Timeout - I need to repeat: Löwe did split lower and higher audio frequencies in their high end radio designs in order to enable a resonable feedback in the lower frequencies. A very good sound resulted.
The transformer is the weakest point. It gets worse and worse, when you raise the energy, it shall transport.
Philips invented a direct drive of speakers, which they designed as companions for the EL86 tubes, created for this design, too. Brilliant idea, but the speakers have been unique, nobody else produced such. I remember, they had 10W power in one low impedance EL86 pair and in the speaker. No transformer, brilliant sound, but 10W limit.
The transformer is the weakest point. It gets worse and worse, when you raise the energy, it shall transport.
Philips invented a direct drive of speakers, which they designed as companions for the EL86 tubes, created for this design, too. Brilliant idea, but the speakers have been unique, nobody else produced such. I remember, they had 10W power in one low impedance EL86 pair and in the speaker. No transformer, brilliant sound, but 10W limit.
Some class D amps can swing that.
Part of using an existing class D amp is the pre-existing quality of getting from analog or digital to a PWM. It may be better starting from a well refined methodology / purpose built IC implementation than reinventing the wheel via discrete linear ramp generators and comparitors. Not as much fun, though.
OK, so your tube amp will sound nothing at all like a conventional tube amp - got it. I was thinking along the lines of keeping some of the "charm" of an analog tube amp, along with a functional section replacing all the signaling upstream of the power tubes. A true "cement half A to half B" proposal; then iron out the kinks to make it work.
Just missed a Heathkit 6BQ5 stereo amp for ~1/4 of what they usually go for. Was thinking of getting it to try, but hesitated. Soon enough, someone else saw it too, but didnt. Oh well. Merry Christmas, everyone!
Hi, jjasniew,
I will try to remove any own sound component. My target is, "make that thing work as best objective data from measurement instruments as possible".
You can view it as an amp with lots of power for home use, quality like a very good semiconductor amp. Linear, wide frequency range into cellar and up to 20kHz, low distortion. The analog to PWM is not complex at all, just needs 2 tubes for each opamp. Tuning the feedback loop and the windings will take its time.
Merry Christmas!
Hi ximicas,
I spent some minutes to check the situation. Recap the circuit, there is one capacitor right at output of the diodes center and that can be viewed as DC source. This is due, when low audio frequencies are fed through at high power. That DC will go straight to the opposite path and shortened as soon, while not earlier, as the transformer of the opposite side goes into saturation.
That needs to be circumvented.
Greetings,
Joachim
Last edited:
I got rid of one entire transformer and double the winding count on just one core. I also got rid of the ugly semiconductor diodes.
Please note the very different switching on/off of the tubes and the use of more windings, in order to generate true opposite polarities. The magnetic discharge is not completely shown, but you see that I expect to do it during the time not shown. When both tube arrays are switched off in magnetic discharge, voltages at anodes will raise to 2kV, limit of the GU 50. Again, that energy will fed back into the supply by an extra winding and EY500 diodes. Using this, I can extend the 66/33% to 90/10% or alike.
Creating proper timing is possible.
Result: Less complexity, halfed transformer count.
I will do more checking and repeat drawing this stage, when it looks promising. Critics welcome.
Enjoy holy days,
Greetings, Joachim
Last edited:
Hi .Because of push pull primary winding connection , there will be always 1kv at anode of tube which is off at that moment ,but at pause (deadtime) may be higher .Normally class D amps can have wide range of modulation ,i think up to 95-99% of time ,but they are transformerless. Transformer may add upper limit because of saturation ,and lower limit because of inductance and capacitance , which will simply absorb short pulse .But this needs to be tested first . Do you have any simulation software ? I would recommend to simulate this circuit first ,and see which new problems / difficulties will arrive .
Yes, you are right, during off state of the tube, on state of the other polarity, anode voltage rises to double value, because of symmetry.
I can make use of some timespan above 2/3 of the time axis complete cycle, then both tubes must be shut off and magnetic discharge must be done. This is due to the very low frequencies, which will raise flux from cycle to cycle.
I estimate 1/3 of the cycle to discharge, because the anodes will not allow more then 2 kV, 4 times the nominal anode voltage. So there remains some security. Anode voltage will be higher without load and I do not want to regulate that. Therefore peak voltage at anode might reach some 1800V.
The magnetic discharge circuitry is the same as before, the discharge winding count will have 1/3 of the primary winding count.
I will make a better drawing, in order to get the full circuit.
I have Cadence pspice for TI here, quite good, but:
No tube model within their generics. As soon, as I pull in a foreign model, it stops working usable (license caused limitation).
I have seen in this forum, another tube circuit simulation from another semiconductor manufacturer and I would love to simulate.
It will be enough, to run with any pentode, a transformer model which allows winding counts, inductances, flux level and magnetic discharge cycle. Usually I put my parasitics with aproximate values. PSpice is critical for generic components, if you omit these parasitics.
Can you make a proposal for a good spice simulation environment?
Greetings,
Joachim.
