If you fuse the secondary, you fuse the AC, not the DC. The fuse opens more reliably. But then you still have to deal with cap charging current, which is somewhat less than the momentary decaying DC component on the primary at switch on (unless you happen to switch at the zero crossing). I guess at these lowish B+’s it might not matter.
My experience with power trafo ringing seems to be the opposite - never had a peep out of a typical +/-60 to 80 volt supply using an 800 or so VA trafo. When I started messing with 400 volt power trafos, I start getting that annoying “hissing buzz” that you can’t track down with a scope to save your soul. Snubber on the secondary and POOF it’s gone. I have, however, had a lot of trouble with ringing in small cheap power trafos for low power solid state amps. Bad enough to be convinced the amp is oscillating.
My experience with power trafo ringing seems to be the opposite - never had a peep out of a typical +/-60 to 80 volt supply using an 800 or so VA trafo. When I started messing with 400 volt power trafos, I start getting that annoying “hissing buzz” that you can’t track down with a scope to save your soul. Snubber on the secondary and POOF it’s gone. I have, however, had a lot of trouble with ringing in small cheap power trafos for low power solid state amps. Bad enough to be convinced the amp is oscillating.
I like your suggestions. A relay on the output would be simple in terms of protection... Are there any salient scientific reasons for why a relay would be undesirable in the "signal path"? I can't see any reason why a relay should degrade the signal in that position. Alternatively I like the idea of the DC condition putting the whole amp into protective shutdown, although I will have to really sit down and come up with an elegant solution that integrates all the necessary protections (and activates them in the right order). The MOC3010 looks handy, and I wouldn't have thought to use something like this otherwise. I'm not an EE and I don't have experience integrating a project of this complexity, but this is how I learn
My plan was to fuse the primary windings of the HT transformers. Since it's a two channel circlotron the design requires two separate trafos that have two isolated secondaries on each. The output stage B+ and Screen supply is derived from the same transformer, so in the event that a primary fuse blows the B+ and SG+ should both go out. I will definitely integrate the heater supply transformers into the fused circuit as well (separate heater and HT supplies per channel).
I have heard mixed recommendations on the necessity of snubbers, but I figure since it's being mentioned I will look into adding them to the HT circuits. The DCR of the secondary windings of the transformers I'm using are pretty low (~1 to 4R for the output trafos and ~30R on the preamp HT winding).
I got the red compactron sockets from two separate ebay listings a few years back, 24 in total at about $1 each. These sockets omit pins 6 and 8, so they were likely meant for high voltage on plate pin 7 where the risk of arcing between pins was a real concern. They were the perfect solution for my 20 power tube design, and I don't think I will soon find a deal as good again.
I will check these threads out, thanks for the resource.
Feature creep is an issue for me, but ideally I would like the ability to easily monitor all the cathode currents. I may add something more sophisticated later.
The nice thing about the circlotron topology as opposed to a Futterman design is that failure of one tube will not wipe out the others in a chain reaction. All you get from the resulting imbalance is more current shunted through the speaker and balance resistors. Still not good for the speaker, hence the need for some sort of offset failsafe. The question is whether to put this failsafe on the front end or back end of the amplifier
Unfortunately, it can. A small amount of oxidation on the contacts can introduce distortion. Actual nonlinear distortion, due to the junction partially rectifying.
Even in you ignore that, the relays that you would usually use are not rated to open a typical “the amp failed” DC fault current. A typical 20 amp AC relay will usually do the job, but not with 100% guarantee of reliability. Worst case is the contacts can weld shut.
MOST of the time when I have a speaker relay trip on one of my amps it is on a big solid state PA amp, and due to a large transient or relatively small DC offset coming from the preamp. Garbage in, garbage out - and *click*. If the amp is dead on start up, the relay just won’t engage. None of this will kill the relay or prevent it from opening. If I get oxidized contacts and get distorted sound, I just “blast” it for a few seconds and the resulting high current clears the issue.
Could the oxidation risk be lesser with sealed or solid state? Or does the solid state relay introduce other problems? Another thing, is the oxidation of relay contacts worse than a speaker wire to a binding post? I would think that there is more inherent risk and variability with a hand tightened binding post and speaker wire.
I should clarify that the relay control I'm envisioning would theoretically be controlled based on some sort of DC balance reading and the signal current would pass through the relay switch contacts.
I should clarify that the relay control I'm envisioning would theoretically be controlled based on some sort of DC balance reading and the signal current would pass through the relay switch contacts.
I'm not in to Circlotron design but it appears there is direct feedback from each output 'side'. If so, then if one output stage valve goes awry then the remaining valves on that side could be stressed in order to compensate, as well as the general balance concern about wanting a total output but not knowing what each individual is doing (ie. one valve could be nearly red-plating if the others are weak and you may not know).
The power supply appears to use one power transformer to supply screens for one half, and anodes for the other half of the output stage. As such, you don't want one power transformer primary fuse opening unless it is interlocked with the other power transformer!
The power supply appears to use one power transformer to supply screens for one half, and anodes for the other half of the output stage. As such, you don't want one power transformer primary fuse opening unless it is interlocked with the other power transformer!
