Kevlin Lead Capacitor Wiring - Last Cap

[L0rdGwyn]

I am elbow deep in a SET build with an unregulated power supply. The final capacitor in my power supply is a Clarity Cap TC Series 130uF 600V, it is a 4-lead Kelvin style capacitor, one per channel. These capacitors have an V+ in and out, and V- in and out leads. It is a CLCLC supply.

Is anyone familiar with these capacitors? What I am wondering is if I should leave the last V- lead disconnected, jump it to the other V- lead since these are the final caps in the supply, or ground it independent of the other V- lead?

I've included the picture Clarity Cap uses to show how to wire these capacitors. Typically you would connect the V- in to the negative terminal of the proceeding capacitor and V- out to the negative terminal of the following capacitor.

Any help is appreciated, thank you!

 

[brightcity]

You will loose many of the Kelvin benefits if you cannot wire as in and out. In many cases it is easy to split the cap into two stages. If not minor edits to the schematic will greatly benefit.

 

[trobbins]

Are you saying you have two independent secondary windings, and each winding is rectified in to its own CLCLC filter? It would make sense to then use the Vout terminals of the last C as the feed to each channel's B+ supply (which would be the B+ connection to each output stage - pos to say OPT, and neg to cathode node).

The cathode node of the output stage is then a 0V node that would be stared or distributed stared out to driver and preamp stages.

Where the B+ supply is 'grounded' to chassis is a separate issue, and is only done at one point on the 0V distribution - often it is done at the preamp end to minimise earth loop length and noise with external equipment.

That final C cap then has pretty much steady current flow from the Vin end, and the Vout terminals would have max signal currents that circulate through the output stage.

 

[TonyTecson]

oh, boy, here we are again....

 

[L0rdGwyn]

You will loose many of the Kelvin benefits if you cannot wire as in and out. In many cases it is easy to split the cap into two stages. If not minor edits to the schematic will greatly benefit.

Are you saying you have two independent secondary windings, and each winding is rectified in to its own CLCLC filter? It would make sense to then use the Vout terminals of the last C as the feed to each channel's B+ supply (which would be the B+ connection to each output stage - pos to say OPT, and neg to cathode node).

The cathode node of the output stage is then a 0V node that would be stared or distributed stared out to driver and preamp stages.

Where the B+ supply is 'grounded' to chassis is a separate issue, and is only done at one point on the 0V distribution - often it is done at the preamp end to minimise earth loop length and noise with external equipment.

That final C cap then has pretty much steady current flow from the Vin end, and the Vout terminals would have max signal currents that circulate through the output stage.

Thank you both for your help. This is the first amplifier I have ever built and I am no engineer (clearly!), so I am learning quite a bit as I go.

trobbins - I have included some simplified schematics below with two different possible wiring of the Kelvin caps. What I have is a single secondary winding. The power supply splits into two rails at C2 (100uF). My driver tubes are cascode CCS loaded, the B+ for both drivers is also pulled from the 100uF cap.

I am star grounding this amplifier, just a matter of the best way to wire the Kelvin capacitors. After discussing it with another DIYer, right now I am prepared to forgo the advantages of the Kelvin capacitors and simply wire them like normal caps (see Option A below). I am told this is the better of the two options by him.

The other possibility (Option B below) is to wire them as they have been intended with -Vout connected to the cathode node of the output tube, which would then be on a bus with the power supply ground and connected to star ground via the negative terminal of the resevoir cap.

Right now I am planning for Option A with a separate connection to chassis ground from the output tube cathode node, but if there is no downside to connecting the output tube cathode node on a bus with the power supply ground, I could do that as well. Please let me know what you think.

I have included some pictures of my amplifier too, it is mostly done, I have been working on it for almost three months, just need to wire it up

Thanks again.

 

[trobbins]

I would recommend option B, and move the chassis link to the 100uF negative terminal, as that would be the least noisy common distribution point for both channels.

Why are you not taking the driver circuits (and any preamp circuitry) from the channel's B+ feed and 0V node, as that is where the driver and preamp signal current loop to. You should read up on distributed star layout, where each stage has its own 0V star, and the 0V nodes are linked to each other in a sequential manner.

The issue with stereo amps is that they often end up with a compromise for 0V connections, especially if any external equipment connecting to them has a common 0V node for the two signal channels (rather than two separate channels each with an isolated 0V). That causes multiple 0V loops, and you may notice ground related noise/hum.

I can't see any secondary side fusing, or use of protective ss diodes for each valve diode plate - it would be sad for a dodgy valve rectifier to take out your PT, or a dodgy output valve to take out an OPT.

