8647	2013-05-12 18:48:38	r.ewton@att.net	old step attenuator
Message	Date	From	Subject
This is my first post to this group and I want to say I’m just getting started in emrfd, being still in the study and tool building phase. I have been a ham for nearly 60 years. I guess it’s about time I really got into rf. In any case, I’m enjoying the mental workout. My question needs a little preliminary explanation. I have had an old Navy radar test set (TS-107A/TPM-1, for anyone to whom this might mean something) in my garage for over 30 years. Since I am not into UHF, its 500-1500 mHz range let me forget it. But, a few days ago, it occurred to me that it had a step attenuator built in. It has seven 3 dB steps from 3 to 21. I took it out for testing. One amazing thing: The set still had a pair of 6-volt cells. On a whim I stuck the DVM probes on one and, lo and behold!, 5.4 V. No load, of course. And this thing is probably over 60 years old. I can vouch for it just sitting in a series of garages for 30 of those. Wonders of electrochemistry! But, back to the attenuator. I still don’t have rf power measuring capabilities whose accuracy I fully trust, so I tested with DC. I put a 50 ohm resister across the output and fed in exactly 1.000 V and at -3dB read 0.711 V. The calculator says that’s down -2.97 dB, off just -1% from 3. The remaining data sets are: (-6 dB, 0.493 V, -6.15 dB, +2.5%) (- 9 dB, 0.342 V, -9.33 dB, +3.67%--the worst); (-12dB, 0.244 v, -12.3 dB, +2.17%) (-15dB, 0.168 V, 15.5 dB, +3.35%) (-18 dB, 0.124 V, -18.1 dB, +0.78%--the best) (-21dB, 0.086 V, - 21.3 dB, +1.52 %) So I’m appealing for an educated and/or experienced answer to the following: 1) Is this thing worth keeping for use in HF experimentation? Maybe I should use it, but just on the 3, 18 and 21 dB settings where the inaccuracy is probably tolerable. 2) Are the DC inaccuracies likely to translate to the same (or close) inaccuracies when measuring rf? In other words, are these inaccuracies I can “depend on” and factor out? In any case, I’ll keep the thing as a personal museum piece. Just the attenuator weighs some 4 pounds in its massive cast metal enclosure. The switch is heavy and authoritatively child-proof, almost old man proof, with deep detents and impressive loud clicks when switching. Thanks in advance---Ralph, W6DV [Non-text portions of this message have been removed]
8648	2013-05-12 21:18:17	Kerry	Re: old step attenuator
Definitely worth keeping; you can never have enough attenuators, fixed or variable. DC results may not be quite the same as RF results; there are stray capacitances & inductances that won't show-up at DC. Some of those "strays" may even be intentional for frequency compensation purposes. None of your results are "bad". Even fixed attenuators are never X.0000 dB over a frequency range; my good ones, by MA/Com, Texscan and other good companies, show at least 0.2 dB variation over their frequency range. My quality variable attenuators, by Telonic Berkeley, Wavetek and Daven, would have errors similar to, if not greater than, your attenuator; that is the nature of a variable device. One lesson that I think we have all had to learn is that nothing in electronics is "exact"; all components have some tolerance/variation from an exact value and the real question is "How much variation is acceptable in the use to which the component will be put?". There are more variables in your test setup than just the attenuator under test; the "exact" 1.000 volts and the "exact" 50 ohms for instance. I think that your results, all of them, are very good; I would be happy to use that attenuator on any of its settings. For most work the errors are negligible; if you come to a high-accuracy/precision job you simply should know the characteristics of your attenuator and all other instruments in the setup. Your attenuator sounds like something that you would use with pride on both its accuracy and its "feel"; I know that I get a lot of satisfaction from using something like that. Kerry.
8652	2013-05-14 22:31:55	r.ewton@att.net	Re: old step attenuator
Thanks Kerry. You really boosted my pride of ownership. More importantly, though, you caused me to actually think about those numbers. My impulse, when something isn't exactly as desired or expected, is to see what percentage it's off by, a resistor, for example. However, it now occurs to me that a percentage of a logarithm is a fuzzy concept at best, probably even useless. I see now that my worst case was actually at -15 dB where I measured -15.5, off by just 0.5 dB (out of 15), which doesn't sound so bad. All of the others were off by 0.33 dB or less. I'm used to building attenuators for a single fixed attenuation, which I can usually do quite accurately, but with a question in my mind: Is it better to realize the table resistor values within 0.1% or so, even if it means parallelling 2 or 3 resistors, or miss by a greater percentage but use only one resistor. Since I intend to work at HF, I go for the more precise multi-resistor solution. If I were working closer to the daylight end of the spectrum, I know I wouldn't be able to afford the reactance involved in multiple leads, even if quite short. Even at HF, I sometimes wonder if less resistor accuracy with only 1 resistor would be better. I'll try to resolve this experimentally at a later stage. Again, thanks Ralph ________________________________
8653	2013-05-14 23:30:57	Kerry	Re: old step attenuator
Attenuators are not greatly sensitive to variations in resistor values; you can "get away" with a fair bit. But I have a perfectionist streak I suppose; I use resistors in parallel to get the required value. I assume from your reference to "multiple leads" that you refer to leaded resistors. There are, or can be, advantages in using resistors in parallel. Parallel connection reduces the total inductance; this inductance may be present in the leads or in the resistor construction which is usually a spiral in carbon-film and metal-film devices. Parallel connection, however, increases the stray capacitance so it's rather "swings & roundabouts". Good attenuators can be made with SMD resistors; these are usually thick-film types with inductance much lower than that of leaded resistors. I've just finished a test fixture that incorporates impedance-changing attenuators; 50 ohms one end, 12.5 ohms the other; http://i40.tinypic.com/x5yqzt.jpg Here is the "innards; http://i40.tinypic.com/33xyj46.jpg and here is a close-up of one of the attenuators; http://i40.tinypic.com/11b4u8h.jpg Construction is simple; I just cut-out PCBs like this with a sharp knife. This kind of construction will work well to at least 1 GHz. Calculators are useful; I like this one; http://chemandy.com/calculators/matching-pi-attenuator-calculator.htm It will do normal attenuators too; just enter the same Zo for each end. It;s interesting to fiddle with the resistor values to see the effect on attenuation and return loss. This one takes the drudgery out of finding suitable resistor pairs; http://www.coolcircuit.com/tools/resistor_closest/index.php?rslist=E24&rd=1234&cal=Calculate&sub= I measure the value of each resistor before I install it; "just in case". Kerry.