Not all solid state relays are suitable for switching speakers. You can’t use a triac to switch a speaker. - and the old-school SSRs used them.. Mosfet types must be used, and they need to be good for switching AC. Not all of them are. A single polarity mosfet has a body diode. AC rated SSRs would need *two* mosfets. Sealed mechanical types, mercury-wetted types, be prepared for sticker shock.
Oxidation is much less of a problem on binding posts because you *can* tighten them. It’s not a big problem in my typical applications because I’m not after the last two parts per million distortion.
Oxidation is much less of a problem on binding posts because you *can* tighten them. It’s not a big problem in my typical applications because I’m not after the last two parts per million distortion.
We use 6AS7Gs which have a mu of 2. So we are running large Bias voltages and big swings from the driver, since the tube is relatively insensitive. That's why we don't match the tubes. Our Novacron amp uses 6C33s and that one does need matched tubes and also individual bias controls for each tube.
Your power tubes have greater mu and they are of course high transconductance. At the very least I would install some cathode resistors to help reduce current hogging! You'll find that doing this helps reduce your output impedance even though the resistors are in series with the load. This is simply because you force the tubes to work together better. It also gets you more power, lower distortion, cooler tubes and longer tube life.
Anytime a circuit change offers you so many benefits like that, its worth looking into. Since your power tubes are a bit higher current, I suspect you'll need 10-watt cathode resistors in order to survive an arcing event or internal short that causes the tube bank to lose bias.
We're using 5 ohm devices which work fairly well with the 6AS7G sections. Since you have greater mu that value will be considerably more helpful than it is in our circuits.
We use a separate power transformer to run our output section. This is so that one fuse can control the B+ for the power tubes which in addition to preventing damage to the amp is also our speaker protection fuse. This has worked quite well for us in the last 45 years; if we were damaging voice coils we wouldn't be in business very long!
You'll have to fiddle with the parameters but you might be able to use a fast blow fuse for this application; that might give you pretty good speaker protection without a lot of complexity.
If not that then a fuse in series with the output itself will also work. But you'll still need to protect the circuit from burning up if things do go south.
Since you are running feedback you also have some power supply rejection. This means you can run lower capacitance values in the supply. The fact that the filter caps are in a Circlotron is helpful too- the sawtooth waveforms (if you don't use a Pi network) cancel in the load. These facts are helpful if you go with a separate power transformer to run the output section B+.
Nice build BTW!
As it stands I currently have 4.7R 1W resistors in the Cathodes, which may be enough resistance but not high enough power rating. What I'm gathering is that it is not beneficial to have cathode resistors fail as a result of overcurrent and that the circuit protection should be baked in elsewhere. Bit of a half-baked idea on my part.
When I have some time, I plan on testing my collection of tubes so that I can at least bin them based on how close to nominal they operate. Not matching, but hopefully I can weed out problem tubes since I opted to not be able to individually control the bias.
My current design uses a separate output B+ transformer per channel, and I plan to fuse each one individually on the primary side. Since I'm employing a soft-start circuit I think a fast blow fuse could work (I will have to work out the specifics). I was also advised to put the heater transformer on the same fuse so that everything on that output channel shuts down.
I'm open to the idea of placing protections at the output as well. I had some earlier conversations about relays but it's hard to pin down how much of a sonic impact they have in practice. I doubt I could hear it, but it's something I've marked as worth researching.
I figure my output power supply board will see a revision before this is all said and done (I jumped the gun and no sooner was drilling holes and jumping traces). So the possibility of reducing the capacitance in the bank is intriguing to me. Is this possible because of the feedback introducing ripple rejection on top of the inherent cancelation of the circlotron?
Thank you for the kind feedback, it means a lot to me since your threads provided much of the inspiration to start this project.
Jerdy, the comments so far indicate that it is in your best interests to put some pre-design effort in to protection. That way you start your testing having some of the known failure modes covered by a protection mechanism, and that you are well aware of what could happen.
At the moment you are knee jerk responding to forum comments, and may not be suitably aware of what failure could happen (and why), or be suitably aware of what protection mechanisms can cover a failure mode and to what extent (as there are often a few practical methods and they each may have their own quirks).
Of course you could just continue to set up the basic cirlotron circuit and start testing as is, and go through a practical learning curve. I guess the disadvantage of that is that you are not following a well trodden path, even though there are similarities with other valve amps and even specific circlotron amps.
It may seem pedantic to go through a failure cause and effect assessment, and then assess protection mechanisms for each issue, and then decide if you could or would alter your basic circuitry, but imho you could easily end up with some substantial collateral damage and quickly become disheartened about why you started this project.
At the moment you are knee jerk responding to forum comments, and may not be suitably aware of what failure could happen (and why), or be suitably aware of what protection mechanisms can cover a failure mode and to what extent (as there are often a few practical methods and they each may have their own quirks).
Of course you could just continue to set up the basic cirlotron circuit and start testing as is, and go through a practical learning curve. I guess the disadvantage of that is that you are not following a well trodden path, even though there are similarities with other valve amps and even specific circlotron amps.
