UK Vintage Radio Repair and Restoration Discussion Forum

UK Vintage Radio Repair and Restoration Discussion Forum (
-   Components and Circuits (
-   -   1MHz output impedance query? (

mole42uk 23rd May 2021 7:37 am

1MHz output impedance query?
1 Attachment(s)

I am breadboarding my latest GPSDO using a ublox NEO-7 GPS receiver. I have a very stable phase lock using a 100MHz output from the receiver and a divided 10MHz OCXO clock.

I have used the attached circuit to ptovide a 50Ω output, and I'm experimenting with the square-to-sine converter also shown. What I don't know how to calculate is the output impedance which I would like to be 50Ω.

Would anyone help please?

Radio Wrangler 23rd May 2021 8:03 am

Re: 1MHz output impedance query?
Bipolar LS TTL isn't ideal. It can pull down strongly, but up only weakly, so its output impedance is different in each state. Later CMOS logic has a better output stage and gives a better defined output voltage in the logic 1 state. So it's worth trying 74HC family.

As for the filtered sine output, the RC RC RC network is looking for a much higher load Z than 50 Ohm cable. It will see 50 Ohm coax as yet more capacitance added to that final 1000pF. and if the cable goes into a genuine 50 Ohm load, then the amplitude will fall dramatically.

Yes, it can be calculated by hand by traditional means, but it's a lot easier to just put it into LTspice and then you'll see. You can easily play around with loading values and see their effects.

RC filtering is cheap and easy in some respects but it ain't half lossy, and offers a poor amount of filtering per dB of insertion loss. An LC filter is much less compromised, but then inductors cost a bit more.

You can design filters with unequal input and output impedances, but the load you'd need to put on the LS TTL driving the input of the buffers above and the RC filter makes it a bit extreme. Filters are also transparent, so changes in loading will affect the load it applies to whatever drives it... but here the large loss of the RCRCRC fixes that problem.

It's a compromise, in other words.


Ambientnoise 23rd May 2021 8:39 am

Re: 1MHz output impedance query?
A quick and dirty method practical method would be to load the output with a resistor and note the drop in output level. Output impedance can then be calculated but as David says, the cable capacitance is likely to dominate and the impedance will of course be frequency dependent.

mole42uk 23rd May 2021 9:13 am

Re: 1MHz output impedance query?
Point noted about the use of LS TTL - I have some HC types which I’ll try.

The sine output is only ever going to be 1MHz but I do want a low-impedance so that I can shove anything onto it. Should I use a transistor output driver - that should offer a steady impedance to the RC network, and buffer whatever load gets onto the output?

Radio Wrangler 23rd May 2021 12:13 pm

Re: 1MHz output impedance query?
If you've got the power supply volts, an emitter follower would be good.

If your voltage is limited, then there are video speed opamps which will run off a single 5v supply and drive 50 Ohm lines. 741s need not apply! Have a look in Linear Technology, Analog Devices databooks and websites. TI and Natsemi also do some.

If you do this, I'd also shift to an LC filter and reduce the need for gain.

There's a lot of this sort of circuitry sprinkled around in Horowitz & Hill.


Radio1950 24th May 2021 6:24 am

Re: 1MHz output impedance query?
Hi Richard,

good to hear that you are progressing with that project.
I assure you that it will be a very important "standard" in your workshop that you can rely on.

As Wrangler says, the Zo of that circuit will be variable depending on a host of factors, and AmbientNoise has given a method of measurement.

The basic concept is that if you load up a Source with a test resistor of value equal to the source Zo, then the Source open circuit voltage falls to half value.
Best done at RF with a CRO and 10X probe, and with a load pot to start with, progressing to small resistors with low stray capacitance and inductance, once you have found a ball park figure.

The principle holds true for Nuclear Power Stations and "potato cells", although you would never carry out the actual test of either.

I would suggest you move away from the TTL driver and use a Video Driver chip or discrete amp.

Although it doesnt help you here with your circuit, here's what I did with my NEC PCB GPSDO which has a uBlox Lea 5-T and a 10 MHz output buffered by LVDS differential amps giving a 400mV PP square wave.
It may give you suggestions.

I broke quite a few design and construction rules, but the outcome is a working unit.

To take signals “off GPS PCB board” I just “isolated” them with 0.1 MF ceramic caps and sometimes with 100R resistors, or both C and R.

As suggested by others, I used a DIY 2 pole 10 MHz tuned LC filter to convert the square wave to sine.
I wouldn't bother with an RC filter.

The filter is two 9.5 uH coils on 1cm tubing, 50pF adj parallel tuning caps, top coupled with 22pF, and input and output taps at 10% up the windings for quasi Z match.
This works very well in practice, with a "sine looking" output.

