George,
thank you for your intervention. I've always been used to simulate before building, simply because I always had more time for the former than the latter, so it has always been an optimisation of the time to build/mod. The guitar and bass amps I've built (a dozen) and modded (a hundred) followed the same process. The Hi-Fi amps I've built (three) followed the same process too, even more in depth being simulators more detailed when working linearly. But the final goal is always share ideas and build something better than I did before.
Concerning the fact that "common" tubes can be easily triode connected, whilst it is for sure true, I've found (again, only on LTSpice by now) this a-g1 feedback to be far superior in terms of performances compared to 100% a-g2 feedback, in terms of rp, linearity, available power, and the possibility to adapt it to the driver.
dgta, SpreadSpectrum, smoking-amp,
thank you for your interventions. You all gave me great support to optimize details.
I've already bought four matched 6N23P (russian ECC88 equivalent), four matched 6E5P and four matched KT88 to be ready for tests.
Bandersnatch,
thanks for your interventions too. I would probably move some arguments in a dedicated thread, in order to be able to better develop them.
Considering other interventions, moderators time ago suggested me not to consider and not to reply anymore to the italian guy (don't feed the... you know) that is just chasing me everywhere on the forum to say everything I post will sound bad without even trying to understand how it works. I extend the suggestion to you guys too.
thank you for your intervention. I've always been used to simulate before building, simply because I always had more time for the former than the latter, so it has always been an optimisation of the time to build/mod. The guitar and bass amps I've built (a dozen) and modded (a hundred) followed the same process. The Hi-Fi amps I've built (three) followed the same process too, even more in depth being simulators more detailed when working linearly. But the final goal is always share ideas and build something better than I did before.
Concerning the fact that "common" tubes can be easily triode connected, whilst it is for sure true, I've found (again, only on LTSpice by now) this a-g1 feedback to be far superior in terms of performances compared to 100% a-g2 feedback, in terms of rp, linearity, available power, and the possibility to adapt it to the driver.
dgta, SpreadSpectrum, smoking-amp,
thank you for your interventions. You all gave me great support to optimize details.
I've already bought four matched 6N23P (russian ECC88 equivalent), four matched 6E5P and four matched KT88 to be ready for tests.
Bandersnatch,
thanks for your interventions too. I would probably move some arguments in a dedicated thread, in order to be able to better develop them.
Considering other interventions, moderators time ago suggested me not to consider and not to reply anymore to the italian guy (don't feed the... you know) that is just chasing me everywhere on the forum to say everything I post will sound bad without even trying to understand how it works. I extend the suggestion to you guys too.
Last edited:
To Tubelab
I am sure that your circuit are working fine.
Never said something other.
But you have simulated then built then tested then presented.
And this make a great difference.
I have in my lab two AP ( Sys One and Sys Two), five tube tester ( Sofia included), three generators, plus some other stuff.
Then I have around 5.000 pcs of Holco resistors (H4,H2 and H8, in the past around 20.000 pcs) , 10.000 pcs of Resista ( thre different wattage of WK serie), around 2.000 caps plus (I don't remeber the number) the tubes plus around ten tube power amp as protos ( sometimes I change the configuration)
And some dozens of kilos of OT ( always protos)
And I always test live on desk the circuit.
Then I write here
AudioReview - Il sito ufficiale della rivista
the projects with all description and test lab, words on paper.
Around 30 different ones that aren't (probably) the most excited circuit in the world but they are running fine and everyone can build them with the results as describe dand expected
Ciao
Walter
I am sure that your circuit are working fine.
Never said something other.
But you have simulated then built then tested then presented.
And this make a great difference.
I have in my lab two AP ( Sys One and Sys Two), five tube tester ( Sofia included), three generators, plus some other stuff.
Then I have around 5.000 pcs of Holco resistors (H4,H2 and H8, in the past around 20.000 pcs) , 10.000 pcs of Resista ( thre different wattage of WK serie), around 2.000 caps plus (I don't remeber the number) the tubes plus around ten tube power amp as protos ( sometimes I change the configuration)
And some dozens of kilos of OT ( always protos)
And I always test live on desk the circuit.