For simulations i use microcap ,it become free some time ago .I just checked , there are just generic triode ,pentode ,no concrete models ,but if you find spice data ,you can specify that in circuit .For transformers there are some parameters enterable also ,but need to know them .Many here use LTSpice ,but i'm not familiar with it and what components it have .
I feel good with is circuit strategy. Anything is used at it's full potential and compared to the last one, I was able to strike out a complete transformer and 2 last ugly semiconductors 😜!
Now, any component is within reach of available technology from the mid 50's years.
Astonishing.
I know, this is an absolute energy waste. But if I get my targets, the amp can run DC to 20 kHz in high audio quality.
Now, any component is within reach of available technology from the mid 50's years.
Astonishing.
I know, this is an absolute energy waste. But if I get my targets, the amp can run DC to 20 kHz in high audio quality.
I would not expect to work in very low frequencies ,because of transformer limitations ... Maybe if you will wind it in some special ways , many primaries ,many secondaries ,as litz wire ,then connect them properly ... This needs testing
It will work to produce even DC, that is for shure! But for getting low distortion high frequencies, you are right, the transformer needs little parasitics.
The litz wires are better then solids up to 2 MHz. Above, solids are equal. Their multiple isolation needs room. At rectangular signals at the GU 50 anodes, I am not sure, how much (frequency) I can expect in full swing.
The pentode was designed for 300 MHz (lower amplification). There is potential, but it was designed for analog, linear designs. I will see. I will use the thermometer at the transformer - coil and core.
Secondaries will tell the truth 😎
I made a small step towards the power stages. Here is a symbolic schematic. It has one error in it, that is the merger of triangle and discharge control signals. It would be more precise, to let the triangle run uninterrupted. So ignore this detail. Triangle will run with fixed frequency, uninterrupted. The discharge control will fire with fixed frequency, for defined time, which also needs to be well above high audio limit, in order to fulfil Nyquist requirements. The frequencies must have a difference of more than high audio frequency limit. Example: basic triangle (PWM) 200kHz, discharge control as asymmetric multivibrator with ~ 40kHz.
This is easy to generate and will not generate an audible mixer frequency.
Here is the base circuit. Please omit the inverter after the comparator. It is moved into the group of three tubes, which do gate driving . Gate 1 of GU50 signal swing will be between -1V and -50V. GU50 must switch off - single or both. The left triode will cut off, magnetic discharge will take place. If it is off, the differential amp will work. Three diodes could pull from one common resistor.
I hope, the PCC89 will come in, handy for high speed switching.
The GU50 will be 2 times 2. So 600mA will be coming in to the transformer. Swing of 450V should be expected. That is a potential of roughly 180 W due to magnetic discharge loss time.
Next week, I will try simulations.
Greetings, critics welcome!
Joachim
This is easy to generate and will not generate an audible mixer frequency.
Here is the base circuit. Please omit the inverter after the comparator. It is moved into the group of three tubes, which do gate driving . Gate 1 of GU50 signal swing will be between -1V and -50V. GU50 must switch off - single or both. The left triode will cut off, magnetic discharge will take place. If it is off, the differential amp will work. Three diodes could pull from one common resistor.
I hope, the PCC89 will come in, handy for high speed switching.
The GU50 will be 2 times 2. So 600mA will be coming in to the transformer. Swing of 450V should be expected. That is a potential of roughly 180 W due to magnetic discharge loss time.
Next week, I will try simulations.
Greetings, critics welcome!
Joachim
Last edited:
I think in class D deadtime must be minimal , because at that time audio signal is not generated or feedback can compensate .Too big time for deadtime may cause artifacts like no-idle current in class AB .Many years ago have tried simple class D amplifier and noticed that ,when increasing deadtime delay,to prevent slow transistors heating without signal ,some distortion appear when with signal .To prevent distortion and maximize output , duty cycles at idle must be close to 50% . Maybe different way exists , like both legs 10% at idle , then at positive in example ,one leg increasing ,second decrease ,even cut off ,but this is like DC operation for transformer then ,similar to low frequencies amp .First assemble it without any feedback and try what you will have at output , feedback may make more problems , because it will get delayed output signal after filter , this makes instability .For control signals with 50V amplitude - who will create them ? Tube input capacity wil defeat any resistor load in driver ,driver needs to be totem pole ,maybe from CRT monitor or tv .It will be easier as someone recommended use cathode modulation method i think.
Hello ximikas,
Thanks for input. To make it clear: an idle amp like this shall output a high frequency square wave, which is interrupted during magnetic discharge of the transformer. Both fixed frequencies, but different and apart from each other more then audible. Output filter is important and not detailed.
Those 3 triodes shall run with 20mA and as VHF triodes, I expect sharp transitions. G1 capacity is 13pF. It will be a high slew rate, I hope. Otherwise, I will select higher valued RF triodes with higher current.
The triple can be thought of as a treefold amplifier 😎
I will maybe get other surprises at simulation, let's see!
Cathode modulation with 600mA requires something, which is slow (and Semiconductor). I want to avoid them.
