Hello Gnobuddy,
In answer to your questions:
1) Feedback does tend to make transfer functions more angular and, when used in large amounts, produces high linearity followed by sharp clipping. Here, however, I am using it in small amounts as a tool to control the curvature of the transfer function. In the first example, the loop gain around Q1 is about unity and in the second about three.
The effects of small amounts of feedback around a curve with mainly second order distortion are to raise the input level required for a given percentage of second harmonic and to add a small amount of third harmonic. The first effect helps with signal to noise ratio, while the second effect actually mimics the behaviour of a triode. Too much feedback would give the wrong harmonic spectrum, but a useful amount can be applied without harmful effect.
Remember that a triode incorporates its own internal feedback. 1/mu of the output voltage is added to the effective grid voltage. This is why the anode output resistance of a triode is relatively low, compared to a pentode where this feedback loop is broken by the presence of the screen grid. The first circuit uses feedback in the same way to reduce the modeled "anode" resistance.
2) The difference in the overload behaviour of the two circuits is caused by the placement of the zero signal point on the transfer function curves.
There are two clipping mechanisms in place; clipping on positive inputs due to the source resistance and the grid diode, and clipping on negative inputs due to cutoff. In the first circuit, cutoff is modelled explicitly by the cascode arrangement, while in the second circuit, the clipping characteristic arises from the natural transistor curve sharpened by feedback.
In between clipping points, we have the natural rise in second harmonic due to the curvature of the characteristic. As clipping occurs, there is a gradual transition to a squareish waveform. The exact placement of the clipping points determines the mark-space ratio of this waveform and hence the harmonic spectrum. The same effects will occur with a triode, depending on the supply volts, anode current and anode load.
Arthur
Thanks for your reply, Arthur. Clearly you've put a lot of thought and effort into this, and I think you've ended up with a rather unique approach to cracking this particular nut.
The only real downside is the number of components and degree of complexity needed to mimic just one single triode. If only someone would put together an integrated circuit containing a fistful of your emulated triodes!
-Gnobuddy
The only real downside is the number of components and degree of complexity needed to mimic just one single triode. If only someone would put together an integrated circuit containing a fistful of your emulated triodes!
-Gnobuddy
I follow a couple of these threads with interest. As a professional guitarist of nearly 50 years I have owned and used lots of different amps. On tours and concerts you often get offered a selection of various ones (2 or 3) to use for shows. They can be anything from well maintained to unserviced shite.
My 1st amp back in 1967 was a Goldentone 60 (locally built unit) which GJB will be conversant with I expect followed by a Strauss 300w Polka/Warrior which was complete overkill
. Playing more RnB a la Stax, about 1972 I bought a 2nd hand Vox AC30 Top Boost which I used for around 4 years before getting a blackface Fender super reverb which was much better than the Vox for playing country music which I was doing at that time. I moved to a Mk1 Mesa Boogie with no graphic about 1981 and at the same time bought an original Roland Cube 60 to keep as a practice amp at home - so early it was the original wire wrap terminals !!
I replaced its speaker with a Celestion and did a couple of small gigs with it and was impressed with its sound - it was also a lot easier to cart around weighing only 12Kg, less strain on a back. After getting the circuit schematic from Roland (at great expense for a photocopy) I was intrigued with the use of JFETs for the preamp - even using the overdrive channel it was passable - and carting around the Boogie became less and less desirable. As electronics has been my hobby since my high school years I have always done most of my amp maintenance myself along with trying various mods suggested by different tech-heads.
So started the mods of the Cube 60:
1) I removed the spring reverb unit and IC driver and built in an Aux send via a split load drain resistor, and kept the reverb return for and Aux IN so that I could run an Echo or Delay unit in a loop.
2) I raised the back to back diodes after the Volume control from ground via a small value resistor to round off the square wave edges when overdriving on 10.
3) Changed the tone stack to the original FMV style using the Fender Twin values.
I have now used this amp (I actually have a 2nd almost identical model with the first 2 mods incorporated) for over 30 years. It was almost indestructible however the airlines won in the finish.
As I need to build a new PCB for it I figured I might as well look into some of the other JFET/MOSFET ideas on various forms. I’ve tried a few variations of the trioderiser, a couple of the runoffgroove circuits and AMZ ideas, plus some hybrid ideas of my own experimentation using a JFET or MOSFET as a source follower direct coupled to a BJT to achieve 30x gain as well as the FET in common source mode direct coupled to BJT emitter follower for low impedance out. I also tried variations of values in both the FMV and James/Baxandall style tonestacks. I’m now looking into LoudThuds JFET Bender as a front end.
After trying both the Cube and a Fender Deluxe (reissue model even) side by side at a concert sound check, the main difference I noticed tonally was the Fender was brighter probably less mids due to the tonestack scoop, where the Cube was a little darker in sound similar to the AC30s. Whether that is due to the speaker difference I’m not sure but level wise there wasn’t much to choose from.