Then I write here
AudioReview - Il sito ufficiale della rivista
the projects with all description and test lab, words on paper.
Around 30 different ones that aren't (probably) the most excited circuit in the world but they are running fine and everyone can build them with the results as describe dand expected
Ciao
Walter
In my post #118 I was stating that my UNSET circuit had been through rigorous testing and listening for nearly 2 years. I actually prefer it to my TSE-II 300B amp. The higher peak cathode current capability (1.4 amps) of the large TV sweep tubes make for more realistic dynamics, especially percussive sounds.
I guess that I failed to state that the circuit shown in post #1 of this thread IS my circuit that has been modified for use with a KT88 in UL, something that I have neither tested or simulated. I have never been a big fan of UL, but it does have it's place. I can switch on the UL mode with cathode feedback in my SSE KT88 amp when I want to crank up some Pink Floyd to near clipping and hear the bass inside the house across the street. For more "normal" listening the same amp runs an EL34 in triode.
I do have versions both tested and simulated that use a CCS load on the input pentode to get higher gain. I do not need this much gain in a HiFi amp, so the CCS is not used in my UNSET. It is useful in a guitar amp that will be used well into clipping, and my little 4 watt guitar amp uses a CCS loaded pentode in it's input stage.
The schematic in post #1 states 30 Wrms at 1.34% THD. I would venture to guess that putting a current probe across the 1 ohm resistor might reveal excessive peak cathode current, at that level.
As stated by others LTSpice can not do a decent simulation of an OPT. Even if it could, where would you get the data to feed it? Leakage inductance and distributed winding capacitance are the big hitters in the phase shift department, and can differ widely between two identical OPT's.
Models for tubes can vary from "useless" to "pretty good" but again that's only when you are using them in a "normal" manner. Grid currents are usually poorly modeled, and vary considerably from tube to tube even two of the same type and brand.
The simulator knows nothing about secondary breakdown in semiconductors, which is a limiting factor in this design. Getting 30 watts from a single KT88 requires putting about half the total dissipation into the mosfet. This is easily done in the simulator and will give good results.
The mosfet will short when driven hard under these conditions which will result in full B+ across your tube / OPT under zero bias conditions. I have seen one tube shatter, and another EXPLODE when I did this, but I have a real big power supply (650 volts @ 1.7A) with a slow current limiter and 1000 uF of output capacitance.
These things can only be learned by playing with real parts.
I guess that I failed to state that the circuit shown in post #1 of this thread IS my circuit that has been modified for use with a KT88 in UL, something that I have neither tested or simulated. I have never been a big fan of UL, but it does have it's place. I can switch on the UL mode with cathode feedback in my SSE KT88 amp when I want to crank up some Pink Floyd to near clipping and hear the bass inside the house across the street. For more "normal" listening the same amp runs an EL34 in triode.
I do have versions both tested and simulated that use a CCS load on the input pentode to get higher gain. I do not need this much gain in a HiFi amp, so the CCS is not used in my UNSET. It is useful in a guitar amp that will be used well into clipping, and my little 4 watt guitar amp uses a CCS loaded pentode in it's input stage.
The schematic in post #1 states 30 Wrms at 1.34% THD. I would venture to guess that putting a current probe across the 1 ohm resistor might reveal excessive peak cathode current, at that level.
As stated by others LTSpice can not do a decent simulation of an OPT. Even if it could, where would you get the data to feed it? Leakage inductance and distributed winding capacitance are the big hitters in the phase shift department, and can differ widely between two identical OPT's.
Models for tubes can vary from "useless" to "pretty good" but again that's only when you are using them in a "normal" manner. Grid currents are usually poorly modeled, and vary considerably from tube to tube even two of the same type and brand.
The simulator knows nothing about secondary breakdown in semiconductors, which is a limiting factor in this design. Getting 30 watts from a single KT88 requires putting about half the total dissipation into the mosfet. This is easily done in the simulator and will give good results.
The mosfet will short when driven hard under these conditions which will result in full B+ across your tube / OPT under zero bias conditions. I have seen one tube shatter, and another EXPLODE when I did this, but I have a real big power supply (650 volts @ 1.7A) with a slow current limiter and 1000 uF of output capacitance.
