26 THD analysis

As I’m proud of my 26 DHT pre-amp and also looking to use this valve as the first stage of my 4-65a SE amplifier shortly, I looked at how linear this valve is.

Unfortunately I don’t have an extensive set of valves of same brands, so albeit I have about 40 valves if this type there is a big mix of different brands and many of them are used ones. Having said that, I think this may be the case of many of you out there, so I think that probably the results of this test may be relevant to you (if you are still reading this post).

So in summary, this is the sample set I tested:

39 valves
12 brands
ST (34 valves) and Globe (5 valves)
Type: 26, 226 and 326
State: used (80%) and NOS (20%)

All DUT were tested with the same test set and operating point:

Operating point: Ia=5.5mA and Vg=-10V
Anode load: Cascoded DN2540 CCS set to Ia
Test signal: 1kHz, Adjusted to obtain Vo=10Vrms (+22.22dBu)
THD analysis: Audiotester via Pete Millett’s interface:
- 32768 sample FFT (2.96Hz resolution)
- 256 averaged windows
- Van Hann window
- THD for H2+H3+H4

So in summary the results showed that you should expect a 26 to have around 0.08% (Std Dev = 0.00047). Here is an histogram showing a summary of the tests done:

Looking at brands, the following average THD ranking was produced:

26 THD by Manufacturer/Brand

Interesting to find National Union leading the chart. I found RCA and Sylvania to be my preferred ones in terms of sound. THD shown above are average of sample sets of 4 valves or more. Although I tested 12 brands in total, some of them were just a pair so they are not a representative subset.

Finally, here is a sample THD of a very nice 26 valve:

I wish I could have a bigger collection of 26 to improve the accuracy of this statistical analysis. Either way you can get a view of what you should expect from this great valve…

6CB5a output valve in triode mode

6CB5a, as many TV power pentodes, is a great valve for a single ended amplifier. Specially now that popular valves of similar anode dissipation are getting really expensive.

Thomas Mayer came up with a great design for this superb TV pentode that wired in triode mode is very linear and can sustain at least 25W of anode dissipation. If you want to check Thomas’ design, have a look at his website which is very interesting.

If you want to look at the specifications, you can download the datasheet here.

When testing this valve on the curve tracer it was evident the fantastic features and linearity at every point.

As a power valve, testing a 6CB5a for harmonic distortion showed THD=0.07% @ Vo=+22.22dBu (10Vrms) at the following operating point:

Ia=40mA
Vg=-50V
Va=253V

Transconductance in triode is high: 6.4 mA/V @ Ia=40mA, Vg=14.6V, Va=100V. In summary:

Gm = 6.4 mA/V
μ = 4.5
ra = 700Ω

Here is a set of triode-strapped curves that you may find very useful if you’re planning to use this valve:

Also you may want to try the following spice model. It can be improved, but I think is an accurate representation of the curves shown above. Please try it and let me know what you think about it…

26 DHT THD (continued)

Looking for the optimal operating point

From an audiophile’s perspective, this is not the right approach to determine the optimal operating point. However, minimum distortion is a good indication of a good starting point for further refinement with your ears.

I have used mostly the 26 DHT with filament bias in the following point:

Vg=-10V
Ia= 5.5mA

26 DHT operating point — 26 DHT THD as a function of Ia

We can see in the diagram above that distortion decreases with the increase of anode current (lower ra and higher gm) and between 6-7mA it’s at its minimum of 0.04-0.05% at full output swing/

As posted previously, is well known that starving the filaments is a good approach to reduce microphony of the valve and THD as well. At the expense of increasing Ra.

26 DHT operating point vs filament current — THD impact of filament starvation

From the picture above we can clearly see that a typical 26 running at 5.5mA and with filaments at the normal level (i.e. 1050mA) can achieve a reduction of distortion of about 0.02% by starving the filaments to 700mA (66%).

I still need to test how this level of starvation will sound on my preamp, but is quite promising…(at least in theory)

12P17L THD analysis

THD analysis for 12P17L in triode-mode (left-handed)

Looking at the THD for the 12P17L in triode mode (left-handed) using a CCS load and driving the input with the TEST SET oscillator to achieve the output at 10Vrms (+22.22dBu) to look at valve’s distortion in particular. This valve is more linear at lower currents (Ia=25mA) rather than an operating point to maximise anode power (e.g. Ia=50mA) as would be in an output stage:

Interesting to see that there is a point where there is a second harmonic cancellation and only H3 component is visible achieving very low THD (circa 1.3%):

12P17L thd3 — Minimum distortion (H2 cancellation)

A compromise point to obtain maximum output power whilst minimising distortion was found to be:

Ia=35mA
Vg=-15.2V
Va=241V
THD=0.13%

This puts the valve under 8.4W anode dissipation. Looking at the specs you can see that anode dissipation is 7.5W and g2 dissipation is 2W.

12P17L 3 — 12P17L THD at maximum output power

I’d rather operate this valve at a lower dissipation point.

12P17L curves and Spice model

A great russian pentode valve similar to 4P1L, but with indirectly heated cathode. You can check the valve specifications here.

I tested transconductance in left-handed triode mode: Gm=9.6mA/V @ Va=150V, Ia=50mA , Vg=-5.6V.

