music, guitars & other obsessions

Test Gear

OPT Characterization

Today, I build a simple dufferential instrumentation amplifier consisting of TL072 and INA128. With it, I can now measure the impedance response of the OPT, choke, etc. The task is made simple by using the Syscomp CGR201’s built-in network analyzer function, which is set to sweep from 10Hz to 1MHz, and the voltages across the reference resistor (3.3k) and the primary winding the OPT are fed to Ch 1 and Ch 2 respectively. After exporting the resulting data to Excel, something like the following impedance vs. frequency plot can be plotted. From which, all the parameters of the OPT can be estimated.



Shanghai XinJian Instrument & Equipment Co.


While doing research on the QT-2 curve tracer, I found some interesting tidbits on the factory that made these test gear (and many more).

Originally founded as the Shanghai No. 21 Radio Factory in 1951 by the Shanghai Municipal Governement, Shanghai XinJian Instrument & Equipment Co. (“SHXJ”), became known nationally with its successful development the GT-1 Tube Characteristic Curve Tracer in June 1960, the tracer was capable of measuring the dynamic parameters of vacuum tubes. Mass production began in 1962. [Never seen one in existence yet…]

In 1962, SHXJ launched the GS-5 Tube Parameter Tester, mass production began in September 1966. The tester made use of different punch card to measure the static parameters of many different types of vacuum tubes, both foreign and domestic made. The measurement error was ± 5%. From 1962 to 1990, a total of 34,292 units were produced. [The GS-5 is a direct copy of the LS3-3.]

The early 1960s, the development of semiconductor parameter measurement was in demand. In 1962, the SHXJ began development of an experimental prototype transistor parameter measurement instrument. In 1963, JZC-1 (630-type) Transistor Tester began its production run. Thereafter, the transistor parameter measurement instrument categories expanded rapidly to form the QC and QG series in the early 1970s. These products can measure the cutoff frequency of the small power transistor, including noise figure, power gain, automatic gain control characteristics, etc., but are also capable of measuring the FET’s transconductance, pinch-off voltage, the drain saturation current, the input capacitance, the feedback capacitance and other parameters. The units are capable of measurement frequency range from the low-frequency, high-frequency, UHF, to the microwave band. The JS-7-type transistor parameter tester sales, with an annual output of 18,507 units, accounting for more than half of the total production the entire semiconductor measuring instrument production.

In 1962, SHXJ began development of a dynamic transistor curve tracer. And in 1963, the JT-1 Transistor Characteristics Curve Tracer was put into production. The instrument was not only able of observing the dynamic responses of the transistor under the different conditions, but was also capable of observing the base source voltage for reliable, quantitative and rapid measurements under dynamic operating conditions. By 1978, SHXJ also introduced higher power transistor curve tracers. These products were popular from the mid-1970s to the 1980’s [and some are still in production today such as the QT-2.]

SHXJ also tried to develop testers for solid-state circuits and CRT color picture tubes. In 1970, it released the 905 Solid State Circuit Analyzer, however, the sales were too small, and only a small batch was produced, the production was discontinued after two years. In 1976, SHXJ did produce a color picture tube tester, but only 7 units were made.

Another important product category for SHXJ was the noise tester noise measurement device with the growing domestic demand from the aerospace, communications, radar, and other industries in the 1970s. The noise measurement of high-frequency transistors placed high demand on the proposed noise the coefficient magnitude unified requirements. 1978, SHXJ launch the XO11 Coaxial Thermal Noise Measurement Standard Device, with output stability of 0.003 db, resolution of 0.01 dB, the noise generator can be precisely and continuously adjustable in 0.003 dB increments, was a high-stability, high sensitivity, high precision, high-resolution instruments that compared favorably to international standard in 1979. Saturation diode noise source development began in 1982 and mass production began in 1984, the Q03 Transistor Noise Standard Generator was used by the national standardization board.

By the end of 1990, SHXJ produced 23 types of measurement instruments, with annual production volume of 1819 units

Prizes and Awards.

1964 – Ministry of Machinery Industry New Product Award , Fourth Prize for the GT-2 Tube Characteristics Curve Tracer [which is the direct copy of the venerable Tektronix 570].
1979 – National Metrology Scientific Research Center, Third Prize – XO11 Coaxial Thermal Noise Measurement Device.
1979 – Ministry of Machinery Industry New Product Award, First Prize for JT-1 Transistor Characteristics Curve Tracer [which is the copy of the Tektronix 575].
1980 – Major Scientific Research by Shanghai Factories Shanghai Government., First Prize.
1983 – Ministry of Electronics Industry, High Quality Product for JS-7B Transistor Parameters Tester.
1988 – Ministry of Machinery Industry National Quality Award, Silver Medal for the the XJ4810 Transistor Characteristics Curve Tracer.