SimSmith simulator gives approximate figures of -25dB and -35dB for second and third harmonics for the filter.
The loss is approx 5dB.

It is very worthwhile to have your GPSDO ref signal as sine, because some monitoring test equipment eg Freq Counters can be tricked by second harmonics etc of square waves.

I aimed for a final ref output of 10MHz and +5dBm50.
I used an amplifier after my 10 MHz filter.

The RF Amp is a cheap LNA PCB ex eBay with SPF51 MMIC type amp, with input and output coupling caps changed to 1000 pF and collector inductor changed to 1 mH approx.
This to modify the response “start” down from 50 MHz to below 1 MHz. The caps are just soldered in parallel with the existing SMDs.
Amp gain is approx 20 dB.
The output Z is a nominal 50R, but in practice, for this type of project, it can be 1-150R or so.
If you actually have a very low Zo, I would insert a 47R resistor in series with the output as protection against a host of unusual "load" scenarios.

My RF Amp has the output of the 10MHz filter signal as an input, and produces an unbalanced 50 ohm output at a nominal level.
The Amp output level is set by slightly mis-tuning the filter output tuned circuit, and watching the wave shape on a CRO for sine/level compromise.
I was lucky with my circuit levels.
In practice I get +7dBm.

An emitter follower is also very useful as a defined Zo output driver, after you set your level.
If you use an emitter follower and an output level of 0dBm, the signal input will have to be 0.25 V RMS and the stage has to handle a swing of approx 700 mV PP.

Is your uBlox ref Freq actually 1.0MHz or 10 MHz please??

Very good luck to you.
Hope some of this helps.

mole42uk 24th May 2021 10:57 am

Re: 1MHz output impedance query?
I'm using a 1MHz output to sync the Marconi 2955 Radio Test Set. Fortunately, that doesn't appear to care too much about the waveform so I'm using the 50Ω square-wave for that. The OCXO is 10MHz which I divide down to 100kHz to phase-sync with the 100kHz down from the ublox receiver. That has been solid on the breadboard for two weeks - I monitor the frequency control voltage at the OCXO and it has been exactly the same every time I've looked.

I do want a 1MHz sine wave-ish output on the front panel for ad-hoc use. I guess that 50Ω again would be useful, especially as all my BNC cables are 50Ω....

G0HZU_JMR 24th May 2021 6:42 pm

Re: 1MHz output impedance query?
At 1MHz I think the output of the triple RC network can be quickly simplified to something based on the final RC network. This means it will look like 1500R in parallel with about 1nF. This means you don't have to do any experiments to try and find the half voltage loading. It would be difficult to do this anyway because the source will be complex.

At 1MHz the series equivalent transformation will be about 16R in series with about -160R reactance and this is 16R in series with about 1nF at 1MHz.

It doesn't look to be a good circuit because it will be very lossy and the source impedance is a long way away from your target of 50R.

mole42uk 25th May 2021 5:56 am

Re: 1MHz output impedance query?
Taking on board the wisdom of you all and discarding the idea of RC filtering, at the moment I am learning the difference between ∏ and T filters at 1MHz and how to use LTspice and the index in Horowitz & Hill.

This GPSDO is becoming quite a project, I have had to clear several others off the bench. It just started as a 1MHz reference for my Marconi 2955.....sigh. At least the original concept is working reliably, it's only the square-to-sine section that's making me learn things I've avoided for 40 years. In school maths I was completely flummoxed by differential calculus and could never find anyone to explain. My maths tutor just said that I could never be an electronics engineer because my maths wasn’t good enough.

Radio Wrangler 25th May 2021 9:11 am

Re: 1MHz output impedance query?

Originally Posted by mole42uk (Post 1377538)
it's only the square-to-sine section that's making me learn things I've avoided for 40 years. In school maths I was completely flummoxed by differential calculus and could never find anyone to explain. My maths tutor just said that I could never be an electronics engineer because my maths wasn’t good enough.

Don't worry. It's not actually difficult. It's just weird and involves a few alien concepts. There is a threshold effect where once you get up to a certain level, you suddenly see how a number of things fit together and all of a sudden everything becomes simpler.

Calculus is rarely taught particularly well. They don't spend much time at the beginning telling you what it is, and then launch into lots of the detailed stuff. You don't see its applicability at the time, so you don't take as much interest as you could. You've also heard that it's difficult and a scary monster.

You're not alone. Some people get out of this trap by teaching themselves when they find they need it.

Calculus is all about things which are changing. Differentiation tells you how quickly something is changing, integration tells you how far something has changed. The bit worst-explained are limits... if you're working out how much change has occurred, then you need to say over what period of change you're talking about.

