DHT day

DHT in excess

We met yesterday at Andy Evan’s with our friend Tony for an interesting set of tests. Firstly we looked at measuring Andy’s 4P1L SE and PSE amplifiers:

4P1L SE LL1682: a great sounding amp overall which sounded as well as it measured. A must amplifier to listen to!
4P1L PSE (O’Netics OTs). This one particularly revealed an issue with the 4P1L driver configuration as it was running out of steam at about 2W before distortion creeped in. I think Andy will look into fixing this shortly. It also showed a slight dip above 10kHz up until 20kHz which may be attributed to the O’Netics.
4P1L PSE which I nicknamed the “Daemon” as it nearly screwed up my measurement gear due to some nasty grounding issues. We decided to give up on testing the response of this amp after this.

Andy’s speakers are Mark Audio Alpair 10s full range in some standing 23L cabinets. They do sound great with a solid bass and detailed treble. Perhaps a bit higher on the treble, but they are worth every penny.

We set the listening session to rotate the amps as well as the preamplifiers:

4P1L Siberian Gen 3
Andy’s 26 preamp, filament bias, LL1692 step down transformer and Rod Coleman regulators.
Tony’s 30sp with Rod Coleman regulators and depletion FET CCS loads
01a preamp Gen2

In my opinion, I think we can draw a conclusion to the 4P1L-4P1L-4P1L configuration. Perhaps it’s the H3 harmonic profile, but it doesn’t sound good – a bit harsh on the treble. The challenge in my view was that one 4P1L stage driver wasn’t sufficient to bring out to life in full Andy’s system. It forced the DAC to swing higher output levels and didn’t sound as good as with the preamps.

Don’t get me wrong here, all the preamps sounded great, however there were subtle differences which showed that 01a was superior in this setup. The 26 was also outstanding as expected, however the 30sp was slightly thin. The three agreed on the evaluation carried out and we concluded that 01a-4P1L-4P1L was a wining formula!

The 01a brought a level of clarity that it was superior. The piano, bass, snare drums, brass and voices we listen to in detail across various test tracks sounded with a level of detail and delicacy which was unique. This was a surprise to all, as we were expecting a system with 2 stages to be the superior combination. I wish the 4P1L could have a gain of 20! 🙂

Here are some few pictures of the messy DHTs spread around:

Andy’s HT supplies

4P1L Gen3 preamp, 01a Gen2 preamp, Tony’s 30sp preamp and Andy’s 26 preamp

Andy’s 4P1L SE Amp

What a mess on the floor!

First tests on the 01a, 4P1L preamps into Andy’s 4P1L SE Amp

wiring 01a, Slagle’s AVC and 4P1L Preamp into Andy’s 4P1L SE

Andy’s 4P1L PSE

Andy’s 4P1L PSE a.k.a “The daemon”

Thanks Andy for hosting a great day.

Here are Andy’s impressions posted in DIYAudio of our great experience testing the preamps and the 4P1L SE and PSE amplifiers:

“I had a very interesting day today with Ale Moglia and a friend of ours, Tony. We auditioned 4 line stages. Amp was a PSE 4P1L, with 4P1L input (plate choke and FT-3 coupling cap). Speakers were Mark Audio Alpair 10s. Full range, 23 litre infinite baffles. We placed them in this rank order:
1. Ale’s 01A
2. 26 into LL1692A stepdown
3. Ale’s 4P1L
4. 30sp

All DHTs and all sounded good – quite alike in having that DHT sound. I think they were all filament bias. Ale’s 4P1L line stage might have suffered from going into two more stages of 4P1L so may be better into a different amp. In that sense it may not be a definitive test, so the jury is out on that. We have known from past experiments that 3 4P1L stages in a row just don’t sound that good.

The 26 preamp was predictably nice, sweet and detailed. Audibly better than 30sp. Just a bit better in this context than Ale’s 4P1L but not far away.

The star without any doubt was Ale’s 01A preamp. It was just stunning. Quite magical. So if it’s a question of building a line stage, this to my ears supersedes the 4P1L. I never expected this – I’ve built 01A preamps in the past, but this is a very clever circuit. you can find it here:

01a Preamp (Gen2) | Bartola Valves

I do urge you to look seriously at this design. It’s a bit special. It was the only one of the 4 we auditioned that actually sounded better in the system than without it. There is enough gain with just the 2 4P1L stages in the amplifier, which I usually drive straight out of a ES9023 DAC. But adding this stage was a better sound. I never expect 3 stages to sound better than 2, but this did.”

(Andy Evans)

Robustiano (V0.7)

Hacked a simple PCB to build the follower to drive the 4P1L as suggested by Rod. I had to play with the LND150 setting resistor (R4) to achieve the 2mA of idle current. I ended up biasing the 4P1L rather hot at about 11.5W which exceeds the specs. The Q2 VBE was not possible to measure as the Q2 would oscillate I guess when I place the tester lead on Q1 collector and the voltage seems to drop when I try to measure it. Should have added a ferrite bead:

When measuring distortion against frequency, I was keen to see that the follower provided some impact in reducing the HF distortion. For example at 20kHz, THD reduced from 0.96% to 0.59% @1W output power and from 7.84% to 3.52%, that is close to half the distortion I had before:

What is nice to see now is the effect of the follower providing sufficient source current to the 4P1L grid. Above 2.5W, the grid current kicks in and we can see how “Robustiano” can deliver 3W at less than 1% until starts to clip about 3.5W:

I found that if I were to reduce the Rf further and therefore increasing the collector current but obviously exceeding the 4P1L power dissipation too much as collector current was about 45-48mA, the distortion at 20kHz falls significantly. I suspect I should increase the collector current to enable better drive of Q2 due to its Cib (30pF). To keep the current feedback arrangement this could be done by reducing the negative emitter voltage source (V1). Should try this I guess…

