There are definitely some tricks that you can play to quiet down an otherwise noisy SMPS, things that I have picked up in the course of a long career. Caging the beast is always a good start, as well as applying a grounded "belly band" of copper foil to reduce the racket radiated from the SMPS transformer.
Speaking of harvesting whale blubber, I'm working at melting away a bit of my own - perhaps not that hard dong power walks with 5lb wrist weights in the sweltering heat we're experiencing on the West coast (100F today). The blubber may be due to some overindulgence to the type of Vitamin B that's fizzy and comes in bottles...
Layout is a big factor in SMPS EMI/EMC problems in my experience of designing Functional Safety Industrial Control subsystems. I never did any high power SMPS, mostly below 10W, but they had to pass world wide agency approvals including TUV/SUD. EMI was always a hair pulling experience. EMC not so much. Although I have seen sensitivities up to 2GHz on some regulators that were layout/component value sensitive. Feedbnack loops need to be kept as short as possible to prevent noise pickup upsetting stability. Layout is a "Black Art". High current loops likewise. Fortunately, a lot of IC manufacturers have sample layouts, and app notes with proposed layouts and advice for issues.
I'm taking the lazy route and using DC-DC converters of my own design, so that all the safety and EMI concerns are stuck with the adapter that's powering the DC-DC. If the adapter is made by a reliable firm (someone like Delta or Meanwell or Phihong), they are generally pretty good about addressing safety and EMI concerns in their adapter designs. I tend to avoid the scary Chinese fly-by-nights, which can hide a host of horrors inside. The advantage of using an adapter with a DC-DC converter is that you also get line regulation free of charge, and the load regulation can be better than your usual line frequency power supply. The adapter also provides protection against overload. I do my own layout for the DC-DC converter, and I've been doing SMPS layout since around 1995-96, so I'm pretty good at it.
Last edited:
My first stereo amp was a 6BM8 SE. Way too much B+, but it could handle it.It was similar to yours except old fashion resistive load for the triode and Shade feedback. As I recall, it sounded great despite the really small Output transformers. Still love that tube even though they're getting expensive.
I scored an octet of the Russian 6BM8 equivalent for what amounted to beer change.. They're Svetlana tubes, too, so the quality is likely pretty good.
They say the most important thing in real estate is location location location.
The equivalent in SMPS design may be layout layout layout.
Trying to shield a bad design is little like closing the gate after the horse has bolted.
I'm not saying shielding is not required at times, the design should be the best possible.
The shield should be the icing on the cake..
Last edited:
I agree with that assessment - there are few people on this flea-bitten planet who really know how to lay out a switching power supply - sadly, it shows in the results.
Imagine putting an SMPS inside of a spectrum analyzer metal "can" enclosure. look for noise using very sensitive 10Hz Resolution filters, and noise floors of -120dBm or better.
Then put a vibrating cooling fan inside the enclosure, and look for "Fan Sidebands" on all of the 3 local oscillators.
An Art and a Science.
Work hard, and learn how to "pound-in" the quality.
Then put a vibrating cooling fan inside the enclosure, and look for "Fan Sidebands" on all of the 3 local oscillators.
An Art and a Science.
Work hard, and learn how to "pound-in" the quality.
I have found in practice that vacuum tube amplifiers are a lot less touchy with respect to SMPS placement than something really touchy like a spectrum analyzer. The first job I had out here in Cali was for a company that made 2-way radio test gear, and they did integrate an SMPS with a spectrum analyzer.
The first VT amp I ever built was a concoction that I christened "The Shrine". All the innards were contained in a 6" cubical Hammond box, with tubes sprouting out the top. No special attempt was made to shield the SMPS from the rest of the innards, though I did implement several proprietary measures to drastically reduce conducted emissions from the SMPS. It was a hybrid affair, with jfets cascoded by 6C4 triodes, and 1625 finals, run in ultralinear mode with a pair of Edcor output transformers. I started out with the XSE series of transformers, then upgraded to the GXSE series for better bass. It enjoyed an extended residency in my living room, pumping out tunes without incident - no moans or groans or extraneous noises were made. You can view the relevant thread here:
https://www.diyaudio.com/community/threads/shrine-se-amp-using-6ah6-and-1625.93560/
The first VT amp I ever built was a concoction that I christened "The Shrine". All the innards were contained in a 6" cubical Hammond box, with tubes sprouting out the top. No special attempt was made to shield the SMPS from the rest of the innards, though I did implement several proprietary measures to drastically reduce conducted emissions from the SMPS. It was a hybrid affair, with jfets cascoded by 6C4 triodes, and 1625 finals, run in ultralinear mode with a pair of Edcor output transformers. I started out with the XSE series of transformers, then upgraded to the GXSE series for better bass. It enjoyed an extended residency in my living room, pumping out tunes without incident - no moans or groans or extraneous noises were made. You can view the relevant thread here:
https://www.diyaudio.com/community/threads/shrine-se-amp-using-6ah6-and-1625.93560/
Last edited:
OK - all my ships look to be coming in - transformers likely on Monday, and PCBs for amp and DC-DC converter likely coming in later on in the week - interesting things are about to happen. I intend to tailor the cathode resistors on the output stage to conform to the current limitations of the output transformers I selected. This will reduce the available output power, but I'm OK with that.
