... or until Alzheimer, whichever comes first. You can't win this game, just enjoy it a long as it lasts! 😛
Jan
Output section?
Now that I have at least a sketch of the oscillator circuit it is time for the output section.
Was there ever a consensus reached about output impedance, maximum level, attenuation etc?
Obviously needs differ but what is reasonable best practice?
I am inclined to skip 600 ohm these days.
50 ohm, maybe lower?
Calibrated switch, in dbV. from +10 to -60?
Continuous fine trim, Baxandall style?
Any instruments that people have tried with features that were particularly nice?
Best wishes
David
Now that I have at least a sketch of the oscillator circuit it is time for the output section.
Was there ever a consensus reached about output impedance, maximum level, attenuation etc?
Obviously needs differ but what is reasonable best practice?
I am inclined to skip 600 ohm these days.
50 ohm, maybe lower?
Calibrated switch, in dbV. from +10 to -60?
Continuous fine trim, Baxandall style?
Any instruments that people have tried with features that were particularly nice?
Best wishes
David
Hi David,
A 50 ohm output impedance would seem to be perfect. External resistance can be added to create a 600 ohm output should that source impedance be required. The 50 ohm impedance is about the lowest I can think of that might be needed. It matches the input impedance of most Spectrum and Network Analyzers normally seen on a bench. That should reinforce good signal routing standards. The 50 ohm impedance would also help reduce noise pickup from other sources.
The calibrated switch is a very welcome addition, adding a vernier level trim should cover most needs. Were you thinking of using a switch for signal switching, or relays?
Most people are probably familiar with that type of output level control. It's either that, or direct output level control via a keypad. This would be more like the automated oscillators - so not a bad idea either. This type of control might be best when different units are available. Something like V, mV and -dBv or -dBm into maybe fixed resistances like 50 or 600 ohms. I would think that is getting more complicated for control design. Certainly a micro-processor control for that, although maybe a future retrofit to keep from delaying your work any further.
Maximum output level? May as well copy what the current equipment out there has for maximum. An HP 3325 looks like 10 Vp-p, as is an HP 3324A. Nothing wrong with using industry standard levels.
-Chris
A 50 ohm output impedance would seem to be perfect. External resistance can be added to create a 600 ohm output should that source impedance be required. The 50 ohm impedance is about the lowest I can think of that might be needed. It matches the input impedance of most Spectrum and Network Analyzers normally seen on a bench. That should reinforce good signal routing standards. The 50 ohm impedance would also help reduce noise pickup from other sources.
The calibrated switch is a very welcome addition, adding a vernier level trim should cover most needs. Were you thinking of using a switch for signal switching, or relays?
Most people are probably familiar with that type of output level control. It's either that, or direct output level control via a keypad. This would be more like the automated oscillators - so not a bad idea either. This type of control might be best when different units are available. Something like V, mV and -dBv or -dBm into maybe fixed resistances like 50 or 600 ohms. I would think that is getting more complicated for control design. Certainly a micro-processor control for that, although maybe a future retrofit to keep from delaying your work any further.
Maximum output level? May as well copy what the current equipment out there has for maximum. An HP 3325 looks like 10 Vp-p, as is an HP 3324A. Nothing wrong with using industry standard levels.
-Chris
Read the AP S1/S2 or R&S UPV datasheets to get an idea. R&S has 10 Ohm,
S2 has 20 Ohm.
50 Ohm is too high if you want to measure precise frequency response up to 100kHz.
600 Ohm is important to check that equipment behave well with higher source resistance.
.
Hi Carl
Elma makes this, "UniSelector" model.
Specifically claimed to be low bounce on the switch contacts, perfect for this.
I am inclined to believe them, it looks to be a quality product, based on the brochure information.
I expect they are not cheap but don't have a price yet, anyone here ever priced them?
Best wishes
David.
Typical price of Elma switches of that quality are $65 to $99 USD retail and that's for the simpler switches. Can't find any distributor that stocks the Uni Selector but I guess you could special order it.
There are inexpensive industrial relays that meet or exceed any switch out there.
Consider using relays then you can use any cheap rotary switch to drive them.
There are inexpensive industrial relays that meet or exceed any switch out there.
Consider using relays then you can use any cheap rotary switch to drive them.
For truly accurate frequency response measurements you can either do it like an AC calibrator with remote sensing and a zero Ohm output impedance or 50 Ohms matched. The 50 Ohms matched requires a pretty health amp even if 7V RMS open circuit (3.5 Volts into the load) is the target. Remote sensing combined with a clear current limit is a good option and doesn't need as much drive if it just folds back or trips a disconnect relay (calibrator style).
I don't know of issues using 50 Ohms as a source. I do it at home to 500 MHz and use 75 Ohms to 3 GHz at the office. However you do need to know the losses in each leg or calibrate them out.
Switches are the real roadblock for this stuff. The ones you would want are either custom and need a minimum order of 1K or Mil, really expensive and you can buy one (Daven/Shallco).
The best option may be to use an arduino (BBK, etc.) and relays. Find someone to write the code. Resist the temptation to make it do too much. All you need is frequency and amplitude. Sweeping with relays (or switches) can be done and is how the RF network analysers do it but not simple from the front panel. I think you can get the board with a touchscreen etc. for $30/$40. Since that cheap controller may have an Ethernet port you could make it all a web app. Really less than a switch and it would do everything. Davida's uses a PC and USB to control his which is nice but maybe not necessary. Its astonishing what can change in just a few years with technology.