Some classic products made by the company:

This slideshow requires JavaScript.

My New Curve Tracer – The XinJian QT-2

After months of debating, I finally bite the bullet and got myself a “real” curve tracer. The funny thing was, QT-2 was not even my first choice, I actually ordered the CA4810A, a relatively modern tracer that has 2V/step on its staircase generator setting, which I thought can allow me to test many small signal tubes without even a grid amplifier.

But when I got the unit, it did not function out of the box, there was no vertical deflection, so I had to return the unit! In hindsight, that was probably the best thing that could have happened, because when I opened the cover and took a peek inside, I did not like what I saw – cheap PCBs, lousy wires/wiring, cheap parts – in general a total mess… I was happy to return the unit and told the seller that I rather get something else.

After getting my refund, I went back to Taobao and start my search again, and lo and behold, there was a QT-2 available for sale right here in Shanghai, so all I had to do was to go down there and pick it up. The benefit of buying locally was that I could check out the unit in person and make sure it worked before paying for it.

But, nothing was as simple as it should – of course, the unit did not work when it was plugged at the shop, all the while the shopkeeper insisted it was working just fine when he tried it before… He spent the next hour or so, valiantly I thought to fix it, and actually found some broken Zeners and transistor on one of the power supply boards, alas, even after he replaced the components, the tracer still did not function properly. So I just had to left empty handed…another total waste of time, I thought.

A few days later, the shopkeeper called back and said “I got the tracer fixed by the factory, it is all ready for you to pick up if you still want it…” Well that was a surprise, I did not think he would bother spending money on repair it, anyway, as I later learned, he paid a factory technician to work on the unit at his home, paying probably very little for the service as they are good friends. Hey whatever to get the job done, the tracer was finally working! So I brought the curve tracer home…

This morning, I got out the tube tracer adapter and hooked it up to the curve tracer, using it simply as a test fixture for the tube. Guess what I saw – the ugly multiple traces that plagued the adapter was back! Now I don’t feel so bad about the adapter, if even factory unit with full metal chassis and proper shielding could not be immuned to the EMI, what chance did a DIY unit without a proper chassis had? [Edit: now that the tracer has a new X-Y driver board, I am no longer seeing the multiple re-traces, perhaps there is still some issues with the DIY adapter after all, will have to investigate later…]

Well all the joy lasted no more than half an hour, when the curves simply vanished without a trace 😉 while I was trying to take some screen shots, at first, I thought some fuse must have blown (as it did in the shop), but no, they were all fine, so WTF? Do I have another lemon on my hands?! When will this saga ever end! So I call the shopkeeper and told him what happened, well he said, “bring it over and let me take a another look…” So off to the shop we went again.

It took all of 3 minutes for he to figure out what was wrong – the X-Y amplifier board – which he had swap out earlier when he was trying to repair the tracer. After he put back the original board, the traces came right back, so that was it. It’s great how things sometime just sort themselves out, wish they would happen more often!

More on the QT-2 Curve Tracer, this is one of the older units on the market, but much to my surprise, I learned that it is still in production today – right here in Shanghai. I think mine dated back to the early 70’s. The construction style is “heavily influenced” by Tektronix, which makes perfect sense, since the first and extremely rare Chinese-made curve tracer, the GT-2 was an exact knockoff of the famed Tektronix 570 tube curve tracer, guess what, made by none other than XinJiang, the factory even won a national award for its achievement!

The QT-2’s design was (I am guessing) based on the Tektronix 575 but built with transistors, instead of tubes. No matter, even with transistors, this thing is a monster, and weights a whopping 50kg! There is no comparison between this unit and the CA4810A mentioned earlier, it’s like comparing Mercedes to Cherry, the difference in construction and parts selection is night and day. Yes, the CA4810 is cleaner, and built with modern parts, but it just look so un-substantial, like so many of the products that we see coming out of China these days.

The fact I paid exactly the same price for them just reinforce the saying “they don’t make them like they use to…” These old gear are worth far more, at least in my book. May be the new one works just as well, or perhaps even better, but who wants to own a POS (once you know what’s under its cover)?