I’m starting to think that the Roland engineers who developed the original Cubes (not the latest ones with the modelling) were getting very close to the Fender/Vox sounds. I’d be very interested in trying GJB’s SS VOX next time I’m in Perth. I could probably get a Fender Deluxe, Pro, or Twin to A/B it with at a sound check.
My 1st amp back in 1967 was a Goldentone 60 (locally built unit) which GJB will be conversant with I expect followed by a Strauss 300w Polka/Warrior which was complete overkill
. Playing more RnB a la Stax, about 1972 I bought a 2nd hand Vox AC30 Top Boost which I used for around 4 years before getting a blackface Fender super reverb which was much better than the Vox for playing country music which I was doing at that time. I moved to a Mk1 Mesa Boogie with no graphic about 1981 and at the same time bought an original Roland Cube 60 to keep as a practice amp at home - so early it was the original wire wrap terminals !! I replaced its speaker with a Celestion and did a couple of small gigs with it and was impressed with its sound - it was also a lot easier to cart around weighing only 12Kg, less strain on a back. After getting the circuit schematic from Roland (at great expense for a photocopy) I was intrigued with the use of JFETs for the preamp - even using the overdrive channel it was passable - and carting around the Boogie became less and less desirable. As electronics has been my hobby since my high school years I have always done most of my amp maintenance myself along with trying various mods suggested by different tech-heads.
So started the mods of the Cube 60:
1) I removed the spring reverb unit and IC driver and built in an Aux send via a split load drain resistor, and kept the reverb return for and Aux IN so that I could run an Echo or Delay unit in a loop.
2) I raised the back to back diodes after the Volume control from ground via a small value resistor to round off the square wave edges when overdriving on 10.
3) Changed the tone stack to the original FMV style using the Fender Twin values.
I have now used this amp (I actually have a 2nd almost identical model with the first 2 mods incorporated) for over 30 years. It was almost indestructible however the airlines won in the finish.
As I need to build a new PCB for it I figured I might as well look into some of the other JFET/MOSFET ideas on various forms. I’ve tried a few variations of the trioderiser, a couple of the runoffgroove circuits and AMZ ideas, plus some hybrid ideas of my own experimentation using a JFET or MOSFET as a source follower direct coupled to a BJT to achieve 30x gain as well as the FET in common source mode direct coupled to BJT emitter follower for low impedance out. I also tried variations of values in both the FMV and James/Baxandall style tonestacks. I’m now looking into LoudThuds JFET Bender as a front end.
After trying both the Cube and a Fender Deluxe (reissue model even) side by side at a concert sound check, the main difference I noticed tonally was the Fender was brighter probably less mids due to the tonestack scoop, where the Cube was a little darker in sound similar to the AC30s. Whether that is due to the speaker difference I’m not sure but level wise there wasn’t much to choose from.
I’m starting to think that the Roland engineers who developed the original Cubes (not the latest ones with the modelling) were getting very close to the Fender/Vox sounds. I’d be very interested in trying GJB’s SS VOX next time I’m in Perth. I could probably get a Fender Deluxe, Pro, or Twin to A/B it with at a sound check.
I'm Back!
Hi people... been absent for a while... won't bore you with the details.
I have (mostly) read the comments since my last post... some good thoughts in there. Thanks.
I have got a little more active I the last month or so and being doing some more development... aided by a supply of Toshiba FET's that I have been able to secure... 2SK117, 170, 369, 30's... etc.
This has been very helpful as they are:-
a) Graded at manufacture (BL, V, G, etc. versions) albeit that I still INDIVIDUALLY test and select them, and,
b) Especially made for audio applications
Importantly they have helped me refine the design process of biasing in "triode" mode as per the article I referred to some time back
(Google "Dimitri-Danyuk FET Triode Emulator")
I have developed an XL sheet making it easy to get all the components values I need by simply specifying Vgs, Idss, and the supply voltage.
BTW - There are also some "wider" considerations (not mentioned in the above) that need to be considered.
While I could go in to detail... this process seems to work... I have looked at numerous FFT spectrums on my scope... and, much more importantly, had the critical (positive) evaluation of several (good) musicians... unlike myself! :-(
I have also refined the output stage further with (again) a good supply of 2SK1058 and 2SK2221 (audio) power MOSFETs...
I could continue... but perhaps I'll wait for any feedback / questions about specific items before continuing...
Cheers
GJB
Hi people... been absent for a while... won't bore you with the details.
I have (mostly) read the comments since my last post... some good thoughts in there. Thanks.
I have got a little more active I the last month or so and being doing some more development... aided by a supply of Toshiba FET's that I have been able to secure... 2SK117, 170, 369, 30's... etc.
This has been very helpful as they are:-
a) Graded at manufacture (BL, V, G, etc. versions) albeit that I still INDIVIDUALLY test and select them, and,
b) Especially made for audio applications
Importantly they have helped me refine the design process of biasing in "triode" mode as per the article I referred to some time back
(Google "Dimitri-Danyuk FET Triode Emulator")
I have developed an XL sheet making it easy to get all the components values I need by simply specifying Vgs, Idss, and the supply voltage.