The input Triode 1mA current - at 1mA - the characteristics Plate current Ip vs Plate voltage Vp are more non-linear at 1mA.
Push the constant current up to about 3mA and set bias resistors so that there is about 175 Volts across the Triode with no signal input. Ideally the Grid voltage should be around -1.5 Volts negative of the Cathode with no signal.
Plot this curve for your tube:
Last edited:
If I push 3 mA through the input triode, that means I'll have to go back to a resistive load for the input stage, as that will be way too much current for the LND150 current source I was thinking of using. I'm also using only 200V worth of B+, so having 175V across the input triode is definitely not an option. BTW - I HATE backseat drivers....
Oops 😴 that curve in post #33 is wrong. This is the right graph of Anode voltage (Vp) versus Grid voltage (Vg) for lines of constant current:
The LND150 in a TO-92 would just about do it at 3mA. Throw C2A, R4A, C2B and R2B away, then choose R3A and R3B so that Plate voltage is 175 Volt with no signal. So then the input Triode Plate can swing from 150 Volts to 200 Volts if it needs to. At Vp =150 Volts and Ip =3mA the maximum power dissipated in the LND150 can only be 50 Volts x 3 mA which is 0.15 Watts.
Your problem is at 1mA Ip even with constant current you have non linearity and THD whereas 3mA is better and 8mA would be very close to linearity, see graph above (its for a 5670 tube). See how the 1mA curve is bent.
Looking at that graph again (I know you are not using a 5670 but its the method) 4mA would be even better and would work well with a Vp of 125 Volts. Your 200 Volts B+ is fine. Good luck, hope that helps your thinking.
By the way there's nothing to stop you putting a resistor above and in series with your current source, you could drop 50 Volts or so across that resistor and save your LND150 from getting hot.
Take C6a and C6b out as well and let that Pentode have some local negative feedback.
Even better download LTspice from the Analog Devices website, enter your circuit in there and then you can work out how it's going to perform, plot frequency responses etc.
By the way I'm not a fan of negative feedback from the speaker side of the transformer, much prefer local feedback at each tube. If you have to use feedback from the speaker transformer why not AC couple it, then your input valve can do what it likes rather than have its Cathode forced around by the DC feedback.
Don't let me put you off I'm just trying to give you a different perspective to work on. Download LTspice!
-
Last edited:
Not happening - LND150 will NOT make it at 3mA ( did you bother to check its junction to ambient thermal resistance ?) How much voltage do you expect to have across it?. I'm not tearing up this design and starting over because you have some quibbles. The PCBs have already arrived, and I'm going with the design as-is. If it has a little more distortion, so be it.
Dissipation across the LND150 on this simulation would be 75mW with a 3mA constant current. Total Harmonic Distortion is -60dB Gain is about x17 Input is 100mV peak (can of course be more). The Anode of U1 sits at about 175 Volts. R3 would need to be selected for your tube. Here is the simulation showing input of 100mV peak and the output at Op1 frequency is 1kHz but will be flat below and above the audio spectrum The tube U1 would dissipate about 500mW in the Triode excluding heater dissipation:
In electronics you have to learn to tear up hundreds of designs before you get a good one.
-
Last edited:
First you have to find an LND150 that will pass 3 mA - that's at the absolute limit of their IDSS. The ones I have been able to match up are maxing out at around 1.6 mA with no source resistor. Guess which bias point I'll be using? According to the curves, from the Svetlana 6BM8 triode, that still places me in a fairly linear portion of the operating curves.