DSI 5600A frequency counter noise ...

[eccerr0r]

Anyone have a old ancient DSI 5600A frequency counter (late 1970s)?

Just wondering if you see significant noise on the readout now in 2023, and if you replaced the caps in it.

I had huge amount of noise in the readout but after replacing the caps it went down a lot - but not zero... Still trying to figure out how t get the noise down when right now it looks to me to be a design flaw...

[redwire]

What are you powering it from? It looks like a really good freq. counter. 9 digits I don't have one but a few similar ones. FND500's take me back.
The LED display multiplexing does make noise on the 5V rail and there is little for capacitance 47uF and not sure about the OCXO voltage-doubler 9V rail.

What I've done is poke around with a scope to find noisy power or signal. Upsizing to say 220uF can lower the display jitter due to noise on 5V.
The input JFET is protected but maybe put in a sine wave and see if it looks OK, the signal is not distorted. The 10MHz IN has no protection diodes.
I would've replaced all the electrolytics, upsized a few select ones, given it's from 1979.

[eccerr0r]

Powering from multiple sources from a linear to batteries, I ruled out PSU noise. Tried upsizing but still problematic. Just wondered if it was like this when it was new.

I sort of like my Monsanto better because it has adjustable trigger and can do A-B timing, but this frequency counter is cute as it appears it has much higher sensitivity, I can put it next to my transmitters and it shows the transmit frequency!

[redwire]

The better quality freq. counters have a JFET front-end and can pick up a lot with just a piece of wire in the BNC, they need a few mV of signal only. They also need a x1/x10 attenuator switch because they get overloaded with stronger signals- What are you putting into this for signal? It might just be too strong which makes the readings noisy on the Direct Input.

I was talking about noise after the 7805, on its 5V output that ends up in all the freq. counter IC's.
Most noise on incoming power gets cleaned up by the 7805 and of course with batteries there is no hum and ripple coming in but the LED and maybe even the oven heater can also pollute 5V output if the oven heater is pulsing or oscillating.

I had a Mark V Electronics SM-100 8-digit that I had to revamp because it was a cheap kit that (wrongly) copied a bunch of other designs as a kludge and the front-end was lousy. It uses ICM7226 and I upgraded with a TCXO. It will display 3.579545x MHz yet the cheap chinese one I tried PLJ-8LED was missing digits lol it showed 3.57954 and wtf I haven't found a way to increase gate-time if possible and get muh digits.

[eccerr0r]

Well, you asked what it was being powered from - and it's clean. I'm pretty sure the problem is post 7805 but the output of the IC also looks fairly clean.

My input is... shorted to ground, I don't get a 0 count so it's all noise. This should be a design defect as far as I can tell, not a degradation issue though it may well be exasperated by it.

I also recently got a ICM7226 eval kit, had no front end. Thinking about adding one for the heck of it. Just so happened to also get a SP4541 prescaler that I'd need to count UHF, though not sure if it's as sensitive as the one in the DSI5600A. The ICM7226 can do A-B so this should be able to do what my Monsanto can do, if it had input conditioning.

Will still need to calibrate these ... Tried my best to calibrate my monsanto but since it doesn't have a OCXO and I don't think it has a TCXO, it probably is as good as it gets. This 5600A with its OCXO may well be a better instrument to calibrate.

for the CC frequency counter you'd need a fairly long gate time to see that many digits, whether that's acceptable or not is the question. Sounds like 0.1s gate is what you need to get 3.57954 MHz from a colorburst crystal, though I'd think that the CCFC's probably use a microcontroller to do counting and thus require a software hack to increase gate time.

[redwire]

Non-zero counts with input shorted, I would look for noise in the front-end and prescaler. I don't think it's a design issue. There are three transistors. I use an audio signal tracer to listen for popcorn noise or AM radio, hum etc. because a scope trace can be hard to figure out.

For calibration I had access to a 53131A and then set my old MM5314 digital clock TCXO 3.579545MHz and it's a few seconds a month at worst drift in hot weather.
Building a GPSDO is next on the list. I have a cheap NEO6M GPS with 1PPS but need a reference oscillator module VCO etc I think and not feeling rich right now.
I flipped out in a grocery store buying two bell peppers- price was over $6, $4.50lb or something like that, out of Mexico. Ridiculous.

[eccerr0r]

unfortunately no hp 53131a here and likely wouldn't even be calibrated either lol. Still looking for a cheap, reliable, precise way to calibrate.

Actually I do wonder those colorburst crystals, are they all 1PPM or better crystals? I never thought about that but it seem it would be nice if they were indeed very accurate. Unlike say bulk 10.000 MHz crystals that may be used for computers where close enough is close enough, 100PPM is fine...

[redwire]

As I remember, TV colour burst crystals had to be accurate in tube TV's but not in solid-state sets which use a PLL and lock in to the broadcast chroma. Mine is a huge can HC-6/HC-33 from a 60's TV lol.

Old school crystal oscillator design was to use a medium temperature drift crystal paired with tuning caps that had a big tempco like N750 or N220, or trimmer cap like that, to compensate.
But this is a lost art and took a lot of trial and error, no fun to see if you got it right.

It's two things - temperature drift and aging. When they say "10ppm" or 0.001% there is still the parabolic freq. vs. temperature curve.
Everyone went to an oven design or computer trimmed VCO. I only have a time clock that I've watched over years compared to WWV.

