I'm building an Aikido based headphone amp using the White Cathode Follower output stage. All of that is very straight forward. I'm optimizing the White stage for a 32 ohm load. But...I'd also like to use this as a preamp. The difficulty is that the White Cathode Follower will create distortion if the load is different than what it was optimized for. If it's used as a preamp it definitely won't see a 32 ohm impedance.
How can I make the output think it's seeing a 32 ohm load while really presenting a 600 ohm (ideal) load to an amplifier? An impedance matching transformer seems like the logical idea but it's entirely dependent on the impedance of the load. I suppose I could try to impedance match a particular amplifier, but that's not where I want to end up.
Looking for suggestions.
How can I make the output think it's seeing a 32 ohm load while really presenting a 600 ohm (ideal) load to an amplifier? An impedance matching transformer seems like the logical idea but it's entirely dependent on the impedance of the load. I suppose I could try to impedance match a particular amplifier, but that's not where I want to end up.
Looking for suggestions.
A White CF is not an SRPP- a higher load shouldn't increase the distortion (in fact, I'd suspect the opposite).
My suggestion is that you don't try to use a headphone amp as a line stage. They are different, however much some people may wish them to be the same.
I'm not sure which amplifier requires to see 600 ohms at its input.
I'm not sure which amplifier requires to see 600 ohms at its input.
I had a prototype headphone amp in house for testing about a year ago- it was fine driving 32 ohms, THD in the 0.05% range. When I checked it as a line stage, I was surprised to find that the distortion dropped below my test limit of 0.001%. So... as usual, "it depends."
Different applications usually require different circuits employing different compromises. Sometimes we get lucky, and one circuit will do two jobs.
My understanding of the White is the push pull current balance is dependent on load. Thus the value of Ra is critical in the design. Distortion is supposed to be the downside if you stray from the designed load.
I totally understand I'm trying to stretch the use of this design. If it's possible I'd like to give it a try. I thought of an L-pad going into a 1:4.1 impedance matching transformer. I see those available and they'd approximate a 32 ohm to 600 ohm jump.
I totally understand I'm trying to stretch the use of this design. If it's possible I'd like to give it a try. I thought of an L-pad going into a 1:4.1 impedance matching transformer. I see those available and they'd approximate a 32 ohm to 600 ohm jump.
Given the 100% feedback, I would seriously not worry about it. If the unbalance means that (say) 2nd order distortion increases from 0.002% to 0.005%, who cares, really?
The success or failure of the WCF is largely dependent on providing the right signal level to the bottom grid. It should be equal or smaller than the signal on the upper grid. Biggest mistake is to use a big value anode load, which produces a signal level too large for the bottom triode to reproduce without distortion. If done correctly, the lowest possible load will produce a signal roughly equal on both grids. If presented with a higher load (e.g. poweramp with 10k input), the signal on the lower grid reduces significantly, so no problem there and the distortion decreases even more, just as Sy experienced.
So no need for additional buffers, L-Pads, transformers or whatever. Output level will only alter slightly because of the low output impedance, specially when feedback is used.
So no need for additional buffers, L-Pads, transformers or whatever. Output level will only alter slightly because of the low output impedance, specially when feedback is used.
Last edited:
Although you can minimise distortion with most totem pole stages by careful choice of load, its well not to dwell on this too much. Using a load that is "too high" only increases the distortion a little bit, and it tends to level off rather than increase proportionately. Also, even with the optimum load, the degree of distortion cancellation will vary with valve tolerances, so it may not be such a deep null as you expect. And even headphones don't exactly conform to a constant impedance!
Anyway, the solution to your problem is simple. Just add a dummy load that is present when using it as a line stage, but is switched out when you plug phones in...
Anyway, the solution to your problem is simple. Just add a dummy load that is present when using it as a line stage, but is switched out when you plug phones in...
Pretty much the only people who still use 600 ohm terminations are Japanese broadcast studios...
I'm at work and don't have my text books with me - but from memory, the anode resistor on a white cathode follower should be equal to 1/gm for optimised performance. This means that the voltage gain at the anode of top device = 1, which is the drive voltage for the bottom device. So both devices have same input voltage (but out of phase) and currents are balanced.
I don't thinbk that the value of the load comes into this?