I don't know of issues using 50 Ohms as a source. I do it at home to 500 MHz and use 75 Ohms to 3 GHz at the office. However you do need to know the losses in each leg or calibrate them out.
Switches are the real roadblock for this stuff. The ones you would want are either custom and need a minimum order of 1K or Mil, really expensive and you can buy one (Daven/Shallco).
The best option may be to use an arduino (BBK, etc.) and relays. Find someone to write the code. Resist the temptation to make it do too much. All you need is frequency and amplitude. Sweeping with relays (or switches) can be done and is how the RF network analysers do it but not simple from the front panel. I think you can get the board with a touchscreen etc. for $30/$40. Since that cheap controller may have an Ethernet port you could make it all a web app. Really less than a switch and it would do everything. Davida's uses a PC and USB to control his which is nice but maybe not necessary. Its astonishing what can change in just a few years with technology.
Wow! That is a lot of money. Switches were probably the most expensive components in my original THD analyzer as well. Relays are relatively cheap unless you use a boatload of them. As you point out, you can then use a cheap 11-position, single-pole ordinary rotary switch on the front panel, for example. Even if you choose to use some discrete CMOS logic to do translation, that may not be too bad.
Cheers,
Bob
AnArduino is overkill. My autoranger uses a PIC16F886 for a whopping buck and a half, the DCX2496 6 channel volume has a 16F84 IIRC for less than a buck. Singles. And free software development environment.
Of course when you go digital control you may start to think: why not use one of those integrated digitally controlled analog ladders like the CS3318 or that thing from Muse. The CS 3318 has 8 (!) channels with over 100dB range in steps of 0.5dB (0.25 if you insist). Use them in the R position on the integrator and you have incredible fine freq control with awsome tracking.
I don't know whether the CS3318 in-build opamp is up to the task, but the Muse is just the ladder sans opamp, though only 2 channel IIRC.
There's a brave new world out there!
Jan
I know. I used it in high-end type of applications and it didn't leave anything to wish for.
Now for a super-duper audio generator, don't know, never tested it for that.
But worth a thought.
Jan
Maxim makes a "resistors and switches" only part also that might be of interest.
Can't recall the part number but give me a minute ...
mlloyd1
Can't recall the part number but give me a minute ...
mlloyd1
The integrated chips may not be quite good enough for -150 dB applications. Its the internal FET's. And the MUSE stuff is expensive. The MUSE guys were adamant that it needed AC coupling, (I could not figure out why, maybe leakage causing DC offsets?)
With that resolution and some smarts a really good precision frequency synthesizer as well as low distortion source. If the amplitude can be made precise it would cover everything. Can a digital pot replace the AGC?
With that resolution and some smarts a really good precision frequency synthesizer as well as low distortion source. If the amplitude can be made precise it would cover everything. Can a digital pot replace the AGC?
timed out 🙁
DS1882 was the part i was trying to remember.
Costs MUCH less than the MUSE part and is easier to get.
Step size is 1dB though.
DS1882 was the part i was trying to remember.
Costs MUCH less than the MUSE part and is easier to get.
Step size is 1dB though.
Yes, the description of the intended applications being car audio, consumer audio/video with no mention of studio applications suggests the part adds significant distortions.
However, if you take one of these devices rated at 10k max, add a series 10k resistor, and parallel this series connection with a 1k resistor, you have something that varies between roughly 900 and 950 ohms, and a step size of (presuming linear steps, which I know it's not, but work with me - maybe someone makes a linear-resistance-steps version) 1/20th the original part.
But then, the adjustment range is rather small and the distortion may still be too high for a -150dB application.
While I'm here, many companies make analog switches and multiplexers that, combined with precision resistors, likely could do the job. I recall using the CMOS 4066 parts run on +/-5V for switching source inputs in a home-made stereo circa 1977, and the specs are horrid (though not bad for what else I had back then). Many parts available now can run on +/-15V and have MUCH better specs (much lower ON resistance, much lower variation in ON resistance with the analog voltage). These parts and matched resistors cost more than the one-piece digital pots, but the performance increase is surely magnitudes greater than the cost increase.
I think there are more linear digital pot options than audio taper pots. They may be better behaved in this application where the pot voltage is moving with respect to the substrate. If there is an easy way to kludge a digital pot onto an oscillator I think a lot could be learned quickly.
In a very early adventure (80-ies) I used asymmetric power supply voltages to integrated switches to straighthen them out. IIRC I even modulated the neg supply with a fraction of the signal to make it very linear.
Soooo long ago - forgot most of the details.
The Jeff Rowland site has a white paper on the tricks they do in the CS3318 (they used its cousin the CS3308 IIRC) and I believe they (Cirrus) also modulated the substrate beneath the switches.
Jan
Soooo long ago - forgot most of the details.
The Jeff Rowland site has a white paper on the tricks they do in the CS3318 (they used its cousin the CS3308 IIRC) and I believe they (Cirrus) also modulated the substrate beneath the switches.
Jan
Thanks for that. Elma is a Swiss company and the Swiss Franc is stratospheric, nor does it seem likely to drop in the foreseeable future.
So, know any affordable alternatives in a rotary switch with a 90 PCB mount?
Otherwise can use a sub-board to rotate the switch or just live with a few wires.
For my baseline 12 position switch with two wafers, that's 24 relays.
So even at $3 each that's $72, can you really buy relays with low distortion "dry" circuit contacts (Au flashed) at much less than that?
Best wishes
David