These things can only be learned by playing with real parts.
I have 330 mA peak, 160 mA at idle. Are those 230mA Ik limit a peak value?
Yes, it's around 35W to be dissipated. The reason why I went that far with the feedback is to see how it would perform pushing the driver to its limits. I will test the amp with different ratios of feedback, in order to understand the sonic effect of different percentages, from your 10% to "RCA&UNSET's son" 20 and 30%.
Thanks, well noted that a fast fuse is needed on the cathode of the tube to save the tube, the OPT and... me!
Forget TO220 stuff like the 3P50, use TO247 as minimum size such as IXTH10P50 on a fairly large heatsink, probably comparable to a third or half of a single side 4U-400 heatsink just to maintain case temp < 70°C. There is also a good chance that some fuse is not fast enough to prevent damage to other parts in case of mosfet short failure mode, better add a crowbar as added insurance. We are too far behind George.
The IXTH10P50 has a Crss spec of 135 pF with 25 volts across the part. The Crss will vary with the voltage across the device, and is probably much higher at low voltages, but there is no graph in the data sheet. The driver must be capable of driving this directly without distortion or slew rate limiting. A buffered driver, yet another mosfet follower between the driver and the p-fet, or a CCS with an output from it's source may be needed.
I use a mosfet circuit often called a gyrator in my guitar amp, primarily so I can pound current into my unorthodox tone control. See the Q1 circuit in it's schematic.....yes, this is more food for the simulator.
I built the guitar amp completely by the pile of parts and a soldering iron method. I never simmed it. Trying to simulate an amp that will operate in heavy compression / saturation is pretty useless anyway.
The biggest part that I have tried is the Fairchild FQPF9P25 at 27 to 35 pF. It is a isolated TO-220 which is worse case for Theta j-c, but I'm only dissipating about 5 watts. The same part is available in larger packages, but they are hard to find now. I have tried the FQA9P25 with a theta j-c of 0.83. The FQPF9P25 gets me to a 43.75 KHz upper 3dB point with -0.33 dB @ 20 KHz. This is with a Hammond 1628SEA 5K ohm OPT wired to reflect 2500 ohms to the tube (8 ohm load on the 16 ohm tap). This is much better than the same OPT measured when directly driven by a "600 ohm" HP204C oscillator. I get 60+ KHz in the push pull driven amp when I use a toroidal OPT from Plitron.
The THD remains flat between 1.2 and 1.3% from 200 Hz to 25 KHz. All specs measured at 10 watts output VS the usual 1 watt or less commonly used to test amps.
There has been some debate as to the stability of this design. It is quite docile even when hammered with a guitar preamp in my UNSET form, since there is only enough local feedback in each individual stage to make pentodes exhibit triode like curves. Increasing the gain and / or wrapping feedback from the driver plate to the input tube's grid could provoke instability without GNFB or any feedback from the OPT secondary.
I have not seen any in both cases in some limited testing, but I tend to prefer low levels of feedback even if the THD is higher. It just sounds better.
Some OPT's (like the Hammond I am using) have a resonance in or near the high end of the audio frequency band. This presents a phase reversal and an unusual load for the output stage, even into a fixed load resistor. A multi way speaker can present a really weird load due to it's crossover network. Drive the test amp with a sine wave while varying the drive level and look for bursts of oscillation on the leading edge of the peaks. Again a sim will not find any of this stuff.
Last edited:
Post your proposed idea and the simulator guys will play with it, or I may even build one to see how it really works.
I have taken a pair of TV sweep tubes to the 250 to 300 watt range in an UNSET board with the two channels driven out of phase. The mosfet in the screen regulator blew when I pushed it into hard clipping. I got some big IXYS fets to fix this, but they will not be needed in the final amp due to it's multi stage power supply. I'm looking at 500 WPC with 4 to 6 tubes, 4 if I can get away with it, since the amp will never really see much use beyond 50 WPC.
Fortunately the mosfet sees it's peak current when there is little voltage across it, and peak voltage at near zero current. The papers discussing the mechanism for secondary breakdown failures in mosfets show that this may not matter much, it's localized hot spots and thermal lag that kill the fets.
