uTracer (Micro Curve Tracer)

[martin manning]

The uT NXT construction manual is now available so I can start looking into the build.

The communication interface for the uT NXT is virtually identical to that of the uT3. Accordingly, the conversion to use the FTDI USB-TTL should be the same. I plan to omit the MAX232, its socket, and the caps related to it, and just solder the jumper and resistors directly to the PCB.

I'm not running a Windows installation at home, so I will be relying on a future version of Ihor Smal's uTracerJS for the UI.

[martin manning]

Looking more closely at the Grid Loupe accessory first shown in prototype build here: https://ampgarage.com/forum/viewtopic.p ... 63#p403463, and finished PCB here: https://ampgarage.com/forum/viewtopic.p ... 43#p438843
I've concluded that it is not compatible with the uT NXT. The problem is that the cathode in the NXT is at ground potential, where in the uT3 it is at supply voltage (+20V). That means the Loupe does not have the negative voltage that was available in the uT3 to get the roughly -10 to +5 Vg. A new circuit will have to be designed, preferably one that includes a feature to zero the grid current measurement. Recall the original circuit required running through a number of LM348M's to find one with minimal offset.

[Poimandres]

I am considering selling a brand new utracer NXT board that was soldered by Ronald. If anyone has interest please shoot me a pm.

[martin manning]

A couple of updates on the uT NXT just came out. One is that Ihor Smal has adapted his uTracerJS UI for the NXT. This is good news for me since I'm exclusively using Mac OS. The other bit of news is that the first Bach of 50 NXT kits went out. I believe I'm somewhere north of 100 on the list. so it might be a minute before I get mine.

[Corerftech]

Here is another test result from the same tube. I've figured out how to dump the entire data set and get it into Excel (turns out it's easy).

This comparison shows how this EH 6V6 matches up to the GE data at 250V Vg2 and matching Vg1 values (uTracer is limited to eleven curves in one test).

The EH tube's emission is 10% high across the board as previously shown, but here the screen current has been added and plotted along with the anode current. The screen current level matches quite well at the spec point, but the anode current knee and consequently the break in the screen current curve are quite a bit sharper than the original. What this means to tone quality, I don't know.

It will be interesting to try some other brands, some preamp tubes, and some NOS tubes and see how close they come to contemporary published data. Of course I'll have only a small sample size, and the old data was gathered using less sophisticated analog equipment, and it was reduced and hand-plotted by a draughtsman using a french curve...

Dredging up this topic hoping Martin or other will clarify. I would like to capture and dump data to excel, full curve sweeps, for critical comparison. Use Excel to filter data from multiple tubes of same type, in the hundreds count. This of course for matching. With 100 or 200 numbered tubes, it should be relatively easy to filter tightly matched candidates via excel. It would be too laborious to run multiple quick tests and load multiple test points, document and then rifle through the mess. I dread the task and have put off the attempt. I inquired recently about the matter with Ronald and he noted he didn't have a method but rather reach out in the forums.

I have seen the posts surrounding the quoted one and would like to know the procedure. I don't need graphics, just cells of data. It appears Martin Manning has a way to dump data. I use a uT3+ and also a uT6 on the standard gui. I have been researching the JS gui and if that in any way improves data management, I'd gladly move over

This would be an answer to prayer for me.
Thanks in advance for any assistance in the matter.

[martin manning]

Either UI can support what you are wanting to do. The Dekker version writes a .utd file which you can copy and paste into Excel. uT JS also writes a file, and also writes into the terminal window. It won't necessarily be in a format you like or that is convenient for sorting. Full traces take considerable time, maybe 3 minutes or more depending upon the data density.

I suggest you set up a QT for a given tube type and go through your stock. That will screen for good tubes. From those results, select the ones you want to run full traces on.

I would standardize on a 5-minute warm up for each tube and let the uT idle for 30-40 minutes before starting to test them. At that rate testing 100 tubes will take 500 minutes, some time to run the QT and save the data for each tube (30 sec ea or 50 minutes), another 30 sec to remove the DUT and place a new one (another 50 min). Total test time 600 minutes or 10 hours plus 40 minutes uT warm up.

I'm sure you are aware that you will have to let each tube warm up, and the uTracer itself takes a while to stabilize. You can see that in a post above in this thread: https://ampgarage.com/forum/viewtopic.p ... 38#p327038

[Corerftech]

File management was the only real concern in this case. Tubes get burned on a small aging rack for 24 hours under plate current. 10x at a time. Tube removed from aging rack and transferred to uT. The tubes under test are GU50's, NOS, in the first matching phase. Thereafter another hopefully 1000, over time of course, of a few other Russian large lot purchases.

Sorting and other verification completed already, as is coarse matching in larger groups. Initial shorts tests are completed on bastardized bk747 and hickok 800 with custom settings. QT has been completed at a common bias point, 380v/70ma but again that is just one point. Like flying to Montana, lots of paths to get to same point, I need more than one flyer to follow the same path. If I can get some 2 percenters (curve/multipoint) paired up with a max 5 percent pair split in a quad, I'll be really happy. Simple Pairing by single point idle current won't cut it.

When home I'll take a look at the guts of a data file in a text editor. I assumed file would be gibberish contents due to Ronald not suggesting the contents were in plain English. I have probably contemplated the task far more than necessary and complicated it in my head! Was hoping for a simple csv though... LOL

The directive for either UI being ok but JS being less than Dekker version for data handling/ease of manipulation makes it sound like Dekker has the better data package for export. Did I read that right? I have read speed is higher for curves with JS and I'm not time rich but am heavily burdened with the task. Unsure if JS data delay to move data to excel is worthy of its tracing time savings .... Is it a push in this particular case, for time consumed between the differing UI? Test data fast but import slow, or test slow and import faster. No net gain in time perhaps?