It may seem pedantic to go through a failure cause and effect assessment, and then assess protection mechanisms for each issue, and then decide if you could or would alter your basic circuitry, but imho you could easily end up with some substantial collateral damage and quickly become disheartened about why you started this project.
I agree with this. I was aware of the lack of such necessities in my original designs, so the crowdsourcing of criticism here has given me some much needed direction. I appreciate your input a lot and I’m going to do some more legwork to make sure this project has a better foundation.
Happy to help!
I think you're doing well with this. You will want the cathode resistors to survive (hence 10 watts for these tubes). It gets old opening up the amp every time a tube fails- much better if its a B+ fuse rather than a resistor.
You're actually better off having the filament separately fused. If the B+ transformer fuse opens up, the B+ will fade off and if the filaments are lit, the output section will go into cutoff, limiting current much faster. If the filaments fade with the B+ the tubes will continue to conduct as hard as they can until the cathodes have cooled off.
If a tube shorts, the output section can dump a surprising amount of current- much like a solid state amp and for the same reasons. So you have to be careful about the fuse value for the B+ transformer, but if that is the only thing its controlling, your speakers will be fine unless you are using a very low power full range driver that only handles a few watts (Such drivers should simply have a fuse in series with them if used with any amp that is a multiple of their power handling abilities)!
Yes. If using feedback you only need to satisfy the timing constants between the impedance of the output section vs the capacity in the power supplies since the feedback will give you some extra rejection on top of that innate in the Circlotron. That should be lower than the bandwidth of the amp (which should be set to 2Hz) to prevent excess modulation.
You're really not able to install the amount of feedback you really need (which is why we don't use any at all in our designs) so you have to help things along a bit. So the power supply TC should be good to a second or two to be well outside the lowest the amp will need (2Hz is required to prevent phase shift at 20Hz; if you were able to have enough feedback you could cut off higher but that value is a good 35db which will exceed the phase margin of the design).
Do you have some form of indication that a particular tube has failed, or just replace them all, or test each tube external to the amp, or ?
Could you elaborate on this a bit more please?
If B+/screen supply is stopped by a fuse, the voltage rail capacitors would discharge. For a typical output stage with a substantial bias current through the tubes, that discharge would typically happen quite quickly even with the filament supply also being stopped. I'm guessing this amp is different, but would the discharge rate depend on signal level at the time (ie. no signal would cause the slowest discharge)?
Would that fault be a 'hard' internal short in a tube (ie. melted or loose metal touching other metal). That appears to raise one speaker output to a rail, so the fault current would then be limited by the speaker resistance and the signal applied to the other side.
Just a failed bias or a grid voltage fail to a tube would cause a 'soft' short where current is still related to the screen and anode voltages and tube V-I curve.
Patrick Turner went to great lengths at providing safety features to his 300 watt monobloacks....300w-monobloc-about
Patrick's goal was not to allow any tube to exceed 35 watts, this is very important from a reliability standpoint..it is the idea, not the actual circuit, what what Patrick trying to achieve?.
if you can somehow monitor cathode current and plate voltage at the same time real time, that would be awesome.....Pete Milette i think made such an amp...815 AB2 Amp
if you can somehow monitor cathode current and plate voltage at the same time real time, that would be awesome.....Pete Milette i think made such an amp...815 AB2 Amp
I’m familiar enough with my circlotron that I think I can avoid a goof like that. I’m grateful for freely given resources—it’s beneficial knowledge even if I can’t put it to use immediately.
I’m hearing your advice clearly, and I’m working on it
That depends on the amp. Since the tubes are paralleled, you can remove pairs and see when the fault is cleared. Or place test jacks at the cathode resistors to measure the voltage drop on the resistors. Or you could wire those to a multiposition switch and use their outputs to drive a meter.
A cathode/pate short is really rare and will be less so with tubes employing plate caps! So we don't need to worry about that. But what can happen is the cathode coating can have flakes that fall off and touch the grid of the tube- that will cause arcing and loss of bias in that tube bank.
Now if the filaments remain lit and the bias voltage continues to be present, but the B+ fuse has failed, the B+ will decrease rapidly as the tubes are otherwise set up to conduct. At some point if the bias is present, the tube will simply go into cutoff. The B+ will not have to discharge all that far for this to happen, and the signal level won't be important although it will happen faster if signal is present.
If the bias is removed with or without the filament current, the tubes will conduct for a longer time because the B+ will have to discharge to a much lower voltage before the tubes won't conduct.
Protecting the speaker is probably the most important thing going on here. If a tube has shorted in such a way as to cause the amp to push DC, if a B+ fuse is to offer protection, it can't be rated to also be supporting some other function such as the filaments. It should be for B+ and B+ only.
(The fact that Atma-Sphere has done this for 45 years is pretty good evidence that this technique works. We'd be long gone if the amps were damaging voice coils!)
Its helpful for the amp to be as easy to troubleshoot as possible. By having a fuse for the B+ transformer, you've limited the possibilities to what's going on if that fuse blows. Most of the time it will simply be a power tube.