The peak currents seen in a simulator, or even in real amp testing with sine waves don't compare to what's seen when a big kick drum transient tries to instantly reverse the woofer's cone travel at high power. Stick a scope across a small unbypassed resistor in the cathode of your favorite amp, then crank up some bass heavy music and watch. Different speaker / amp combos react differently.
I built a solid state car amp for a design project when I returned to college for an engineering degree in 1989 after becoming an engineer without one. It used a push pull boost converter to generate +/- 35 volts, and +/- 50 volts, which ran several typical class AB amp boards. I used this amp for several years on my daily 80+ mile commute during which time the power supply blew up three times. Every time it was the same song.....ZZ Top's Sharp Dressed Man. Coincidence?
I did a lot of experiments concerning real loudspeaker impedance about 15 years ago. I was pretty well convinced that the instantaneous impedance could approach zero, and maybe even go negative on a big efficient woofer under extreme conditions, before I realized just how far I had ventured down a bottomless rabbit hole, and gave up. This was to become my PHD thesis, but Motorola pulled funding for PHD's and MBA's before I got in too deep. A pro audio woofer can generate several volts at several hundred mA just from sharing a closed cabinet with a similar driven woofer. Just assume that your amp needs to drive anything from a near short, to about a hundred ohms (guitar speaker at resonance).
My main discovery.....don't put a Shure SM57 microphone INSIDE a kick drum.
I have taken a pair of TV sweep tubes to the 250 to 300 watt range in an UNSET board with the two channels driven out of phase. The mosfet in the screen regulator blew when I pushed it into hard clipping. I got some big IXYS fets to fix this, but they will not be needed in the final amp due to it's multi stage power supply. I'm looking at 500 WPC with 4 to 6 tubes, 4 if I can get away with it, since the amp will never really see much use beyond 50 WPC.
Fortunately the mosfet sees it's peak current when there is little voltage across it, and peak voltage at near zero current. The papers discussing the mechanism for secondary breakdown failures in mosfets show that this may not matter much, it's localized hot spots and thermal lag that kill the fets.
The peak currents seen in a simulator, or even in real amp testing with sine waves don't compare to what's seen when a big kick drum transient tries to instantly reverse the woofer's cone travel at high power. Stick a scope across a small unbypassed resistor in the cathode of your favorite amp, then crank up some bass heavy music and watch. Different speaker / amp combos react differently.
I built a solid state car amp for a design project when I returned to college for an engineering degree in 1989 after becoming an engineer without one. It used a push pull boost converter to generate +/- 35 volts, and +/- 50 volts, which ran several typical class AB amp boards. I used this amp for several years on my daily 80+ mile commute during which time the power supply blew up three times. Every time it was the same song.....ZZ Top's Sharp Dressed Man. Coincidence?
I did a lot of experiments concerning real loudspeaker impedance about 15 years ago. I was pretty well convinced that the instantaneous impedance could approach zero, and maybe even go negative on a big efficient woofer under extreme conditions, before I realized just how far I had ventured down a bottomless rabbit hole, and gave up. This was to become my PHD thesis, but Motorola pulled funding for PHD's and MBA's before I got in too deep. A pro audio woofer can generate several volts at several hundred mA just from sharing a closed cabinet with a similar driven woofer. Just assume that your amp needs to drive anything from a near short, to about a hundred ohms (guitar speaker at resonance).
My main discovery.....don't put a Shure SM57 microphone INSIDE a kick drum.
Last edited:
A small piece of what I read showed MOSFET die's with what looked like bullet holes in 'em. Localized heating and its positive FB effects that I think caused a current concentration and then failure. Ultimately I began to worry about how they work at all...and so contemplate the efficacy of running a huge one at light current.
cheers,
Douglas
cheers,
Douglas
I am running a configuration similar to the "Kulish driver" that smoking amp mentioned (aka "supertriode" on Broskie's tubecad site). In push-pull this has two mosfets per channel continuously dissipating 30W each (~150V@200 mA). I've used 2SK-3746 TO-247 mosfets rated at 1500V and 110W. The high voltage rating keeps them well away from the danger area. They are mounted on fan cooled heatsinks to keep the size down.
