Power Amplifier - High frequency power loss.

[tec5c]

Okay, after some well needed rest I have realised that I have made a crucial error in my original post. The "reflected" power I mentioned should be the FORWARD power, which was being measured on a power reflection meter and in my sleep deprived stage, I wrote the wrong thing.

More on this once I get to work and have a spare moment.

My apologies for the silly mistake.

[MagicSmoker]

You stated that the only power meter that could be trusted is a Bird 43....

Yeah, I sure did, and in retrospect that was a poor choice of words on my part. I made that recommendation both because I interpreted the OP's misuse of reflected power as a sign he might not normally work with RF PAs and therefore not have the best equipment at his disposal, but also because the Bird 43 is the only wattmeter I am familiar with that isn't total crap.

Consequent to your and G0HZU's criticism of the Bird 43 I agree that it isn't quite the paragon of wattmeters, but merely dismissing my equipment suggestion without providing an alternative to the OP doesn't really help him.

[KJDS]

OK, that's making a lot more sense.

Most amplifiers droop at higher frequency, both in gain and maximum output power. Usually, on a broadband design that needs it, various compensation strategies are used to minimize the droop. Assuming that there didn't used to be a reduction in gain and output power and there is now, then something has changed. A big change, more than about a dB would indicate that there has been a failure somewhere, possibly on a matching or decoupling cap. If the gain is still ok at lower powers then that probably rules out a problem with the input side of the amp. If feedback is used, then it's also possible that something has gone wrong with the feedback circuit, but I hate using feedback unless the amp has to perform at very low frequencies. It's also possible that the active devices have degraded. One of the advantages of PA fault diagnosis is that when something goes wrong, it's usually very obvious where the fault lies, unless it's buried in the active device.

[tec5c]

Are you using a Bird model 43 wattmeter to make these measurements?

No, a Rohde & Schwarz NRT Power Reflection Meter is being used to make the measurements.

Tec5c, are you a tech in the ADF?

No, I work for a Defence contractor.

[tec5c]

Right, so the trainee who was working on this PA was out running errands today so I thought I'd take a look at the faulty PA as I didn't have much else to do.

I found the reason (at least my interpretation) for why the power was dropping at the higher frequency range. Within the diagnostic board there is a BIT test that monitors the collector voltage of the RF transistors on the 4 amplifier boards. It cycles through every ~1.5ms and checks that Vce is not >= 110v. If it is, then it backs off the power/reducing Vce, so come the next cycle (1.5ms) things appear fine so it tries to operate like normal, then the next cycle Vce is back to being > 100v and the whole process is in a continuous loop giving a ringing effect.
I monitored the collector voltage on a scope and sure enough, when you key the exciter you can clearly see the ringing which peaks above 110v.

I suspected some degrading 47nF polycarbonate caps to be the cause (I can't recall their function without the schematic in front of me sorry) so I stuck 1000pF (all I had on hand at the time) in parallel across 1 of 8 suspected caps and sure enough, I noticed some improvement in the ringing.

It was at this time the trainee returned so I left him to replace the caps and go from there. Hopefully this should fix the problem.

Here is 1 of the 4 amplifier boards with the suspected bad caps. Image of the board taken from eBay, so my apologies for the quality.

[MagicSmoker]

...I found the reason (at least my interpretation) for why the power was dropping at the higher frequency range. Within the diagnostic board there is a BIT test that monitors the collector voltage of the RF transistors on the 4 amplifier boards. It cycles through every ~1.5ms and checks that Vce is not >= 110v. If it is, then it backs off the power/reducing Vce, so come the next cycle (1.5ms) things appear fine so it tries to operate like normal, then the next cycle Vce is back to being > 100v and the whole process is in a continuous loop giving a ringing effect.

I monitored the collector voltage on a scope and sure enough, when you key the exciter you can clearly see the ringing which peaks above 110v.

Lost in all the hullabaloo over my Bird 43 recommendation was what I suggested you check in the amplifier itself:

...If so, a decline in power with frequency points to a problem either with the gate biasing network or the drain supply...

P.S. - that R&S power meter looks a very nice.

EDIT: corrected "R.S." to "P.S."

