A few years back I went down the path of exploring different DHT in pre-amp stages for sonic improvements on detail and dynamics. I explored all sorts of different DHTs as have posted in the blog and many more I never got around to write up about.
An interesting brick-wall was hit when testing DHTs with cathode degeneration resistors of considerable size (e.g. higher than what normally is used under filament bias topology) which puts at its limit the traditional filament circuits which are based on CCSs and do want to see a low impedance to GND.
When working with my beloved RE084, I found this issue. Tried the Rod Coleman and Tentlab regulators without success. Below is an extract of my e-mail exchanges with Rod Coleman and Guido Tent.
The issue
Back in 2018, I built the below pre-amp. This needed a 470Ω degeneration cathode resistor:
To my dissapointment, the filament regulator based on CCS, didn’t work as expected with this high ra triode. The LF pole was up to above 20Hz and the HF just to 60kHz. Distortion was above 0.2% for just 2Vrms output:
Now, if you remove the CCS filament regulators and replace them with a low impedance ones (e.g. a bench DC power supply), things look very different.
You can see that the distortion is down to 0.03% for 20Vpp which is improved thanks to the degeneration of the cathode resistor.
Scratching my head, I switched back to the old filament bias, which obviously worked fine:
You can see below, the response is really good (LF and HF). If you remove the Source Follower still is good up to 100kHz or more. When loaded with 100K the LF pole moves up obviously to around 10Hz, which is what is expected.
The distortion profile was really good, below 0.01% for 10Vrms, great performance. It actually sounded very very nice indeed!
So, it was a clear issue on the regulators. I moved back to try Tentlabs modules which are very good. To my surprise, result was the same.
So, had to rig up a test setup on the bench to measure noise the best I could. Here is what I’ve got:
I used an LDO regulator I designed for the RIAA stage which has very low impedance and also a slow startup. They sound very good, and they work very well in this setup.
Despite the Tentlabs have a slightly better performance than the Rod Coleman (in the HF band above 1kHz), I could hear a “hiss” which I didn’t hear with the Rod Coleman regulators. The quietest setup was with the LDO regulators.
I fed this back to both Rod and Guido back into 2018. Cleary, I pushed their regulators into the unexplored territory. Fine, they work brilliantly in other topologies, but not on this one. Obviously for more traditional designs (Choke, resistor or transformer load) you wouldn’t hit this issue as you wouldn’t look to implement degenerated cathode bias with a high Ra triode! Only it’s possible under the mu-follower topology, which I obviously have drained out like not many out there 🙂 – BTW: this was JC Morrison’s comment at the last ETF!
So, what happened next?
I obviously used my LDO regulators when wanted to implement this topology. Recently I posted this in which you would struggle to keep them quiet enough with the traditional regulators, unless you do what I did.
There is some light at the end of the tunnel
Now, fast-forward to 2021, Rod came back to me on this topic to my surprise. Here is an excerpt from our email exchange:
[..] I'm returning to this email from 3 years ago (my apologies for the slow work!) - because I think I have found out what is going on with Low-current filaments and my regulator. I was looking at some nice old Tungsram (Tottenham factory) LP2s with a 200mA current filament, and built up a 150V stage to test them. That's when I confirmed the excessive spur at 50Hz in the spectrum. I started by measuring the same circuit that you used to test the RE084 with Guido's regulators, and compared to mine, and your voltage regulator (the schematic and noise spectrum attached to your message of 31 March 2018). with my V8 regulator, the 50Hz spur was at ca. 100µV - just the same as you measured. Guido's measured the same - because it has the same problems. It turns out that using a current regulator in the negative line means that you have weak coupling between the Raw DC negative, and the system ground, that allows noise to get coupled in from the transformer, or from the circuit wiring, and it has no means of suppressing it. The noise then gets pushed through the positive side of the filament. This also causes higher distortion, since the filament is bouncing around with any noise it picks up... Identifying the suspect quickly suggests a solution! So I re-designed the regulator to allow the negative side to have low impedance, and the current regulator moves to the positive side. At the same time I added some nice upgrades that I found in the development of the Transmitter regulators. Of course, this is the same polarity that your voltage regulator has - hence no noise problems. It's not the differential impedance of the regulator that matters - it's the ground path. With this new V9 prototype, the new regulator measured 40dB better, or ca. 300nV at 50Hz (instead of 30-100µV for V8, or Tentlabs - it can get worse, depending on where the safety earth is connected). The attached spectrum shows the new results - between V8 and V9, connected to a LP2 filament, and 27R bias resistor. The spectra are calibrated for 0dB = 700µV - it's a high-gain Mic amp. Same circuit as your RE084 fil. bias testing. The V9 is sounding better than the V8 - because of the circuit improvements, and solving this problem no doubt adds even more to it. With filament bias it will be quite an improvement in the sound, since even without hum, there is some corruption of the anode current. I am hoping it will also fix the problem with unbypassed cathode resistors - but I have not done full tests yet. [...] Rod Coleman / August 2021
Here are the two FFT plots Rod sent me:
Well, I’m excited to hear that there are some V9 regulators I can test on this setup. Could be a great improvement from Rod Coleman, great efforts from his side indeed!