The old QT-2 could use some top to bottom cleaning, the chassis is scuffed up a bit, some pots should be replaced, and it has probably not been calibrated for ages – I will do them all in good time, now I am just glad to own a piece of Chinese electronic history, plus it can trace tube curves too!

This slideshow requires JavaScript.

DIY Valve Curve Tracer Adapter v.2

While working on the grid drive amplifier, I discovered that the plate sockets were arcing to ground when the plate voltage exceeded ~150V, so it’s time for a new panel! In addition, I replaced all the mini-banana jacks with standard size, recessed banana jacks that that are fully insulated from the front panel – now no more arcing!

Here are some pictures of the new front panel with the completed adapter in action:

This slideshow requires JavaScript.

while the front panel did turn out better than the first one, it still has numerous blemishes and defects – some letters did not transfer properly or were rubbed off during the drilling/cleaning/assembly stage. I will try and get one made professionally next time, since I do not think the diy heat-transfer method is ever going to get the kind quality that I would like to have…

While the adapter is working ok, I was never entirely happy with the grid drive amplifier – especially the fact that I have to use large grid stop resistor to prevent parasitic oscillation during the tests. I tried several circuit topologies, layouts, wiring, etc. But none really did the trick… Weeks passed and little progress was made, so I finally decided to go with a power opamp – the design is already proven by Alfredo in his adapter, so I don’t see why not give it try. Instead of making my own board to mount the LME48910 like Alfredo did, I ordered a pre-made PCB driver board from Taobao, so it should take no time to put the whole amp together with some associated parts.

In addition, I will put the rest of the adapter circuit on a PCB – this one has to be DIY and the whole assembly once it is finished in a shielded chassis – that ought to get rid of most if not all the problem that I had with the EMI interference. I am expecting perfectly flat grid drive to the tube and nice, clean curve traces on the scope. We will see if it will turn out that way soon…

Finally got the PCB for the LME49810 yesterday, had to make a bunch of changes to the wiring with jumpers – no trace was cut which is a good thing, in case, I want to use it for a real amplifier one day…

Unfortunately, I messed up the power supply, which may have kill the chip, will need to take it apart and run some tests to see if it is really dead. The part isn’t expensive, it is more of a time-wasted issue, oh well.

I think I should be able to fit everything under the existing footprint, so no modification need to be made to the existing boards or the “chassis” – but it all depends on whether the EMI can be reduced to an acceptable level and if the grid drive signal can be cleaned up with the new grid drive amplifier, but I think I am pretty close to completing this project, finally…

Got the LME power amp working, unfortunately, the result was no better than the simple MOSFET driver. Since the board also took up more space and requires a separate power supply, the ROI simply does not add up. So I put back the MOSFET driver board and wrap up the project – just have to live with the grid drive as is.

After the tribulations of the past few months, I have learned a lot about the inner working of the curve tracer, while the DIY effort did not pay off fully as I had hope, the experience was still worth it (a bit hard on the pocketbook though). But to proper measurements, I decided to purchase an used transistor curve tracer with some mod for tube curve tracing. More to come on my new purchase…

Scope Probe Adapter/Buffer for Soundcard

Since most soundcards have low input impedance, they load down the circuit being measured. The solution is a buffer/amplifier for the soundcard inputs. It’s a very simple circuit using a dual opamp. I built one this morning…

Tube Curve Tracer Adapter

This slideshow requires JavaScript.

After reading the DIY Curve Tracer thread at diyaudio, I decided to build one of these adapters for the oscilloscope for myself. The adapter has a pretty small footprint, using an A4-size PCB that I have laying around, I try to make all the sockets, knobs and switches fit, it is a bit tight, but it should work just fine. For the larger tubes and other tube with odd sockets, they will be hooked up via jumper cables and tested off the board.

Besides the top/front panel that holds all the sockets, pots and switches. The adapter contains:

1) Power Transformer.
2) Power supply board, the on-board power supply provides voltages for the logic circuits and the grid bias. The plate, screen and filament voltages are all supplied by external bench supplies.
3) Grid step generator and driver board.
4) [TBA] gm/THD test rig a al Moglia.

