be made this way.

Although if you want to experiment with variable CPWG stripline resonators on 2 layer boards, you can try to document the Q , loss and impedance errors but it is better to use a VNA and measure scattering parameters.

 

[ronsimpson]

How would I go about producing super precision ref Vs with 0.25% precision?

Can you see 0.25% on a o-scope?

 

[Tony Stewart]

If that R controlled offset V, and you have a gain of 100 it becomes 25% error in apparent offset voltage.
If that error controls sweep speed on expanded sweep then the error would be amplified again.

In order to calibrate a DMM, one needs an accurate reference better than the DMM.

The <1ns fast pulse 50 Ohm differential output is classic Tektronix discrete design.

Here using four precision cascaded differential amps with output +1.00V to 0V and -1.00V to 0V output into 50 Ohms. Note the symmetrical 16.5V supplies and the alternating symmetrical complementary stages of NPN then PNP, then NPN then PNP for the positive fast pulses and for the negative NPN > PNP>PNP>NPN where the 1st two stages are shared.

Brilliant speed , linearity and symmetrical low differential propagation delay or skew.

The circuit has an unusual precision differential Schmitt trigger in the 1st stage that serves as a limiter and before the last stage a 40mA constant current source to drive 25 Ohms (source and load) for a switched current source with gain and offset adjust both outputs from common controls.

 

[spec]

Hello everyone,

Thanks for your kind words. Firstly can I say that I joined ETO because of an interest in electronics and also because you all seem to be a good bunch of guys and kindred spirits, whether newbee or electronics guru. I also joined to learn some more about electronics and that has already turned out. I can't find the post again, but someone posted about programmable resistors- I had never come across those before. The other thing is, as I have said before, I don't have a workshop at the moment as all my electronic stuff was lost in a house explosion. So it would be interesting if some of my designs were built to see how they turn out. Willen has already done this and, to my surprise, he says it worked!

Mikebits, you didn't cause me any problems in the slightest, in fact, quite the opposite. I appreciate discussions on any aspects of my circuits. I particularly like errors or flaws to be pointed out as it helps greatly to get the design right and that's the only objective; if it hadn't been for you and MrAl I probably wouldn't have noticed the error with the 25K resistor. And Stuart has pointed out a major problem with current steering diodes where I had been blasé with the design. Just to let you know it's fairly simple to sort out.

Stuart, your post didn't cause me any problems either but I may have misread your intentions. I see from your later posts that you know your stuff. The other thing was that I got the idea somehow that you were an ETO moderator, and that you were pointing out a transgression of the ETO protocol, probably because of your confident style and the number of your posts: 1912 quite something. Also, it had been one of those days, you know the type, the car threw a big end, the wife ran off with my best friend and I'm missing him, and the dog just chewed through the TV power-cord and survived. Oh, and someone just deleted an hour's work on Wikipedia and replaced it with a load of gobbledygook!

Over the years, I have written, edited, and reviewed quite a few specifications, mainly for the military: feasibility, study, cardinal point, design, procurement, test, through life cost, safety and on and on. Many of the peripheral specifications were full of boilerplate and write only, but they took up an extraordinary amount of time. So my feeling is that I didn't really want to write any more on EON. As I said, I'm only really interested in electronics so that's why I was considering bailing out- I could still read the posts though, thousands of them. But if someone cared to post a succinct specification for the calibrator, that would be a different matter. The other point is time; I have a lot of things going on; I expect you are all the same.

You are spot when you stress the importance of specifications, but it's a complicated area for a many reasons which I could write a long post on. Specifications are like dynamite, chain-saws, and emulators: powerful tools if used well, but devastating otherwise. What made me smile is that you were reflecting exactly my point of view about specifications. Also, by pure coincidence, my user name!

I have been going on a bit, so back to the scope calibrator.

There are some interesting subsequent posts, especially Stuart's giving the circuit diagram of the commercial scope calibrator. Do you notice anything familiar about Q920/Q935 and R935?

I will post some more about my proposed approach, both the macro and micro aspects, and perhaps cover some of the points you have all made. Also, I hope to cover some of the misunderstandings about the purpose of the circuit. I see now that some of my statements might be misleading.

 

[Tony Stewart]

I think this mainframe calibrator is a universal plug-in unit and not just for calibrating scopes.

I think you meant the parts in green circles. Yes each output final stage is shielded by metal with feedthru's to prevent E-H field crosstalk for a clean signal/noise ratio and low jitter from dV/dt of >1V/ns = >1000V per microsecond slew rate with sub nanosecond slew rate. ( which is faster than "Damn") or dI/dt >40mA/ns or >40A per us or ...40kA/ms like a power supply short cct. transient in EH fields.

There are alot of biasing arrangements going on here to get a super linear super fast current switch with a common mode adjustable current source with 2 pots so that the output impedance is exactly 50 Ohms ( laser trimmed disc) at 1V , 0V and in between. i.e. high s22 broadband.