?

Dave’s explanation is correct and concise; I can add some detail for those who are interested, or not familiar.

The first converter in the 8566 is a single diode mixer. A single diode is the simplest implementation of a mixer. In this case, the RF signal spans 2-22 GHz and the desired IF is 321.4 MHz. That makes the required LO signal 321.4 MHz above or below the RF at any particular measurement frequency. The options for generating this LO signal are to either build a 2-22 GHz signal source with sufficient power to drive this mixer, or to use a signal from a YTO, which in those days could not come even close to a 20 GHz span (“those days” is 1970s).

A mixer like this will actually also produce an IF by mixing the RF signal with harmonics of the LO. Its performance won’t be sufficient unless some additional measures are taken. Just looking for a mixing product with a harmonic of the LO will not produce a sufficiently strong IF signal. To improve its efficiency (reduce the conversion loss), a d.c. bias is applied to the diode to slightly forward-bias it. This bias adjusts the conduction angle during which the LO can turn the diode on. By doing that, each harmonic mixing band can be optimized for the lowest possible conversion loss. So, that’s what is done in the 8566 first mixer.

The 0-2.5 GHz band in the 8566 has a different conversion chain and its own mixer. This difference allows for some troubleshooting observations. If the noise floor is excessive everywhere (0-22 GHz), then I would look somewhere after the point where the two signal paths join. If the noise is high only in the microwave bands (2-22 GHz), I would first look for signs that someone tried to fix the problem by making adjustments. If not, I would observe if the problem is apparent on all bands. From there, paths diverge so it would take a lot of work to speculate.

Be aware that the various 8566 front end adjustments are tedious, so don’t try to adjust things unless you are willing to spend a few days doing this and you have enough equipment to do the job.

But first, see if your 8566 is meeting specifications. There should be a few dB margin.

Vladan





On Mon, Jun 6, 2022 at 06:42 PM, Dave McGuire wrote:

These analyzers use several bands to cover their operating frequency range. The stepped noise floor is due to the increased mixer conversion loss in the higher bands. The higher the band, the greater the conversion loss, and so the IF gain is increased to compensate for it, to correct the amplitude. This higher IF gain raises the noise floor.

Vladan, please correct me if I'm wrong on any details here.

-Dave

These analyzers use several bands to cover their operating frequency range. The stepped noise floor is due to the increased mixer conversion loss in the higher bands. The higher the band, the greater the conversion loss, and so the IF gain is increased to compensate for it, to correct the amplitude. This higher IF gain raises the noise floor.

Vladan, please correct me if I'm wrong on any details here.

-Dave

"pianovt via groups.io" <pianovt@...> writes:

Matt, here is what a typical 8566B does. This unit is average, i.e. not particularly "hot", nor has it been calibrated recently.

Thanks for sending that along Vladan. My unit appears to be a bit worse
than that, but not wildly different. Have you noticed an improvement
after calibration? I sort of get the impression it won't do much but I'm
about to perform the "Last Converter Adjustments" so I guess I'll find
out.

Here's what I see on my HP 8566A - definitely a bit better than my B. If
I become determined I may look into swapping some parts between them.

Matt

Interesting. I saw the stepped noise floor also on the recently acquired 8566b I setup recently, and would like to know mire what you find out. It does look strange.
On Mon, Jun 6, 2022, 20:58 Matt Huszagh <huszaghmatt@... <mailto:huszaghmatt@...>> wrote:
I have 2 HP 8566's: an A and B unit. The noise floor on the A unit is
noticeably better than the noise on the B unit. I've attached 2 images
of my B unit corresponding to the 0-2.5 GHz and 2-22 GHz ranges.
A few things stand out to me in comparison with the A unit. The noise
floor in the lower range slopes up slightly, which I don't notice at all
in the A unit.
In the higher frequency range, the noise floor is just overall higher.
What causes this and is there anything I can do to improve it?
I'm currently going through the performance tests and doing the
corresponding adjustments where necessary. I'm hoping this will improve
the situation, maybe someone has a proper understanding of this.
Matt

I'm currently performing the YTO Main Coil driver adjustments (preferred
procedure) for my HP 8566B. This apparently sets the frequency endpoints
of the YIG. Leaving out some details, the procedure basically has you
key in zero spans centered at 2.3 and 6.15 GHz, disconnect the YTO loop
phase detector output and check the frequency of the 1st LO output.

I've performed this procedure probably 2 or 3 times in the past, and
each time I've found it difficult. The documentation says to wait for
the 1st LO output frequency to stabilize when adjusting the trimpots to
get the 2.3 and 6.15 GHz frequencies on mark (within 100 kHz). However,
I've always found the frequency to slowly drift in one direction. This
time I was able to get a pretty stable reading on the 2.3 GHz endpoint,
but the 6.15 GHz endpoint has been steadily drifting up over the last
10+ minutes. This long drift period obviously makes the procedure
difficult because you can't just adjust and check, you have to try to
predict where it will go.

Does anyone have any advice for how to perform this adjustment
effectively? There's an alternate procedure which I'm tempted to try.

Thanks
Matt