During the October holidays, I made some good progress on building the adapter, of course, there were numerous mis-wirings, in-correct parts that threw me off track, but in the end, I managed to de-bugged it, and may be even improved a bit over the original design…

Anyway, as usual, the biggest problem now remains the mechanical aspect – the front panel and chassis – same old story again – the unit still looks like a high-school science project (may be even worse!) I tried the head-transfer lettering for the front panel, of course, the alignment was off, not only that, the holes for the various connectors also did not line up straight – not surprising, considering I was using a small table-top drill press and just eye-balling the template for the hole locations, a few millimeters off here and there, and that was it…

Before I toss in the towel, I think I will give it one more shot, this time, I will do the heat-transfer lettering first, clear-coat-and-seal, then drill the holes after it dries, hopefully by drill the holes with the texts already in place, I can get a better alignment, also thinking about whipping up a simple jig to help with the hole alignment – I find that even a few millimeter mis-alignment could be easily detected, so getting the holes to line up is a must.

The problem with doing thing this way is that the finished “product” still may look amateurish. But if I spend the money to get either a CNC-mill or send out the front panel/chassis to have them made, I might as well just buy the damn test gear second-hand!

As is, the extra parts that I had to pick up because of the minimum order size and the number of trips that I had to make to the stores locally, probably already exceeded what I could buy without going through all the hassle (of course I do not get the learning experience gained along the way by build them myself, but that’s a separate issue). From a cost-effective angle, it’s at best breaking-even, but much more likely, I am way in the red… That’s why it is called a hobby and not a business 😉

Ran a quick test on the adapter – it worked but the traces look rather bad due to the multiple retraces, need to clean them up first before getting too far along.

Jumbled curves due to retrace

Still trying to clean up the traces, but the adapter is usable at least.

Improved curves, but still have double-trace per curve

I think I finally manged to sort out the issue with double traces, as suggested by the good folks at diyaudio, the problem was with the grid drive – its waveform got distorted as the plate voltage increased. The step voltages were not constant thus creating a loop when the curves were traced. The fix was pretty straight forward – re-biasing the mosfet driver and connecting the source follower to ground instead of 15V, resulted in a more solid grid drive.

Now I just need to clean up the mess of wiring and tidy things up a bit, then I am ready to do some curve tracing… Finally!


Simulation Result

Most of the double-traces are gone, fuzziness is due to line interference which seems to occur between the hours of 8AM-9PM!

Slowly getting there… Most of the loops are gone

Still working on the grid drive amplifier, I did not think such simple circuit would require so much time – there are only a handful of parts, for Christ’s sake… Parasitic oscillation is a real problem here, I think I need to move the driver FETs closer to the grid as the current wiring is bit too long, whatever it is, I need to get to the bottom of it, so I can wrap up the project, which has taken far too long than I had expected.

During the trouble-shooting and experimentation, I discovered that a simple unregulated power supply for the FETs just did not work, even though on paper, that’s all it was required, the transformer was already over-rated for the current draw expected. Anyway, with the unregulated supply I got ripples on the grid steps, which of course caused double traces on the curves.

I also tried a simple DC-DC converter for the drain supply, but it generated too much noise and the display was all distorted. So it appears that the current topology is very sensitive on its power source.

[11/12/2012 #2]
What happened was so strange, I just had to post it – when I tried to trace some curves this afternoon, the current kept shooting straight up, so I thought something must be faulty, mis-wired, or shorted, etc. First, I swap out the INA128, thinking perhaps it was burned out during one of the earlier tests… No that wasn’t it. Could it be a bad tube? Swap the tubes, still the same result, WTF! Took everything apart, and discovered that the plate sockets were shorted to ground! Whoa! How could that happen?! The sockets were isolated from the front panel/ground by plastic washers (top & bottom), and tighten down by nuts, so how can they even move around? But obviously they did somehow!

Based on this totally unexpected event and the fact that the wiring needs to be shorten to reduce parasitic oscillation, I think a new build is called for – new PCBs for the controller, new front panel and layout – the works! I will also take this opportunity to make the unit larger, so it is not so cramped…

After moving the grid drive amplifier onto its own board and closer to the grid sockets, the grid drive signal is much improved – clean and free of line interference artifacts. So now it is a matter of final packaging to lock everything into place. Will also wire up the volts/step selector as well as the gm/THD test rig into place. Still need to pick up a few more parts – momentary switches, on-off-on switches, non-polarized capacitor and may be a small +-70V transformer (for the built-in B+ and B- supply for the grid drive amplifier).

New grid driver amplifier on its board:

Grid driver amplifier board mounted as close to the plate sockets as possible

Nice clean curves (this one of the 6P14P):

6P14P Curves with new grid driver amplifier