(rp + 2Rload)/mu
That's the formula from John Broskie. I suppose I could do something where I switch in different resistive values for headphones and line out. A lot of the feedback I've heard is I'm in good shape if I design for the low impedance and use it for both as is.
That's the formula from John Broskie. I suppose I could do something where I switch in different resistive values for headphones and line out. A lot of the feedback I've heard is I'm in good shape if I design for the low impedance and use it for both as is.
Well JB knows more than I do ... a LOT more than I do. I have read most of his stuff and I've obviously fogotten this part so please remind me - this is the formaula for what?
Optimum anode sampling resistor. If Rload is very small then it simplifies to ra/mu, which is 1/gm as you mentioned earlier.
Well, the only real problem that I see is that using a headphone amp as a line amp means that it's way more powerful than needed for line amp use. Assuming you're operating in single ended class A then you'd be wasting some power using a headphone amp as a line amp. It's not a tragedy, but maybe not the most efficient use of resources.
Yeah, I don't see ideal results using that formula. My simulations produce different values for the lowest THD. I am using accurate spice models also.
The plate resistor pretty much determines the clipping level for the circuit, as I understand it (though I don't know the details).
Last edited:
Yes, of course!
The main reason I am playing around with tubes now is because of a friend of mine who is really keen on them. He wants us to produce tube products, but I am reluctant, even if we had interested customers. They have some nostalgia for me though, and they are much more of a challenge to use than say opamp chips. The challenge makes it much more rewarding to come up with something that actually works instead of spending 3 minutes designing something with chips that actually works better.... I wonder why I bother sometimes.
Let me revive this topic a bit.
Hi.
My current goal is to test Aikido as the driver for a hybrid. Not a linestage, but full voltage amplification stage: only current amplification is done in silicon.
The driver must supply full voltage range - in my case it's +-25 or even +-35Vpp, and simultaneously have smallest output impedance possible. The output stage is actually a follower with some current amplification factor Ki, and gain near 1, so it merely translates the load impedance to the input impedance of itself.
My measured input impedance of the output stage loaded with 4 ohm resistor is about 10k. With a real speaker, which in general is a cumbersome L/C/mechanical circuit, the Z can drop well below 2 ohm on a good punch (mind the air inertia), or rise above 16 ohm in midrange where crossovers do their nasty job. So we have 10x factor of impedance drift which can lead to humps and drops in the frequency response, if not driven properly. Say 2.5 to 25kOhm.
The question is - will the White scheme help, or it won't? Most White followers are designed to drive headphones, with hundreds of millivolts on the output. I need some dozen volts. On other hand, the lower the output impedance is, the better the output stage is driven. If I implement classical Broskie follower loaded with a CCS, the second tube must have quiescent current not less than 40mA (5x load current in worst case) - ie, to be a real samovar like 6AS7.
Thanks for thoughts!..
Hi.
My current goal is to test Aikido as the driver for a hybrid. Not a linestage, but full voltage amplification stage: only current amplification is done in silicon.
The driver must supply full voltage range - in my case it's +-25 or even +-35Vpp, and simultaneously have smallest output impedance possible. The output stage is actually a follower with some current amplification factor Ki, and gain near 1, so it merely translates the load impedance to the input impedance of itself.
My measured input impedance of the output stage loaded with 4 ohm resistor is about 10k. With a real speaker, which in general is a cumbersome L/C/mechanical circuit, the Z can drop well below 2 ohm on a good punch (mind the air inertia), or rise above 16 ohm in midrange where crossovers do their nasty job. So we have 10x factor of impedance drift which can lead to humps and drops in the frequency response, if not driven properly. Say 2.5 to 25kOhm.
The question is - will the White scheme help, or it won't? Most White followers are designed to drive headphones, with hundreds of millivolts on the output. I need some dozen volts. On other hand, the lower the output impedance is, the better the output stage is driven. If I implement classical Broskie follower loaded with a CCS, the second tube must have quiescent current not less than 40mA (5x load current in worst case) - ie, to be a real samovar like 6AS7.
Thanks for thoughts!..
Last edited:
So you're saying the load on the WCF will be 5k to 20k? That's a pretty easy load for a chunky valve like the ECC88. I wouldnt expect prblems.
- Status
- Not open for further replies.