We live in a world of volts amps and ohms. These are all nice instantaneous things, firmly rooted in the here and now.

Then along came capacitors and inductors. They upset the applecart.

The voltage on a capacitor is directly related to the amount of charge stored in it. This is easy. But the amount of charge has history. To know it, you need to look at the profile of current versus time you've fed into it/out of it since you last knew its voltage. You could keep records, measuring the current very frequently and then doing a book-keeping exercise on it. This is called numerical integration. It won't be exactly right, but it can be right enough if you measure the current frequently enough. But if you go more and more frequent, the answer tends to settle on some value.

If you knew the current versus time fitted some equation, calculus is just a set of rules and some pre-calculated examples you can build together which will tell you what value the numerical integration would converge on.

So calculus does book-keeping at the equation level, not just the data file level.

It's actually a very neat trick.

In a world of sinusoidal and pulse signals, our signals are defined by equations. We have equations for the behaviours of resistors capacitors and inductors, so calculus is EXACTLY the right tool for the job, and inevitably a lot of what we do gets explained in calculus-speak.

There is a level above throwing straight-forward calculus at electronic circuits that involves a stunning new level of weirdness that actually makes things simpler. It involves a ridiculous concept that only a true vandal could have done :-) .... what would happen if we put in imaginary and complex values for frequency? Thinking how that might relate to the real world is mind-blowing, but it simplifies a lot of things into patterns and the human brain is a pattern recognition engine par excellence. We can use those patterns. This is where words like poles and zeroes crop up. They are a neat way of sneaking past differential equations.

So from the point of view of electronics, calculus isn't some scary beast you have to fight past, it's a very useful tool to add to your workshop. You don't need all of it at once, you can just pick up a bit at a time as you need it.


mole42uk 26th May 2021 5:20 am

Re: 1MHz output impedance query?
Thank you David. I’m absolutely in the “teaching themselves when they find they need it” camp - have been for years.

I’ve had a couple of days off the head-scratching but now I’ve checked that I have around 12V available there will be an LC filter with an emitter follower to provide the 1MHz sine output. Radio1950 provides useful values, but I’m interested to make this a design and learning exercise - it really is time I knew how to calculate filters, and to use the arcane dot commands in LCspice.

Although I know about filters in audio - I've worked with them for many years to roll-off, cut or boost above or below certain frequencies, I don't quite understand what sort of filter I need here. I think it's pass-band but a very narrow one for a single frequency and I get that the action of the filter is to 'round off' the edges of the square wave but that's as far as I've understood. Yes, I know how capacitors work and to a certain extent how inductors work, but the inductors I've used and designed in the past have been for SMPS and working at much lower frequencies. I used some in a 459MHz system I was developing years ago but a pal held my had a lot on that project and did all the tuning circuit calculations.

Radio1950 26th May 2021 5:38 am

Re: 1MHz output impedance query?
1 Attachment(s)
If you are using an LC RF Filter for square to sine conversion, can I suggest you use an online calculator.

Sample filter circuit for your situation attached.

This was calculated with RF Tools "LC Filter Design Tool".

I checked the sample circuit with the SimSmith analyser program to be OK, but values can be slightly optimised.

If you go ahead with this and have difficulty with making or testing the inductors, I can make them up for you and post them off FOC.
If you make a different design with different values, I can check the circuit then make up the inductors and post them off.
You may need to tell us the circuitry immediately before and after the filter, so I can adjust the filter source and load impedances for a final design.

If you make up the inductors yourself, I suggest using type 61 ferrite cores in 50 size (one half inch). The core may be marked "FT50-61".

I have not gone into some filter aspects deliberately, but you are in the right forum to ask questions, even very difficult (radio) ones.
I may not be the one to answer however.

Good luck

Radio Wrangler 26th May 2021 7:09 am

Re: 1MHz output impedance query?
You want to pass 1MHz, and reject 2, 3, 4, 5, etc. MHz so you could use a lowpass filter with its 3dB point chosen to be a bit above 1MHz, but a high enough number of meshes to get decent attenuation in place before 2MHz. Getting artful, you can design an 'elliptic' type (aka complete Cauer) filter and plant its nulls on some of your harmonics.

Alternatively you could use a bandpass centred on 1MHz. You can trade off between having a narrow filter of fairly low order, or having a wider filter of higher order. Narrow filters create more insertion loss, and get fiddly to tune. Higher order, wider filters are less touchy, but use more parts.