[eccerr0r]

Well, it's not just aging and temperature coefficient, but there is the tolerance out manufacturing. I have a pile of 10MHz crystal oscillators and they all show different numbers on a frequency counter, so who knows which is right, just like a pile of resistors and a dmm that's not calibrated. So of course the frequency counter itself is the dmm and calibration is needed.

Though the Monsanto frequency counter I have is probably 1960s, I wonder what temperature compensation it has if any. Probably should see what error I get if I heat the time base...

[eccerr0r]

LOL I just tried sticking in a fake Nichicon 3300µF capacitor on one of the rails. It didn't completely fix the problem but got noise counts down to 1 - close enough to 0!

This is clearly not the proper solution IMHO as the original design never had a capacitor this large in it, nowhere near it (largest on board is only 47µF, and the power supply bulk filter cap was even just 1mF) so something else is wrong ...

Swapped out two more 22µF caps with 17Ω ESR. For a small cap it's not that bad but have a few that are better -- but the symptoms remained...

[megaraider]

Originally Posted by eccerr0r

LOL I just tried sticking in a fake Nichicon 3300µF capacitor on one of the rails. It didn't completely fix the problem but got noise counts down to 1 - close enough to 0!

This is clearly not the proper solution IMHO as the original design never had a capacitor this large in it, nowhere near it (largest on board is only 47µF, and the power supply bulk filter cap was even just 1mF) so something else is wrong ...

I'm sure you already seen the schematic ...

Most likely 3300µF cap will cause problems to the 7805 because the in rush current will be to high...!
Either way if you connected it to the 9V power rail you'll know who to blame. If it was connected to the 5V power rail it's anyone guess...

Don't overlook those multi contact switches, they are often, specially over time, a source of problems.

[eccerr0r]

Nah 7805's are almost indestructible, only thing that kills them are too high input voltage and shorting the *input* without protection ... But yes I connected it on the 5V rail.

[redwire]

The freq. counter has a few 5V rails - usually you have a noisy 5V and a quiet 5V.
+5VA, +5VB, +DIR. I consider the LED display board a major polluter and add 220-470uF to the 7805 output, although there are almost nothing for 0.1uF caps for HF noise coverage in the unit. The 47uF parts don't do much in the MHz.
I would try add a 0.1uF here and there to see if it's decoupling that is the problem. Upsizing the 47uF parts will cause the switches to suffer and maybe they are the problem- power is through high resistance for some sections?

[eccerr0r]

What's bothersome is that if it was a local decoupling problem, this is a design defect and should have shown up after the person hand soldered the device for the first time.

shorting the power switch while on (to reduce the switch resistance further) didn't seem to help either...hmm...

[redwire]

Are the smaller caps 22uF, 47uF etc. tantalum? Because those do much better at high frequencies than old electrolytics. They might have later cheaped out in production.
It's no biggy to tack a few SMT MLCC 0.1uF or bigger 1-10uF here and there, I've done it on some older gear and completely stopped noise problems.
They didn't always have the room for more caps for some builds. But 1979 was a bit large for the whole thing?

[eccerr0r]

They are aluminium and I'm not going to solder wire extensions to get smt MLCCs on lol.
I think I may try adding some inductors in places but they too need to be fairly large, at least in the mH range and not microhenries, which I thought would have been the proper solution to begin with, IMHO.

[redwire]

It's pretty much the tech for the late 1970's. Sabtronics 8000, 8610b is similar for the power section - a few caps.
I meant soldering a say 0805 ceramic cap across the electrolytic's pads, on the solder side.
I've only seen an inductor used on the prescaler section.

[Th3_uN1Qu3]

Originally Posted by eccerr0r

What's bothersome is that if it was a local decoupling problem, this is a design defect and should have shown up after the person hand soldered the device for the first time.

shorting the power switch while on (to reduce the switch resistance further) didn't seem to help either...hmm...

It is not a local decoupling problem, it is a problem that affects even us humans - an old age problem.

3300uF on the OUTPUT of the 7805, you say? Yes, 7805s are almost indestructible, but they DO age. Try replacing the 7805 with a brand new one then test with your large output capacitor and also without it. I think your problem is worn out silicon in the regulator, not anything to do with the filters.

[redwire]

Hi-fi DACs commonly put huge 1,000-3,300uF caps across the output of LM317's or 781x voltage regulators, to get less ripple and noise. Once I put 10,000uF on 7815/7915 outputs on a recording studio mixer to get it below 2mVpp. The engineer could hear 5mV and complained about it. Don't forget the input-output diode though if you're doing this. Grounding had to be single point too.

It could be the OCXO heater is stuck on or oscillating? They are around 1-2W running and twice that when cold, which would stress the 7805. If a cap in the module is bad, the heater (current) could be making a nice square wave.

Poking around with a scope would be worth it, I'm not 100% sold the original design is this bad to give phantom counts.

[eccerr0r]

I don't think the ocxo heater runs that fast in pwm, and it's primarily using the 5V for heating. The 9v is for op amp bias (hooray for missing schematics!) since RRIO op amps in 1979 were not available.

In any case, it's the display noise most likely. Now why is it not being attenuated (anymore) is the question... BTW 470µF, 1mF were not sufficient to quash the noise, the 3.3mF was the smallest I tried that was successful, and only in certain spots on the board!