EIN broadband power amp.: attenuator?

[Skywave]

Introduction.
I have three EIN broadband power amplifiers, model 500AP. Their stated B/W is 1.5 → 520 MHz with 27 dB of gain. I am not altogether sure of their input and output impedances, but 50 Ω seems most likely. Each amp. consists of an aluminium box with two BNC connectors, a PSU connection (+24 v.d.c.@ 400 mA demand) and an encapsulated block with seven legs, which is the 'bones' of the amplifier. On test, all three behave the same way (dynamic range 2 µV to 100 mV (approx.) input with 50 Ω load at output; B/W and gain as stated). Broadly, they all meet their stated spec. I say 'broadly' since each one has a strange dynamic frequency-dependent instability. I shan't go into that here: it'll simply over-complicate matters; it is a concern, but to focus on the reason for this thread, I'll leave that problem to another, later post.

This problem.
Between the input BNC and the input leg of the encapsulated amplifier, there is a network of three resistors in the form of a pi attenuator / Z-matching network. The two 'vertical' legs of that atten. are 390 Ω; the common 'top' R is 12 Ω. The d.c. resistance at the input to the encapsulated amp. reads infinity. (Probably due to a d.c blocking cap. inside the encapsulated amp.).

What I don't understand is the values of those Rs in that input atten. Since the two vertical 'legs' are the same, 390 Ω, that implies that the input Z of the encapsulated amp. block is 50 Ω. However, with the resistor values stated, what is the atten. of that network? - and why is it needed? Moreover, for an atten. of 2 dB, by my arithmetic, those two 390 Ω Rs should be 435 Ω, since the 'top' R is 12 Ω. Perhaps the input Z is not 50 Ω

All responses will be welcomed: thank you.

Al.

[G0HZU_JMR]

It's probably there to improve the input match of the amplifier and maybe it improves the stability margin.

It looks like a 2.15dB attenuator to me. The values imply the design impedance is 48R but this is close enough to 50R to declare this as a 2.15dB (50R) attenuator.

[karesz*]

Al,
it must be a 50 Ohm RF-PA System...
Karl

[Skywave]

Karesz*: thanks - yes, that's the one.

Jeremy: "It looks like a 2.15dB attenuator to me". Yes, that's what I calculated. The point you raise over stability could well 'fit in' with my comment about an identified stability problem. So perhaps -2 dB is insufficient attenuation. Time for an investigation along that line. I'll come back later about this.

Al.

[Radio Wrangler]

Your amplifier has a gain and phase shift characteristic in the forwards direction, as you'll know. any mismatch at the load will reflect signal back into the amplifier, with a below 0dB gain and some phase shift. The amplifier isn't perfect, so some signal leaks back through it, travelling backwards. This sees negative dB gain and some more phase shift. The result pops out of the input, and sees any mismatch at the input. The reverse signal gets reflected from whatever is driving the amp, again with negative dB gain and yet more phase shift. This tedious series constitutes a loop. If there is too much gain in the loop (the losses don't take out the amplifier forwarrd gain at all frequencies and there is too much phase shift, then the system can oscillate. It depends on the source and load impedances.

The 2dB pad reduces the forwards gain by 2dB, reduces the reverse gain by 2dB

So the loop gain is cut by 4dB and that can help a lot.

David

[Skywave]

Having completed a raft of home maintenance tasks, I found time to investigate this amplifier further. It seems to misbehave in a manner that I've never seen before in any amplifier of any specific frequency or frequency range and is a bit difficult to explain in words. But I'll try.

The signal source is an HP8640B; the load is a 50 Ω coaxial load, via a BNC 'T' piece to which a X10 'scope probe is plugged in and thus to my Tek. 'scope. As I slowly change the input freq. over the increasing range 22.0 to 23.0 MHz, at approx. 22.2 MHz there is a momentary transient sinewave, superimposed on the proper amplified signal at the output and that transient is about half the amplitude of the 'proper' output. When the input freq. direction is decreased, 23.0 to 22.0 MHz, that transient occurs at 21.8 MHz approx. So there is some hysteresis. But it is not possible to get that transient as a steady output signal at either of those two freqs.

Up to now, I have assumed that this strange behaviour was caused by the amp. under test. But a further exam. indicates that that is not so. When the O/P of the HP sig. gen. is connected to the load and the 'scope probe alone, the described 'fault' is manifest!! So it would appear that there is a strange fault in the HP sig. gen. - which I find difficult to believe . . . . or am I missing something obvious here . . . again?

Al.

[Skywave]

Re: post no. 5: thanks David. There are a few points in your post that I hadn't fully appreciated - until I read what you wrote.

Al.

[Radio Wrangler]

Ah-HA!