The 2015 datasheet shows the Crss graph - https://www.littelfuse.com/~/media/electronics/datasheets/discrete_mosfets/littelfuse_discrete_mosfets_p-channel_ixt_10p50p_datasheet.pdf.pdf. Zintolo was dissipating 35W @ 230ma so I assume operating Vds to be ~ 150V, probably an FQA9P25 would be the smallest usable fet but a 500V part seems a better bet.
Thanks, I like that tone control, I've used first time in a Carmen Ghia (should come from old Sunn or London City IIRC) and was simple and effective.
I sim guitar stages to see how hard I need to hit them to start saturating, and how they do it plus how they react to frequency shaping. I started using unorthodox stages (cold bias stages with higher currents) and balance differently the signal on different stages thanks to that.
Thanks again George, I will try this FQPF9P25 with some 3°C/W heat sinks I have at your suggested feedback ratio.
Thanks, indeed the idea is to try different feedback ratios just to see how it sounds vs THD plots. With shunt feedback I've always chosen lower values in reality vs best THD on sims, but usually around 15 to 20% depending on tubes.
Thanks, I will do it too.
Hi tikiroo, I haven't found any thread of yours about that on the forum. I'd be interested to read about it, if possible.
At idle it's around 215V and 165 mA, so it swings from 0 to 430V: a 500V is needed with this configuration.
But of course it will be needed to verify how much is too much, and most probably George's values will be the sweet spot (plus minus a variation tube to tube).
I think Bandersnatch wants to propose something like this:
Shown by Nelson Pass at the bottom of the page here: Pass LabsThe Sweet Spot - Pass Labs
Yah, something very like that...but different. LOL Will get some idea up today I hope.
Douglas
In the last few years of my RF design work I helped to design a broadband RF power amplifier for use in a hand held two-way radio for public safety use (police walkie-talkie). It needed to operate from 100 MHz to 1 GHz and produce 10 watts of continuous power for a minimum of 7 minutes with no touchable part of the radio reaching 60 degrees C. This is no easy task, and previous radios barely made it. An efficient RF power amp is needed, and efficiency tends to go down as the bandwidth goes up.
Previous versions used three different transistor die in a single custom package mounted on a single heat sink. Each transistor was tuned to a section of the frequency range, and since only one frequency can be used at a time, 3 packages and three heat sinks are not needed. Myself and a lab tech were granted a US patent for this idea.
As the US TV spectrum kept getting sliced, diced, and sold of to the highest paying cellular phone carrier, new remnant slivers of discontinuous spectrum became available for public safety. We needed a transmitter that could adapt to ANY frequency and power level from 100 MHz to 1 GHz. We used a GaN fet that had been in military service at the 80 watt power level in short pulse radar up to 2 GHz.
This worked well until it didn't. Parts could be transmitting in test boards for days, weeks, or longer into all sorts of overload, temperature, and voltage extremes, then just blow, often under mundane conditions. Bullet holes are a good description of what we saw. Some looked more like a bombing range after practice day (craters missing). We set up a Flir video camera with a $20K lens on a granite slab to capture these failures on video at the die level on unpackaged parts.
In 2014 I accepted a buyout and left Motorola. At that time we had captured two such blow up events on video. Several people from Mot and the part vendor were feverishly working on the issue. Three or four years later the Motorola APX-8000 began shipping. About 1 year of that would be consumed by certification testing once the product was deemed ready for production.
I think I am going to fuse the P-channel, cathode drive circuit. The load for these amps is a horn loaded DL10x, so I don't think any damage is going to happen if one side of a PP output TX goes from full on to off...but I have a feeling it will not be q quiet failure.
George, I have to wonder, is a big, lightly run FET more or less likely to pop vs a smaller one run closer to its limits? From a bullet holes perspective that is...
cheers,
Douglas
George, I have to wonder, is a big, lightly run FET more or less likely to pop vs a smaller one run closer to its limits? From a bullet holes perspective that is...
cheers,
Douglas
Last edited: