If survival of the listener when the headphone's insulation fails is of importance, and it could very well be for DIY electrostatic headphones, it would be nice to have an accurate current limit at some non-lethal value.
Residual-current devices here are usually designed to trip at 30 mA RMS, so 20 mA peak is still below that. However, if the gyrator circuit mimics an ideal inductor accurately, the current through it will just increase without bound when there is a short. An ordinary current source would be safer in that sense, or a gyrator circuit with a built-in current limiter.
The "gyrator" circuit is shown in post #35, the latest version of the main amp circuit in post #39. I want to use a switch to choose between the CCS output and the MU output of the "gyrator", so the current limiting should apply to both switch positions.
What would be a simple-but-good way to achieve the current limiting?
The 5.1k "safety resistors" in the output lines are not enough?
440 V/5.1 kohm is more than enough to kill someone, as is 880 V/(2 times 5.1 kohm).
Actually, because of the symmetrical supply, a current limiter in the gyrator won't suffice, as you can still have a current path from ground, via someone's body, via an output valve to the negative supply rail.
Actually, because of the symmetrical supply, a current limiter in the gyrator won't suffice, as you can still have a current path from ground, via someone's body, via an output valve to the negative supply rail.
So then the current limiters would go into the output lines, just before the headphone connectors.
@MarcelvdG Can you suggest a suitable limiter circuit?
@MarcelvdG Can you suggest a suitable limiter circuit?
Quality Output Transformers, with a primary and a secondary are an extremely good way to prevent B+ (DC) from getting to the headphones.
Nuf Said!
UL no longer allows 1950s style 5 tube AC line operated radios that do not have isolating power transformers to be sold.
Safety First!
Prevent "The Surviving Spouse Syndrome"
Sound Second!
Nuf Said!
UL no longer allows 1950s style 5 tube AC line operated radios that do not have isolating power transformers to be sold.
Safety First!
Prevent "The Surviving Spouse Syndrome"
Sound Second!
440 V across a 5.1k resistor would be 86 mA. The resistor would dissipate 38 W.
At 100 V, the current would be 20 mA, and the resistor would dissipate 2 W.
At 50 V, the current would be 10 mA, and the resistor would dissipate 0.5 W.
The 30 mA limit quoted by @MarcelvdG would correspond to 153 V and 4.6 W on the resistor.
Modern STAX amps use the 5.1k "safety resistors", and James Lin recommended adding 5.1k / 2W "safety resistors" to older STAX units like the T1 in his Audio Express article (aX July 2017). Most DIY amps I found use the 5.1k "safety resistors", and I don't see any other measures for current limiting with these amps. Some of these amps have their output stages biased at similar or higher currents than the OSDEHA (like the Megatron or Blue Hawaii designed by @kevin gilmore , probably also the T2).
It appears to me the 5.1k / 2W resistors work similar to a fuse that opens once there is a (permanent) current that exceeds 20 mA or so.
What am I missing?
You are right, of course. Safety resistors by design act as fuses. Three things to consider:
1. Fuses are quite inaccurate, I guess the same holds for the current that blows up safety resistors - but that's just a guess, their datasheets should provide better information.
2. It will take some time for the safety resistor to heat up and blow. That time needs to be shorter than the time needed to kill the listener. (Safety resistors are normally used to prevent fire rather than electrocution.)
3. If you find a suitable safety resistor and if there will be people building your design, they have to be quite aware that they need to use the exact same type of resistor.
1. Fuses are quite inaccurate, I guess the same holds for the current that blows up safety resistors - but that's just a guess, their datasheets should provide better information.
2. It will take some time for the safety resistor to heat up and blow. That time needs to be shorter than the time needed to kill the listener. (Safety resistors are normally used to prevent fire rather than electrocution.)
3. If you find a suitable safety resistor and if there will be people building your design, they have to be quite aware that they need to use the exact same type of resistor.
Last edited:
The headphones are on your head.
Electro Shock Therapy was not intended to kill anyone, they used less current too.
Do you want to find out on your own what your tolerance level is?
Any lawyer worth his salt would not allow posting such things.
Just my Para Legal opinion.
It may take some time to cause cerebral damage, or to kill . . .
But the Spouse and Lawyer will appear in a Nano Second!
Electro Shock Therapy was not intended to kill anyone, they used less current too.
Do you want to find out on your own what your tolerance level is?
Any lawyer worth his salt would not allow posting such things.
Just my Para Legal opinion.
It may take some time to cause cerebral damage, or to kill . . .
But the Spouse and Lawyer will appear in a Nano Second!
Thank you @MarcelvdG. I am looking forward to testing the high-voltage supply feeding the "safety resistors" into a short to see how much and how long it takes to blow them up.
@6A3sUMMER this is DIY. Everyone dealing with high voltage and other hazards is responsible for his/her actions. Once this project goes anywhere, the documentation will of course have a suitable disclaimer and warning.I will also add a corresponding warning to the first post of this thread according to the diyAudio rules. DONE
@6A3sUMMER this is DIY. Everyone dealing with high voltage and other hazards is responsible for his/her actions. Once this project goes anywhere, the documentation will of course have a suitable disclaimer and warning.
Last edited:
J112 is a JFET whereas DN2540 is a MOSFET. MOSFETs have copious 1/f noise in the audio band. I've measured their noise and DN2540 noise corner is at about 4kHz whereas J112 is at about 50Hz. Nuff sed.
Are we talking about the same thing?
The original comment about the J112 by @PretentiousFood was about the lower part in the "gyrator" load for the anodes of the output tubes, not the CCS tail of the LTP input stage.
- The OSDEHA "gyrator" draft currently has the AOT1N60 part in the lower position (as per the @mogliaa recommendation here).
Should this be replaced with a J112? Why? - The LTP tail draft currently has the DN2540 part in the lower position. Should this be replaced with a J112? Why?
Thanks for your insights!
It's only a sample, which shows the AOT1N60-IXTP "gyrator" (as load of 801a driver) architecture, and simulates it's output impedance.
As you can see, even at 440V B+ the "upper" FET dissipation is significant (the lower is cold), and the bias resistor -in this case 470R- also dissipates.
If you want to use 801a in A2, the grid stopper value should be minimized, or eliminated.
As you can see, even at 440V B+ the "upper" FET dissipation is significant (the lower is cold), and the bias resistor -in this case 470R- also dissipates.
If you want to use 801a in A2, the grid stopper value should be minimized, or eliminated.
Ok, I see. Your circuit is overall very similar to the OSDEHA draft -- did you set this up for the OSDEHA, or did you by coincidence already have this from another project?
Anyways, looking at the bias point of the 801 tube in your SPICE model (260 V, 26 mA) and looking at the 801 curves (see here), the grid would be at -12 V. Feeding that with 28 pp signal would run the grid up to +2 V. That's not a whole lot of "positive grid", and the grid current will be well below 1 mA (according to my curve trace measurements). Still, 1 mA into 1k grid resistor would mean 1 V of distortion to the grid voltage. Thanks for pointing this out! I guess that lowering (or avoiding) the grid stopper is something to test in the first prototype build.
Anyways, looking at the bias point of the 801 tube in your SPICE model (260 V, 26 mA) and looking at the 801 curves (see here), the grid would be at -12 V. Feeding that with 28 pp signal would run the grid up to +2 V. That's not a whole lot of "positive grid", and the grid current will be well below 1 mA (according to my curve trace measurements). Still, 1 mA into 1k grid resistor would mean 1 V of distortion to the grid voltage. Thanks for pointing this out! I guess that lowering (or avoiding) the grid stopper is something to test in the first prototype build.
I've been using Ale's gyrator since the beginning (one of early prototype still working in my 801a preamp).
I tested it with various upper and lower FETs, that's why I recommended to Ale trying AOT1N60 as lower FET (greater transconductance == lower output impedance, than -for example- DN2540).
Nowadays for medium anode current (about 20-30mA) in this position I usually use J310 ... but it's required selection (my batch is between 28..34mA IDSS).
As upper FET I exclusively use 1kV IXTP types (I lost dozens of DN2540 even at low voltage).
For simulation I use the 801a model that Ale developed for me (and to the community).
Enough precise even in A2 mode, several device developed with it.
p.s. tracing DH tubes is need precise biasing. If the tracer use DC filament and don't use virtual cathode (resistors from each filament point to v. cathode point), the "real" bias can be shifted.
I tested it with various upper and lower FETs, that's why I recommended to Ale trying AOT1N60 as lower FET (greater transconductance == lower output impedance, than -for example- DN2540).
Nowadays for medium anode current (about 20-30mA) in this position I usually use J310 ... but it's required selection (my batch is between 28..34mA IDSS).
As upper FET I exclusively use 1kV IXTP types (I lost dozens of DN2540 even at low voltage).
For simulation I use the 801a model that Ale developed for me (and to the community).
Enough precise even in A2 mode, several device developed with it.
p.s. tracing DH tubes is need precise biasing. If the tracer use DC filament and don't use virtual cathode (resistors from each filament point to v. cathode point), the "real" bias can be shifted.
Last edited:
So, should we stick to the AOT or change to the J310? Which is better, and why?
Also, I'd like to understand why you suggest IDSS and matching the J310. The OSDEHA will have the DC operating voltage of the output tubes fixed near 0 V at the anode, and the current is set by adjusting the grid bias. What's the reason for matching the FET?