The good thing is as fast as I can bang thru tests serialized to tubes, the faster I can post-test crunch data and find matches.

Great insight, thank you. May need to try both systems it seems.

Thank you very much for the guidance.

[martin manning]

Check out the uT3 user manual (uT3 page, top menu bar, under “downloads”) for information on data files and formats. See in Ch 6.
https://www.dos4ever.com/uTracer3/uTracer3_pag0.html

[Corerftech]

Update:
Installed uTracerJS for Mac, I have been using Dekker version GUI on a PC for the V3+.

Would you believe that the JS saves data as a numbers file? Would you believe it is already formatted perfectly for an indexed, serialized list of sequential tests that will allow fast searches of hundreds of identical quick tests?

Wow, my jaw dropped.

I did take a look at the output data for curves, I don't think it will be expedient or helpful without some modified logistics. But the quick test is "quick" and if the bias voltage is altered for back to back runs, example, -5, -10, -15 or if plate voltage is altered for 3 values and run back to back, the data file is amended nicely. Each data amendment can have a separate name (serial#).
Just a huge list of rows, easily parsed.

Quick filters of ( plate voltage min/max), (grid voltage), then (plate current min/max) produces the desired result, or at least narrows the field extremely tight. Actual curve traces can be compared thereafter.

Example:
Va:248-252, Vg:-5, Ia:16-17 (filter)
This will pull all fields between 16 and 17ma from the unformated, basic Mac JS quick test.
Then I can use my eyes to validate two other test points, full the herd for a group and sweep them for full curves.

This JS thing is quick to manipulate too.

Thank you for the assistance. I sure tried hard to use numbers (don't use actual excel except for work at work) using the original Dekker GUI data. I did find one format (can't remember if block or list) devoid of text, that could have worked. Decided to try the JS before going deeper, glad I did.

Lastly, rather thatnsweep curves, using the curve tracer to produce specific markers to reference will likely be even faster. I think it will produce points as a multi-quick test would, I'll try tomorrow night maybe. The concern there is the data is formatted differently.

If a close match exists in a reduced 9 point sweep, those complete curves will be very close in the end.

I like the JS deal very much.

[martin manning]

It has taken ~3-1/2 months, but my uT NXT kit is in the US. UPS collected $48 in import fees, which is just about 19% of the PayPal price, which itself includes a 7% up-charge. In total just under $350 US, delivered.

As a first step, I've downloaded the latest uT JS (v. 0.99, build 20260520) from Ihor's site and tested it on my uT3+. That went fine, and Ihor says he has done all of the incremental testing on a uT NXT build, so I should be able to build one without access to a Windows machine. The only bit that has yet to be proved is the plan to ditch the RS232 interface for a direct TTL connection to the PIC using my current FTDI cable (as described above).

The latest uT JS UI is essentially the same as before. Examples of the Quick Test and a full trace of a 6L6GC using the latest version are shown below. This will be the baseline to compare the NXT against.

[luxpaul]

Hi there.
Only recently discovered the uTracer so managed to get myself on the list for the uTracerNXT.
Now built (thanks to Ronald for assisting me with some issues I had during the testing) and running well it seems.

I am using an old Mac (running High Sierra) so I first of all managed to get the GUI running via installation in a Windows 10 image running in VirtualBox.
Then found out about uTracerJS.
At first it wouldn't run due to my old MacOS version but Ihor very kindly built a suitable version for me and it now works fine.

I should point out that I am totally new to valve testing and still trying to learn more about the whole subject so apologies if I say anything silly.

I am only interested in testing valves for guitar amplifiers, and have so far tested the following types:
EL34 , 6V6 , 5881, EF86, EL84, 12A{X,U,T}7

The results were almost all reasonable with respect to the standard operating characteristics for each valve except for the 12AT7's.
I tested around 6 12AT7's of which 5 had been in use and 1 of which was new.
The plate current for all of them was not too far off the expected values but the transconductance for all of them was very low.
I tested at 250V anode and -2v grid and they were all at around 30-40% of expected value.

Of course it is possible that those are the correct values, but the fact they are all low (including the new one) make me suspect there may be something wrong in the calculated value.
Does anyone have any thoughts/suggestions on this ?
Note that for all the other tubes I tested , the transconductances were much closer to what was expected.

Many thanks
Paul

N.B. My uTracer is not boxed and the 9 pin socket is simply attached directly the the PCB without any ferrite beads.

[martin manning]

As long as the leads from your socket to the uT ASGK terminal block are short, you should be fine. Are you only looking at Quick Test results? If so, try tracing a full set of curves and see if they are well behaved.

[luxpaul]

As long as the leads from your socket to the uT ASGK terminal block are short, you should be fine. Are you only looking at Quick Test results? If so, try tracing a full set of curves and see if they are well behaved.

Hi Martin and thanks for the reply.
The leads are roughly three inches long I would say.
The full set of curves did look well behaved from memory but I will run them again and have a closer look.

Regards
Paul

[luxpaul]

Hi again.
See attached plots for the 12AT7 test.
In fact the Gm is not as low as I suggested.
I think these are just weak(old ?) tubes.

Screen Shot 2026-10-11 at 18.08.54.png

Paul

[martin manning]

The strange looking plate curves may be oscillation. I guess you could try shortening your leads or adding ferrites.