 

[L0rdGwyn]

I would recommend option B, and move the chassis link to the 100uF negative terminal, as that would be the least noisy common distribution point for both channels.

Why are you not taking the driver circuits (and any preamp circuitry) from the channel's B+ feed and 0V node, as that is where the driver and preamp signal current loop to. You should read up on distributed star layout, where each stage has its own 0V star, and the 0V nodes are linked to each other in a sequential manner.

The issue with stereo amps is that they often end up with a compromise for 0V connections, especially if any external equipment connecting to them has a common 0V node for the two signal channels (rather than two separate channels each with an isolated 0V). That causes multiple 0V loops, and you may notice ground related noise/hum.

I can't see any secondary side fusing, or use of protective ss diodes for each valve diode plate - it would be sad for a dodgy valve rectifier to take out your PT, or a dodgy output valve to take out an OPT.

Thanks for your insight again trobbins, you have helped me quite a bit by now I'll try Option B and move the chassis connection to the 100uF cap. Unfortunately, my layout has the star ground closer to the resevoir cap, so hopefully the added distance is not an issue, will be 6 or so inches of 18 AWG wire from the 100uF cap to the chassis.

I'm not sure I understand your question about the B+ node and the driver circuitry. It may just be the way I have drawn it, the B+ for both CCS boards for the driver tubes will be pulled from the 100uF capacitor directly, it will be an ~400V tap, 6mA constant current load.

I will try and find some information on distributed star ground, although I think I am somewhat familiar. It is essentially a combination of star and bus grounding, in which multiple star grounds are arranged in a bus, so to speak.

I am hoping to avoid ground related hum, hopefully Option B will suffice, but perhaps some trial and error will be needed.

Again, sorry my questions are novice, that is because I am a novice this is not my field and I have only started teaching myself three months ago with some undergraduate physics/mathematics education, so only a basic knowledge to work from. I have a few texts, from Morgan Jones, Merlin Blencowe, Radiotron's Designer's Handbook, etc., but being self-taught in a short period of time is challenging.

 

[trobbins]

The driver stage passes signal current to the output stage - that signal current loops through the 0V link between those two stages. 0V link noise can enter that loop, so it is better if the driver stage has its B+ and 0V nodes connecting directly to the output stage B+ and 0V nodes (as in a distributed star connection between two 0V star nodes). At the moment it looks like the driver stage connects to the 100uF pos and neg terminals, which is a bit more removed (and hence possibly worse for noise entry).

Do you have a way to measure output hum/noise, or cross-coupling level when driving the other channel? If so, then it can be worthwhile moving connection points around to see the influence.

 

[L0rdGwyn]

The driver stage passes signal current to the output stage - that signal current loops through the 0V link between those two stages. 0V link noise can enter that loop, so it is better if the driver stage has its B+ and 0V nodes connecting directly to the output stage B+ and 0V nodes (as in a distributed star connection between two 0V star nodes). At the moment it looks like the driver stage connects to the 100uF pos and neg terminals, which is a bit more removed (and hence possibly worse for noise entry).

Do you have a way to measure output hum/noise, or cross-coupling level when driving the other channel? If so, then it can be worthwhile moving connection points around to see the influence.

trobbins - I think I am understanding your question. One thing that is likely confusing things is there is added series resistance in the power supply between the driver and output stages that I have not drawn in my schematic. It was a quick scribble to address the Kelvin capacitor question.

My driver tubes are being biased at 6mA constant current and ~200V. To account for the tubes' maximum swing, ~400V is being sent from the 100uF cap to the CCS boards. The resistance in series with the 10H chokes will bring the output tube B+ voltage to ~314V, accounting for OPT primary DC resistance and -45V bias, for 250V on the plates. So that is the reason the B+ is not being pulled from the same node. I hope I am answering your question.

 

[trobbins]

Thanks for clarifying. In a typical distributed star arrangement, the B+ supply for each stage has its own B+ decoupling capacitor, that is buffered from other stages - that buffering is typically done with a B+ distribution resistor or choke.

If you feed the driver direct from the 100uF pos, then the driver stage may have noise coming through depending on the PSRR of the CCS and how PSRR varies with frequency. Just a heads up.

 

[L0rdGwyn]

Thanks for clarifying. In a typical distributed star arrangement, the B+ supply for each stage has its own B+ decoupling capacitor, that is buffered from other stages - that buffering is typically done with a B+ distribution resistor or choke.

If you feed the driver direct from the 100uF pos, then the driver stage may have noise coming through depending on the PSRR of the CCS and how PSRR varies with frequency. Just a heads up.

Not a problem, thank you for the warning and again for your assistance. I will try your recommended wiring of the Kelvin caps. I'll look for some information on distributed star grounding as well, much appreciated!