Here are the curves for this valve:

For the ones who want to test the SPICE model here is my take on it:

You can try the model and please let me know your results! You can always email me

Curve tracer finalised

After a long process, here it is. The curve tracer is finalised. It includes the following features:

Valve curve tracer

Sockets: 4 pin, 5 pin, 7 pin, octal, loctal, 9 and compactron (10 and 12 pins)
Anode sweep: 0-330V
Anode current: 100mA (max)
Grid step generator: 8 steps, 0 to -80V and 0 to -5V steps
Grid output for calibration
Oscilloscope: X (x10 attenuation) and Y
Y amplifier:

x1/x10 differential amplifier
1Ω / 10Ω sense resistor
Polarity inversion optional
Image sync adjust (coarse and fine)

Valve transconductance tester & THD meter

Anode current CCS 0-100mA (0.1mA resolution)
Anode voltage 0-600V (1V resolution)
Grid bias: 0 to -80V (0.1V resolution)
Transconductance meter:

0-2,000 μmho scale (1 μmho resolution)
2,000 – 40,000 μmho scale (100 μmho resolution)
Input test signal: 100mVrms @1kHz

THD meter:

Soundcard I/O BNC connectors
CCS load or external load

Valve THD analysis

Measuring triode linearity

Today decided to do a quick distortion test of on a sample of a variety of different valves. All either triodes or triode-strapped pentodes/tetrodes. As per my previous tests, distortion was measured at +22.22dBu (10 Vrms) at the output of the valve in common-cathode mode. Valves were loaded with the CCS I use in my curve tracer. The operating points were quickly optimised at hand, so I’m sure there may be some better operating points for some of the valves below which may improve their overall THD. If you have any suggestions, please let me know!

valve thd analysis table — Tested valves for THD

Need to retake these measures as the soundcard interface got damaged and results are showing significant distortion

Interesting to see in the chart above, that 6e5p and 6C45p are the best ones. This is in line with their reputation as drivers as they are capable of swinging many volts and producing very low distortion. In terms of harmonics I noticed that 6e5P provides a richer H3 and H5 as being a triode-strapped valve, whereas the 6c45p provide a dominant H2.

Also good to see that my favourite 46, 4P1L and 6CB5A (all triode-strapped) are very linear with anode currents of 40mA (with the exception of 46 as I measured THD on a previous operating point used for transconductance measurement). I should retake the 46 and drive it harder, I’m sure it will perform better at higher current.

Surprised with the results of the 6N6P-I. Was expecting this one a bit better, but perhaps it’s the pulse version distortion, so may need to get hold of an 6N6P and compare the results.

Update:
It looks like I blew up Pete Millett’s interface after measuring THD in float mode and exceeding the 10Vrms limit in this mode. Therefore measures such as 26, CX301a and others are not accurate. When testing 26 with my Ferrograph test set it came out to be 0.05%…
Stay tune until I repair the unit!

46 THD analysis

GM tester modified today to add an option to disconnect the CCS bypass capacitor so can drive the valve with an external signal and measure THD from output in common-cathode mode. The input is calibrated to produce 10Vrms (22.22dBu) at the anode and then signal is fed into the PC through the Pete Millet’s interface:

Did some sample tests with a set of lovely globe CX301a achieving THDs from 0.27% to 0.35%.

When looking at a driver valve such as 46 (triode-strapped) got THD values of around 0.05-0.09% for good valves. When picked up the faulty one I had discovered yesterday with the curve tracer, the THD shown to be 0.20% and over 0.35% in the worst one.

Need to retake these measures as the soundcard interface got damaged and results are showing significant distortion

The PC generates a low distortion sine wave which is fed into the valve grid through the input capacitor. This is the same setup used for the transconductance test. The CCS in the anode is unbypassed to ensure the anode signal is not shunted to ground. The output is then taken out through the output PIO capacitor and fed back into the PC input adapter (Pete Millett’s sound card interface). Audiotester is used then to measure THD at 1kHz.

Finishing the curve tracer

Today I did a bit of extra work on the curve tracer with a view of finishing it. It has been a long and painful journey, but I’m reaching the end of it.

The transconductance tester is working perfect. I need to use the following ranges in my true RMS AC voltmeter:

0-2,000 μmho: 100mVrms scale
2,000μmho-50,0000μmho: 1Vrms scale

It’s probably the DC bias which affects the low scale. As an example when testing a 46 in triode mode (see datasheet for details), I tried the following operating point: Vg=-33V, Ia=22mA and the measure should be around 2.35 mVrms over 220mVdc. But in my bench voltmeter, above 17mA in the 46 doesn’t like it and cannot measure it, so need to change scale. I tested low transconductance valves in the lower AC scale such as CX301a, 26, 4P1L, 71a and then using the high AC scale, used 6e5P, 6C45, 6N6P amongst others.

The tracer now has a common-mode mains filter. This was required as at certain times during the day, specially in the evenings when the mains is really noise or my wife is using the microwave oven!, when tracing curves with the 1Ω sensing resistor and low anode currents (e.g. CX301a) then the noise level was sufficient to impact and distort the traced image. With the common-mode mains filter it works brilliantly.

Now need to place bottom plate and standing feet. Job done then and will move to some proper audio work!

Testing the circuit today, I measured 29 46 valves. Ended up discarding two which measured low and then when tested with the tracer found that curves weren’t good at all. Probably electrode misalignment as they weren’t just with low transconductance. Will upload some examples as it’s very interesting to see the difference

Measuring Transconductance (Gm)

GM tester jig circuit

After a bit of work, got the transconductance jig working fine. Made an obvious omission which was not bypassing the CCS. The CCS present a very high impedance in AC to the circuit, therefore not developing the current variation on the measuring resistor. Bypassed by an electrolytic presents a path to ground.