A Question for an HP 35660A DSA Guru

[Smoky]

I purchased an HP 35660A Dynamic Signal Analyzer a couple of years ago now and I am not very familiar as to how to use it.

I was wondering if it is able to do a frequency response test on my home audio gear.

The DSA's built-in source has random, periodic chirp, and fixed-sine. It does not have swept-sine.

I'm reading chapter 13 of the HP Getting Starting Guide regarding "data tables."

A data table is a list of X-axis values (frequencies) that you input into the DSA, up to 400 of them.

The DSA then calculates the Y-axis values for every X-axis value programmed into the analyzer.

Is the following test possible:

What if my Siglent SDG1032X waveform generator is set up to sending a 20Hz to 20kHz sine wave sweep at 1v into one channel of my stereo equalizer.

Attached to the output of the equalizer would be an RCA to BNC cable going directly into channel 1 of the DSA.

While the Siglent is repeating its sweep, will the HP 35660A fill in all of the Y-axis values of the X-axis points I have chosen?

I'm wondering if it will post a graph and give me the dB down measurements at each frequency.

Thank you for the help 

[Smoky]

Here are the pages relating to Data Tables from the HP Getting Started Guide:

[Smoky]

I decided to connect the Siglent waveform generator directly to the HP 35660A to see if the analyzer could capture an externally-fed sweep.

The sweep was set to 18-30Hz on the Siglent and I adjusted the sweep time until I could get a clear curve above the peaks on the screen of the analyzer. I then keyed in several frequencies for the data table. I then turned on "Calc."

I realized a few things at this point. With each horizontal division being 5Hz, each peak is 1Hz apart.

I typed in 20.5Hz and 21.5Hz. Since those frequencies did not fall directly on a peak notice how their noise levels dropped. So in the future when entering frequencies into the data table, first use the "marker" tool to jump to each peak and note the frequency first.

I'm not sure if there is a way to connect a line to each peak, but I imagine that's where swept-sine software would take over and allow the marker to ride smoothly along the peaks instead of falling into the valleys.

I can also imagine the DSA's internal signal source would be in tune with its inputs and never allowing a "valley" measurement to occur if it produced its own sweep.

Anywho, I'm going to do some reading about the DSA's internal "chirp" signal source too. I recall a quote where it's actually described as a sweep that's super-fast. Maybe it can be programmed? I realize too that I may be trying to make this DSA do things it wasn't intended to do.

But no matter what comes of this, I can wholeheartedly say that this 35660A is one hell of a machine!

[precaud]

Sounds to me like you're trying to do this with one channel. The 35660A computes freq response as FFT(CH2) / FFT(CH1), and the "periodic chirp" is the stimulus to use. The chirp is actually a digitized sweep from DC to (max frequency) within the time record length. Download the user guide and it should walk you through the setup for a freq response measurement. Generally, you put a tee on the output, and connect one of its terminals to CH 1. Connect the output of your DUT to CH 2. Set levels, trigger to CH 1, and select freq response under the "Meas Data" softkey. Then select how you want to see the results displayed. Use the cursor to walk along the curve to see the values at each freq.

Freq response using the chirp works well unless there is a large amount of phase shift at low freqencies in the DUT. In which case you have to shorten the chirp length so that it is all captured. I'm not sure if the 35660A gives you a 'chirp length' option.

[Smoky]

That's great to hear Precaud.

The equalizer that I'm repairing has a couple of filters, an equalizer, and an electronic crossover built in.

I wanting to see how accurate they are and how well they roll-off.

I'm short on BNC to RCA cables at the moment but that won't stop me from powering-up the 35660A to see what options are offered in the frequency response program.

And I notice that when I reduce the frequency span on this thing the measurements become more accurate. So this should allow me to closely focus on just small snip-its of the frequency range that are affected by the filters.

I know Dave of EEVblog had worked on this machine in the past, is there any chance he knows which board the internal clock battery is located? The date and time are not saved on my 35660A after it's been powered-down. Other than that issue, it passes all of its self-test routines throughout. I inspected the inside of the DSA when I first bought it but I couldn't find a battery on any of the boards.

[precaud]

The battery is on the cpu board, I can't remember which one that is. You should be able to see it looking down into them.

And I notice that when I reduce the frequency span on this thing the measurements become more accurate. So this should allow me to closely focus on just small snip-its of the frequency range that are affected by the filters.

That's how time-domain freq reponse analyzers work. Reducing the freq span lengthens the time record and reduces the freq point spacing. If you've not worked with FFT analyzers before, I highly suggest you spend some time reading the manual to familiarize yourself with the basic concepts. The good news is, HP wrote great user manuals 

[cuebus]

The Battery is located on the top CPU board as precaud said.
Remove the case by taking off the side screws (2 each that hold the handles on too if you have them) and the 3 rear case screws.
Then you need to remove the power supply. Unscrew all of the screws around the perimeter of the supply but leave the 4 in the middle alone. There are also 2 or 3 screws on top and one on the side of the power supply holding it in place. You then lift up the supply and tilt it on it's head in the direction of the CRT and rest it on top of the unit upside down. Then, the top PCB that is revealed has a pull tab that you can use to yank it out. On my unit it's pretty stiff and likes to bind up unless it's pulled very straight. The battery is on that board. It's a 3V lithium AA with solder leads. Mouser has them- part number TL-5104/PBP. Have a new small zip tie handy as well as that's how its held down.

[cuebus]

I'm sorry, I own a 35665A and didn't realize how different these machines when it comes to the 2 CPU boards. It seems the battery is not in the same location on the 35660A. You might still take out that top board and see if there is a battery on the bottom board. If not, then there probably isn't a backup battery on this unit, because those two boards are the only ones associated with the CPU. The 3 main vertical cards are the 2 input channels and the source generator.

[Smoky]

Here's an update with results from the HP 35660A.

I read the chapter on frequency response tests and set-up the cables as outlined in the user's manual. I made some new RG-400 patch cables too.

What I'm repairing is an older stereo equalizer from AudioControl. It has a built-in crossover so that the signal can be split, high and low, and sent to separate amplifiers for the main speakers and the sub-woofers. I wanted to see how well each channel rolled-off the programmed frequency. I read that if there was a difference between the two channels that capacitors in the crossover network could be the culprit.

I could only test one output at a time. In this case, I'm comparing the roll-off of the L & R channel "Low" output signal. Setting up the analyzer was pretty easy. I limited the span to show just the area near the crossover point. I adjusted the scale to 1dB per vertical division. The "source" output was set to periodic chirp and the trigger was set to "source."

I think it works pretty good 

[Smoky]

Here is the response from the L & R channel "Hi" output signals of the equalizer. Notice the difference between the two channels.

The only change I made to the analyzer was to shift the frequency span over a little.

I just tested all eight of the capacitors inside of the equalizer associated with the LM4136CN crossover IC's and found opposing pairs of 22nF capacitors with up to 7% of difference.

I kept the marker at around the -3dB down position for no particular reason other than to show the rough frequency at that point.

But to get the frequency precisely at -3dB down I can easily set up a "data" table.

Both the AudioControl equalizer and the HP analyzer were made in the late 1980's

[precaud]

Looks like you're getting good measurements now, Smoky.