The bridge at the discharge winding can be replaced against a double winding and two EY500 again.
Let me add: the entire idea is derived from switch mode supplies, mixed with H bridges built with coils instead of complementary transistors.
Greetings, Joachim
Philips did make several types of 800 Ohm full-range loudspeakers with a power rating of 20 Watt, like for instance the AD5200AM. They even made a dedicated 20 Watt woofer, the AD5201A, nicknamed the Bombardon, with a conus made of stryrofoam and a (as far as I know) unique way in which the conus is connected to the outer rim of the speakerframe. I own two of them and made sealed enclosures for them and combine them with seperate 9710AM's (also 800 Ohm) in open baffles and use bi-amping.
Philips never made a dedicated 800 Ohm tweeter though. For that they used one 800, or two 400 Ohm in series mid/high loudspeakers with whizzer cones, like for instance the AE37011. The 800 Ohm loudspeakerboxes AD5055 contained one Bombardon and two AE37011's.
In 1961 Valvo published a schematic for a 25 Watt OTL amplifier (400 Ohm) with E130L's.
Philips AD5200AM
Philips AD5201A Bombardon
Philips AD5201A Bombardon (my site)
Philips AE37011
The AD5055 (not mine):
Great stuff, if someone understands what they have been after! The larger speakers are new to me, only seen one in my small Philips device handbook. Congratulations!
I hope, you have a Stereo system! Even 10W crisp clear audio is great.
Today, the speakers are creating almost flat frequency response, but the price is efficiency. You need much more power for driving them as in those days.
I remember I had styropor speakers with 8 Ohms. But their purpose was different and sound good and silent, or loud and resonant depending on enclosure.
Greetings,
Joachim
What resistance are you planning to use for driver tubes load ? G1a G1b on the right side . Problem is same as with crt tubes , amplifier had low bandwidth if was resistor loaded ,later AB class totem pole with diodes for some idle current ,that allowed crt tube capacity to be discharged quickly . Lets say 13pf input capacity ,if previous tube is loaded with 10k ,it will make some filter and pulses will be no longer sharp ... Try that in simulation . For 200khz square wave and also if you wanna higher audio quality ,must be very sharp edges ... like it will be 100mhz signal .As i remember from one of amplifiers input filter , 1k 1100pf gives 130khz cutoff (for sinewave ,which has no sharp edges ) .So then 10k 110pf would give same 130k . For 100khz sine wave keep less distorted ,i use 1k 220p and even less input filter .Maybe with resistor load you get sucess ,but that will be very low value , like 2,2k ... Try to simulate square wave 200khz behavior after such filter .
I will try to go as low as possible. If I do not reach reasonable slew rates with PCC89, I will fetch higher performance (current and speed) triodes, I have some. 13pF is not that much, but 50V is a lot. I made rough estimations and resulting times have been in around 40 nsec. PCC89 is cheap, fast, low voltage and a double system. Driving 2 or 4 of them will be easyer. They have 1.9 pF gate/kathode, each. My option is, to use the PCC89 as cascode amp (2 of the 3 triodes) 50V could be in reach, but I don't know.
I have no large signal response application notes about it.
Greetings,
Joachim
For first tests ,i would try cheap ic like pam8610 board as frontend , and irf540 or even irf510 for cathode control . Cascoded power supply , 12v for pam8610 , +50v for output tubes gate , diodes to limit voltage at cathode to +50v level . MOSFETs would drive tube cathodes to ground . Mosfet gates connected to pam outputs . With such simple setup you would design transformer easier , later you may replace mosfets with tubes ,make your modulator and so. But this is just my one cent ,or idea .
Thanks for proposal!
I will start with simulation, simulation, simulation, first for output stage, then driving stage, then the entire circuit.
Then build a minimum: 1 tube GU50, a part of the windings or all and build the gate driver with PCC89 and use square wave / pulse generator to observe driver and GU50 performance, transformer parasitics, find saturation point of the ferroxcube using 2 GU50 tubes with 300mA each. Luckily, I use tubes and they will not blow immediatly. That gives me borders, not to exceed.
If all that looks fine, proceed to the full circuit.
Because I have stocked all toroid transformers, I can do safe starts with lowered anode voltages in case of mistakes.
If anything looks much weaker than simulation, I change that first, before taking the next step.
As last stages, I will try to do the triangle, comparator, magnetic discharge control circuits and finally, the feedback loop adjustment.
When I start to build, I will use a chassis out of PCB material, for ease of changes and reasonable ground system.
Finally, I will use something, which is a polished chrome cover, where all tubes stay visible and wirings are within. The toroid transformers will be hidden inside, too.
It will be stereo, of course. I own big bass reflex speakers, which might withstand 180W peak power for a minute.
My neighbours won't.
Because this thing can and will emit RF energy in big frequency range and power, a copper net with chrome will shield me from airport's military police ☠
I am happy to have instruments, to measure that and stop, before I get stopped.
But it will be a great experiment which is absolutely nuts.
Greetings,
Joachim