The 8640 cavity oscillator tunes 256-512MHz with a leadscrew moving a plunger in the cavity. There is a little (posh quality) pot geared off the tuning knob/leadscrew to see where you're tuned (we'll come back to this)

Lower freq ranges are made using fast ECL 2:1 frequency dividers.

They give square wave outputs.

You want a sinewave, so the 8640 includes a bank of filters.

One filter per divider stage doesn't work. As you tune to the bottom end, the 2nd harmonic would be at the top of the filter's range. So each range has two filter.
The 8640 has to change between the high and low band fllters half way up a band.

It uses the voltage from the little pot to decide when to switch filters. The band you're on is 16-32MHz. Half way in filter design terms (log freq) is SQRT(16*32) =16x1.414 = 22.62MHz

It would be ****** irritating if you wanted the changeover frequency and it sat there switching intermittently, so the comparator on the pot voltage has a bit of hysteresis. The RF filters switch up at one frequency, and down at a slightly lower frequency.

Guess what you found!

David

[Skywave]

Thanks David. I am broadly aware of some of what you have stated as to how the 8640 works - but not the switching between filters. It's actually quite surprising that after all these years of using an 8640 at home and during employment, that to me, this 'problem' has only just come to light.

Originally Posted by Skywave

So it would appear that there is a strange fault in the HP sig. gen. - which I find difficult to believe . . . . or am I missing something obvious here . . . again?

Well, now that all has been explained, it was certainly not obvious to me!

However, from HP's point of view, I can see that what I have found would not have been considered a problem by the designers, since the 'default' application of the 8640 (or any non-sweeping S.G. come to that) is to produce a specified signal of known freq., known amplitude and with known modulation (including none). It was only because my need was to slowly manually scan a range of freqs. and look at the resulting signal level from an amplifying chain that I discovered this 'feature'. (If I owned a scalar spectrum analyzer with its tracking generator, I would have used that instead, of course). The good news for me is that now I know that this 'problem' is not in the circuits that are under my design control, I can make 'allowances': i.e. ignore this effect. But it certainly had me scratching my head for quite a while, with the bits that didn't include the S.G. heavily falling under suspicion.

So to conclude: there is clearly a moral here (I'm tempted to say that that moral is "Assume nothing" - but that's not exactly practical is it?), even if that moral is applicable to expensive, highly-regarded test kit.

Al.

[mhennessy]

The moral is "know your gear". Though "assume nothing and trust no-one" is also pretty reliable

"Knowing your gear" is much easier when you have a modest amount. For many years, I could count my collection on just two hands, and really did know every item intimately. But thanks to forums like this, not to mention eBay and other s/h sources, I'm painfully aware that today I have far too much test gear, and as a result I don't know the individual pieces as well as I knew the older gear. When I'm using the newer (to me) stuff, I'm simultaneously checking/validating the gear as well as the task in hand, and that sometimes does get in the way somewhat.

For that reason, I still prefer to use my original items for routine, day-to-day stuff, despite having much "better" options - it's just more productive. The main example being my Telequipment DM63, which is my least-good 'scope in terms of specification - by quite some margin, on paper at least - but it's an old friend

In my experience, these sorts of things are more likely to happen with the high-end gear. For example, my two favourite function generators are my Wavetek 162, which I've had for nearly 30 years, and a HP 3314A, which is a much more recent acquisition. The former is completely "mechanical", with a big switch to choose the frequency decade, and another big switch to control the output attenuator. The HP is completely automatic - frequency and amplitude is controlled by a single rotary encoder. Obviously, there will be range-changing internally (switching capacitors, and switching the attenuator), but all of that is "abstracted" from the user. As a result, you do indeed get glitches as you adjust the frequency and amplitude. All that sophistication does come at a cost, but it enables remote control and other functionality, so ultimately, it's a fair compromise. The upshot is that for most tasks, the Wavetek is actually the better instrument. But the HP makes better waveforms and has a number of other tricks up its sleeve, so it gets used when appropriate. Frankly, it would get used a lot more if it wasn't for the awfully noisy cooling fan (same comments apply about my HP 8656A) - I really must investigate options for modern substitutes.

Many thanks for the explanation, David. I don't have an 8640, but it's still really interesting to learn how they tick. Noisy fans aside, I've yet to find a poor bit of HP gear

[Radio Wrangler]

3314A, yes a competent function generator with a user interface like having a conversation with the world's worst pedant. The Wavetek is a lot more human-compatible.

The 3325A is easier to drive, a lot larger and has the "analogue phase interpolation" style fractional-N synthesiser in it. A bit of a devil to get your head round, that one! I got landed with redoing the synth in that one to use a new generation of LSI.


Expect the unexpected!

David