Cheers

Ale

Robustiano (Version 0.4)

Finally back home after a long trip and had the opportunity to put the DN2540 at test and try the topologies discussed for the “Schade” feedback 4P1L SE amplifier. So I re-build my test rig and tried the DN2540 and LND150 at various drain currents. It was clearly to see that in order to keep distortion to a minimum, the VDS needs to be greater than 60V to keep the output capacitance of the FET low. Here are the results of the frequency response at nearly maximum output power (Po=2W):

It is interesting to see that the LND150 which has Coss (max) of 3.5pF doesn’t perform much better than the DN2540 which has Coss (max) of about 30pF. Operating points are different for both FETs but the 4P1L is running about the same operating conditions. What is also interesting to verify with this test is that the higher the drain current, the more capability the FET has to drive the 4P1L input (and Coss) capacitance at higher frequencies as the slew rate of the FET is higher.

We can see an interesting improvement from my initial tests at 5mA when drain current was just about 1.5mA. The yellow trace (Id=5mA) shows the best performance of the DN2540. Surely higher drain current will perform better but at a cost as the drain current is part of the OT primary current.

So how do we keep the gain of the FET when increasing the drain current? The natural approach will be to reduce Rf, but this affects the FET gain and the feedback. The alternative is to increase the supply voltage respect to ground. The price we pay here is to increase the cathode resistor and burning the power on it. With -4V as the negative source supply voltage, I had to only reduce RF to 51.5K to set 5mA on the DN2540. The supply power was increased to 350V, the screen (Vg2k) to about 140V (240-98.6V) which is lower than the 150V used before. There is a tad of extra power to extract on the 4P1L but here is close to its maximum dissipation. The Rk is a pair of wirewound 4K7 in parallel.

I think it is now time to try the BJT driver. I suspect that it will need at least 5mA of collector current to get on with the task of the input capacitance of the 4P1L when anode to grid feedback is in place.

cheers

Ale

Robustiano Schade SE Amplifier

Some topologies as discussed in the DIYAudio thread:

Robustiano: 4P1L SE Schade Feedback tests Part I and II

Here are some further tests using automated measurements. Firstly I tried same configuration running the 4P1L at closely 38mA and 160V screen voltage. Second test, I dialled up the screen voltage to 171V and tweaked feedback to get current up to 40mA

Interesting to see that the performance is more triode-like and has a higher THD at lower output levels, however at higher levels the THD is lower. Mainly dominated by H3 and significant rise of H5 given grid current I suspect:

Here is the harmonic distribution of the first test (more pentode like)

Here is the harmonic distribution of the second test (more triode like):

From a frequency response perspective, it performs very good with -3dB from 10Hz to up to 35kHz at 1W tested level:

What surprised me was to find out the increasing distortion at HF. Will this be due to the 4P1L grid capacitance or the DN2540?

4P1L Schade Feedback: “Robustiano” Amplifier

Quick breadboard this evening after watching the world cup match.

Increased screen voltage to 160V and dialed the feedback resistor to achieve minimum distortion without jumping back to 100% pentode mode.

Nice to see 2W at less than 0.5% with a nice harmonic profile:

4P1L – LL2746 driver test

After a recent discussion in the DYI Audio forum about the 4P1L drivers, I decided to do some quick tests on an idea I had around to use a step up transformer (1:4) – 4P1L and step up interstage transformer (1:2) to drive a 300B or similar using the 4P1L in filament bias.

First suspicion is on whether the 4P1L has the grunt to drive a capacitive load which would be a real challenge in a 1:2 step up as load capacitance is multiplied by 4 when impedance is reduced by a factor of N^2=4.

I built a test rig with the 4P1L in filament bias using a 15Ω wire-wound filament resistor and connected the filaments in parallel to obtain easily a nice bias voltage with 650mA of filament current. Also lower Rf will improve the low frequency response as helps keeping low the output impedance:

The valve was biased at Ia=30mA / Va=160V and grid bias is about -10.2V. A 10KΩ resistor was added as a primary Zobel as per recommendation of the datasheet. Then it was replaced by a 25kΩ potentiometer (P1) and the right value was found by looking at the frequency response.

Initial tests showed a very good response at 1kHz with only 0.24% THD @200Vpp output. The gain is approximately 16. The mu of the 4P1L with paralleled filaments is around 8 and lower than when used in series which is approximately 9-10. Albeit the results were promising initially, the real test of this stage is by looking at high frequency response where the capacitance will makes it real pain.

Continue reading “4P1L – LL2746 driver test”

4P1L PSE load line

A pair of 4P1L can be easily matched, so 4P1L PSE is a great cost-effective option to deliver +3W single-ended warm sound in A1. Having investigated filament bias, harmonic content, now is time to look at this configuration in a bit more detail.

I have at hand a nice LL1623/60mA which can be configured to 3KΩ:8Ω. After looking at the loadline here is what I think it should play well to deliver 3W:

Va=250V, Ia=60mA, Vgk=-22.9V
The pair of 4P1L will equate to mu=8, gm=12mA/V and Ra=690Ω
Vg= 41.6 Vpp

A driver with some headroom to provide at least 80Vpp should be fine for this SE amplifier. Without looking at harmonic cancellation, this stage should deliver 3W at about THD=1%. Clearly proven that I will not readapt the 45 for A2 🙂

EDIT – 17th March 2013

Just realised after reading Imzen’s comment that the maximum Pa used is incorrect. 4P1L is a 9W device when triode-connected. So here is the correct loadline for a 5K OT:

As we can see, it’s better to run this valve in PSE as you will get just 1W in SE with 1.5% THD when biased at 220V/40mA…