All about electronic valves and hi-fi
After some tests found that initial design of the power supply had some nasty transient response when HT switch was used. Rod Coleman suggested the addition of snubbers and also end up re-arranging the location of the HT SW at the output of the supply. An UF4007 was also added as a protection for any negative spikes at the output due to leak inductance:
Here is a bottom view of the chassis:
Andy Evans built a pre-amp with the 4P1L and was delighted with the sound of it albeit the 4P1L was running below its optimal operating point: 15mA given the limitations of the 126C interstage transformer.
I went to my workshop to test this configuration and looked at biasing 4P1L with fixed bias and driving it with 1Vrms or more to see what the results were:
So here is the first test at Vg=-4V, Va=74V and Ia=15mA
(all tests were done with the 100k input impedance of the Pete Millett Sound Card interface as the secondary load of the OT. 4P1L had both filaments in parallel and If=600mA)
You can see a richer harmonic profile with the OT and distortion is around 0.13% when driven with a 1Vrms providing an expected Vo close to mu (Vo=8Vrms)
The distortion gets very high when output voltage is higher than 9Vrms:
Now if we bias the valve at a more convenient operating point:
We get a slight improvement in THD down to 0.11%. However the distortion above 9Vrms is still high:
So what if we compare the performance of the OT against the CCS?
As we can see from above the distortion is halved. Now if we look at how well this valve could perform if biased in a better operating point, we can see that distortion can be reduce down to 0.03%
Minimum distortion from a CCS (or gyrator) doesn’t mean that it will sound better. Clearly the OT doubles the THD of the CCS equivalent circuit. Gain here is nearly same on both as OT is in 1:1. Only way of judging both is to do a listening test….
(which is what I’m planning to do next)
Today I finished the bench PSU which can deliver up to 600V @ 100mA. Here are some pics of the final build.
[nggallery id=5]
Here is my first test of modified bench supply providing 600Vdc from existing 250V transformer via a voltage multiplier circuit, gyrator filter and passive MOSFET regulator. This circuit was built with components I had around and supply provides 0-600V with less than 5mV ripple thanks to the 53dB attenuation. Maximum current is 100mA continuous but passive regulator has a 200mA limiter built in.
Above is a classic voltage-doubler configuration. I had a 250V secondary, so had to add one of this to get HT to 600V at least. Resistor equalisers have been added to leverage whatever electrolytic I had around and could use here. The raw supply should provide a low ripple of about 1.3Vpp at full load (100mA).
A classic gyrator to simulate an inductor in an LC stage. Attenuation of this stage and the passive regulator which is configured as a capacitor multiplier can provide at least 53dB @ 100Hz. This will bring down the ripple to less than 5mV!
Finally the passive regulator. It has a current limiter set to 200mA. M2 will dissipate large amounts of heat. Just do the maths: dropping circa 600V to whatever voltage is set as the output by the 1M pot. This will be bolted to the chassis C shaped aluminium frame. Probably needs a bigger heatsink as could melt down when delivering large currents at low output voltage. This is a situation we want to avoid. Output will have a slow fuse of around 100mA, but the circuit has a 200mA current protection to avoid killing the MOSFET whilst charging a capacitor,etc.
Continuing with the design of the 4-65A SE amplifier based on M. Koster design, I’m in the process of tweaking the 46 driver to optimise the operating point and provide maximum distortion to drive the demanding 4-65a. Here is the circuit I’m currently working on.
The current 46 driver will be biased at around 25-30mA using filament bias, so Vgk will be around -16 to -17.5V using a 10Ω filament bias resistor array. This will set the 46 at around 185-210V which will give sufficient headroom (i.e. need about 200Vpp max) to drive the 4-65a.
So today I look at varying slightly both anode currents and Vgk to see impact on THD.
So minimum THD is around -17V and Ia=30mA. So if setting the Rod Coleman filament regulator to ensure that Vgk=-17V and the anode gyrator to set anode voltage to ensure Ia is close to bias current would provide the minimum distortion (which is 0.04% in this example). Pa is close to 7W, but looking at the datasheet we can see that maximum Pa is 10W (as the latter 45 version).
So next I need to build a prototype of this driver with filament regulator and gyrator load.
Just got a couple of NOS EIMAC which I will be using in my SE design
So did some test on distortion, transconductance and driving them to +22.22dBu output to check the quality of these two ones.
I used similar test rig as before. At some point will be able to get a proper filament supply for this valve, but for the time being I will continue to use the hum pot and the big electrolytic cap across my old bench power supply which can gently provide the 3.5A for the hungry filaments!
I tested them at the limit of my CCS and bench HT supply which at the moment cannot provide more than 360V @ 100mA.
Transconductance is in the region of 3,800 – 4,000 μmhos.
So biasing the valve at -2.5V and over 90mA of anode current, the harmonic profile looks like this:
Both EIMAC measured about 0.12% THD.
Before breadboarding a pre-amp with these nice OT provided by Thomas Mayer, I decided to simulate some options in LT Spice to see what results I got.
First one is my preferred CX301a thoriated-tungsten DHT. I wired the LL2745 in 5.6:1 step-down configuration. This should provide a low output impedance which is what we want in this configuration where we need to drive the cables to the amplifier with sufficient capability:
Looks very promising. Of course gain will be much lower than a gyrator-based pre-amp, the 01a anode load is optimised providing very low distortion: 0.015% based on my SPICE model with curves taken from real CX301a. Gain is low at 3.1dB, but we don’t want loads of gain in this pre-amp. Sound is what we are after…
Now it’s time for the revered 26. I used Dmitry’s model based on the RCA manual curves. I’d like to simulate this again using a model based on starved filament curves with a real 26.
Originally planned to bias the 26 in a different operating point based on feedback from Andy Evans, however after playing a bit with the OP I found that a more linear point was around Ia=5mA and Va=114V @ Vgk=-6.85V
26 looks more interesting in principle as the output impedance with this model is lower than the CX301a thanks to having a lower Ra (7KΩ against 11KΩ) so LF response will be slightly better in a side by side comparison.
Will be trying these two with filament bias and Rod Coleman’s filament regulators. The HT will be provided through a Salas HV shunt regulator.
Stay tuned…
New Lundahl LL2745 OT pair arrived today in the post. Very excited, Thomas Mayer kindly send me this pair for trial. These are specifically designed for the low current and high anode resistance DHT used in pre-amps (e.g. 26, 01a, CX112, etc.). Wiring and primary details are similar to the LL1689. Hope we can get a final datasheet in September 🙂
These are gapped at 8mA. So the primary inductance is about 200H in theory (160-180H in practice according to Thomas). The labels are incorrect. This is 2×2.8:4×1, providing the option of wiring it as 5.6:1 if both primaries windings are in series and all secondary windings are in parallel.
Will rebuild the 26 or 01a preamps and do some test.
I would like to build a quick breadboard with the 01a to do a comparison against the gyrator loaded preamp I’m using at the moment.
Music, that’s all about….