[vk6zgo]

...
Basically, don't trust any RF 'professional' who claims/argues that the Bird 43 is some kind of industry standard benchmark for RF power measurement. It most certainly isn't! However, the myths about this meter will go on and on because there is no shortage of gullible hams/owners/techs who 'believe' the folklore/hype about the capabilities of this crude and awful device.

LOL... fine, the Bird 43 sucks. I last designed an RF power amplifier c. 1997 so probably shouldn't have contributed to this thread. That said, most of you all have wasted your time either dithering over the OP's incorrect use of "reflected power" or my suggestion to use a Bird 43 to make the power measurements. Not helpful at all, in other words. You have a *specific* wattmeter recommendation for the OP to use or are you just here to make ad hominem attacks against me?

I'm really just questioning the design integrity of the Bird 43 wattmeter and I'm also warning anyone (including you) not to trust this meter or any of the claims made about it being an industry standard power meter.

I can give you one example of why this meter can lie. It is designed to be used with a pure sine wave signal source as it uses a diode detector that operates linearly up at the FSD end of the meter.

So it will be prone to huge measurement uncertainty if there are harmonics present. eg a single -20dBc harmonic can give a ballpark +/- 20% uncertainty in power measurement on this meter. A -30dBc harmonic gives approx +/- 6% uncertainty. The uncertainty depends on the phase of the harmonic.

It also has an uncompensated detector that operates closer to square law at lower dial deflections and this means it suffers a significant error when used in cold conditions. Ironically it is sold as a field measuring device but if you read the small print on this meter it does not come with a temperature spec and they only specify performance up at the top end of the dial. This is because the accuracy is awful over temperature for lower dial readings. It's not a serious piece of test equipment.

A Bird meter is all these things,but nevertheless,they have been very widely used for many years in circumstances where allowances can be made for their shortcomings.
In many cases,if the "forward & reflected power"appears the same as that measured by the same instrument upon installation,all will be OK.

In testing a complete transmitter with all the required filtering intact,the Bird is unlikely to be confronted by harmonics of such high levels as -20dB,or even -30dB.

Back in the day,we used to check the matching of loads,antennas & amplifier input circuits with a Polyskop RF scanner/display,
For output power measurement of lower power equipment,we used a HP probe section with a HP410C meter.
High power stuff was tested for output power with a water cooled test load.

[tec5c]

Back in the day,we used to check the matching of loads,antennas & amplifier input circuits with a Polyskop RF scanner/display,

Ahh, there is an old R&S Polyskop in our test equipment storage area. I've never seen the thing even been moved let alone been turned on or better yet, been used! You just gave some insight to an application for it (which I have neglected to do any research on) 

[G0HZU]

In testing a complete transmitter with all the required filtering intact,the Bird is unlikely to be confronted by harmonics of such high levels as -20dB,or even -30dB.

It would be confronted with harmonics that are higher than this if it were to be used to measure an agile broadband PA used for certain types of gov/mil service. When used with an antenna this type of amplifier doesn't have much in the way of filtering across several octaves of bandwidth. eg 20-1000MHz. Due to the frequency agility of the system there isn't time to switch any filters in and out so the amp runs 'dirty' apart from a basic roofing filter up at UHF.

[BennVenn]

There are many couplers and antenna tuners between the amp and the antenna. A great deal of filtering is done before it leaves the whip/array

[vk6zgo]

In testing a complete transmitter with all the required filtering intact,the Bird is unlikely to be confronted by harmonics of such high levels as -20dB,or even -30dB.

It would be confronted with harmonics that are higher than this if it were to be used to measure an agile broadband PA used for certain types of gov/mil service. When used with an antenna this type of amplifier doesn't have much in the way of filtering across several octaves of bandwidth. eg 20-1000MHz. Due to the frequency agility of the system there isn't time to switch any filters in and out so the amp runs 'dirty' apart from a basic roofing filter up at UHF.

Indeed,but that is more of a special case,which definitely supports your contention that the Bird is not a "top end" instrument.
On the other hand,the vast majority of Transmitters which would constitute the "constituency" for the use of Bird meters,have harmonic suppression of 40dB or better,so that source of error would be minimal.