On-line filter design programmes are a lot crude. They'll design you an elliptic if you ask, but won't let you play games with placing nulls. They're worse at bandpass designs. There are two approaches to designing bandpass filters. One is to design a lowpass (think of it as a bandpass centred on zero hertz, extending to -3dB point at a positive frequency, and one at a negative frequency) and then to transform it into a bandpass filter. These filters are OK only in fairly mild forms. They have two problems... they soon run into trouble requiring extreme component values and they are very intolerant of strays. Many of the on-line thingies only know about transformed lowpass filters.

The other bandpass structure is the coupled resonator style. These are the only approach once they get narrow. They can absorb strays by twiddling component values and are generally more house-trained. Relatively fewer on-line calculators even know this type exist. These programmes seem to have been written by people more from the programming end of the spectrum than from the filter expert end.

Maybe a 2 resonator design set for 1MHz +/- 50kHz pasband would be a good first try.


Radio1950 26th May 2021 8:05 am

Re: 1MHz output impedance query?
I resign

Radio Wrangler 26th May 2021 1:11 pm

Re: 1MHz output impedance query?
Sorry, I didn't mean to discourage you. No need to resign!

I've seen a number of filter designs where people have just used the first filter design web page they've come across and just accepted the values they got. Sometimes the results are comedic, with very large inductors resonating with tiny capacitors (and vice-versa)

At one place I worked, the boss farmed out the design of a proof-of-concept board to a small design consultancy. Their thing was more software than hardware. This receiver needed a few critical filters but no problemo, they got on the internet and got the necessary values. They checked them by simulating the filters and got the wanted responses. No-one had the background to spot that the values were impractical. 60 Mhz resonators made with sub-picofarad capacitors and huge inductors. They found parts with those values and bought them. They weren't aware that the inductors had self-resonance frequencies a long way below where they were to operate. They built a board and there was no signal flow through their filters! We had to bridge them all out just to see if the rest of the design operated.

Mild filter needs can be met by on-line calculators, but they might not steer you to the best approach.

For the OP's needs a 5th order Butterworth designed for -3dB at about 1.3 MHz for 50 Ohms might be fine but would need a 50 Ohm driver in front of it.

By all means have a play on-line, but you can post the design on here so we can check it for any difficulties.

I've spent 40+ years designing RF hardware and done an awful lot of filters, back from when all the calculations had to be done by hand. There are several people on here who can do a reality check for you.


mole42uk 26th May 2021 7:21 pm

Re: 1MHz output impedance query?

Originally Posted by Radio1950 (Post 1377896)
I resign

Please don't do that. This forum has a number of people with specialities and I prefer to hear from several people when I am musing about a particular query. I think there is room here for all of us with all of our various experience and abilities, I hope no-one gets sidelined.

Cruisin Marine 26th May 2021 9:17 pm

Re: 1MHz output impedance query?
Why is a sinewave required, and for what purpose- what will it be driving?

That may help in deciding the cct. design.

A simple NPN emitter follower with a simple pi LPF on the o/p side should suffice for most purposes.

You can (and often is preferable) have different in and out impedance's on the LPF, you can easily have a say 300- 1000 ohm on the emitter and transform that down to be 50 on the output socket or whatever.

Almost any old TTL or CMOS will be fine to drive the emitter follower with appropriate interfacing/matching to turn the the base of the NPN device on and off (>0.6v), if required, the simple potential divider for the base bias should be enough, or even just a base to earth resistor may suffice (I would use about 560 ohm) perhaps with a series resistor from the square wave drive (if needed) all capacitively coupled in with a say 1-10nF or so.

mole42uk 30th May 2021 7:16 am

Re: 1MHz output impedance query?
2 Attachment(s)
Well, I'm getting a little more familiar with LTspice.....but the more I know, the more I realise that I don't know.

So far I have the circuit shown, which gives me the results in the chart. My emitter follower doesn't quite.....

There has to be a way?

Radio Wrangler 30th May 2021 7:21 am

Re: 1MHz output impedance query?
You need some DC bias on the base.

The filtery bit has a load resistor, but it also needs a source impedance. You can spec this in the generator description string/wizard, but I like to draw in a discrete resistor so its damned obvious (else I forget it!)


mole42uk 31st May 2021 10:40 am

Re: 1MHz output impedance query?
2 Attachment(s)
Thanks David, and thanks also to CM.

I have altered the circuit topology and the results in LTspice look better. Please confirm that I haven't gone completely astray? With 2V p-p across the 50Ω load I should be able to drive most things? Should I put a switched attenuator on the output?

All times are GMT. The time now is 3:54 am.

Powered by vBulletin®
Copyright ©2000 - 2021, vBulletin Solutions, Inc.
Copyright ©2002 - 2021, Paul Stenning.