BTW - There are also some "wider" considerations (not mentioned in the above) that need to be considered.
While I could go in to detail... this process seems to work... I have looked at numerous FFT spectrums on my scope... and, much more importantly, had the critical (positive) evaluation of several (good) musicians... unlike myself! :-(
I have also refined the output stage further with (again) a good supply of 2SK1058 and 2SK2221 (audio) power MOSFETs...
I could continue... but perhaps I'll wait for any feedback / questions about specific items before continuing...
Cheers
GJB
I hope it wasn't anything serious. At any rate, glad to hear you're better now, and welcome back!
While you were gone, I started to build a give-away guitar amp for a friend. For various reasons, it had to be relatively small, as light as possible, and cheap to build - which meant solid state. Along with my build/R&D project, I started posting some of my ideas and results to an existing diyAudio thread.
For some reasons I can guess at, diyAudio moderators chose to break my posts off from that thread, and move them into a new thread, which they titled "Tube Emulation & EQ". That thread then became a place to share and collect ideas on making a good-sounding solid-state guitar amp.
One of the more remarkable things that happened was that another diyAudio member posted links to some really ground-breaking work by a Russian engineer who goes by "KMG" on the Internet. To my ears, KMG has gone further, more successfully, than anyone else when it comes to coaxing FETs to behave like vacuum triodes. Far beyond Dmitri Danyuk's method of coaxing an JFET to have a Child-Langmuir transfer characteristic, for instance.
Here are some links to various bits of KMGs work:
1) Let's start with a sound-clip: KMG's MOSFET re-creation of a Bogner Ecstacy clean channel: http://milas.spb.ru/~kmg/files/proje...mg_xtc_mix.mp3
2) Here's a discussion of KMGs work on the SS Guitar forum: LND150 tube emulator attempt.
3) Here's another discussion, about a MOSFET recreation of a JCM 800 guitar preamp: Fet version of the JCM800
4) KMG's personal website, with schematics and other documents (some is in Russian only): Main page
Personally, I've not experimented with KMG's circuit ideas, but another diyAudio member (mjd_tech) did, and he's posted some of his findings to the "Tube Emulation & EQ" thread. You may find some of his results interesting as well.
I think mjd_tech's first post describing his work with KMG's MOSFET stages is on page 37, post #364 (a little way down the page): Tube Emulation & EQ
-Gnobuddy
Hi Gnobuddy... my main problem is that work keeps getting in the way of my hobby! :-(
Anyways... I have actually seen some of KMG's work a couple of months back via another contact. He has done a lot of work and produced some really good stuff. Clearly he has gone the way of using high voltage FETs - operating at the same levels as valve amps - and emulating the waveforms (and voltage levels) they produce.
Early in my development I made the decision to go the low voltage route and (broadly and in an oversimplified way) my design was scaled at 10:1 - in both in voltages and component values. i.e. 300 volts became 30 volts; a 1M ohm pot became a 100K pot; a 1nf cap became a 10nf cap; etc... (which has some advantages btw)
However, and importantly, I adopted the "Dimitri" FET biasing model and chose actual component values accordingly. Recently I have been aided by the availability of a good number of NOS Toshiba (specialist audio) FETs which have produced much more predictable (and controllable) results.
Not saying which is the "better" method but I have been looking more at FFT plots than waveforms on my DSO - particularly looking to optimise the (desirable) "even harmonic" sound.
Having now done PCB layouts and other detailed design I am now at the point where I perhaps would like to start selling some kits. Just thinking about the best way to go about that. Suggestions?
If I can get that going then perhaps I can give up my "day job"... and here we are back at my opening comment!
Cheers,
GB
Anyways... I have actually seen some of KMG's work a couple of months back via another contact. He has done a lot of work and produced some really good stuff. Clearly he has gone the way of using high voltage FETs - operating at the same levels as valve amps - and emulating the waveforms (and voltage levels) they produce.
Early in my development I made the decision to go the low voltage route and (broadly and in an oversimplified way) my design was scaled at 10:1 - in both in voltages and component values. i.e. 300 volts became 30 volts; a 1M ohm pot became a 100K pot; a 1nf cap became a 10nf cap; etc... (which has some advantages btw)
However, and importantly, I adopted the "Dimitri" FET biasing model and chose actual component values accordingly. Recently I have been aided by the availability of a good number of NOS Toshiba (specialist audio) FETs which have produced much more predictable (and controllable) results.
Not saying which is the "better" method but I have been looking more at FFT plots than waveforms on my DSO - particularly looking to optimise the (desirable) "even harmonic" sound.
Having now done PCB layouts and other detailed design I am now at the point where I perhaps would like to start selling some kits. Just thinking about the best way to go about that. Suggestions?
If I can get that going then perhaps I can give up my "day job"... and here we are back at my opening comment!
Cheers,
GB
I am utterly clueless about the world of business, and if you were to follow my advice, you would probably go bankrupt immediately.
However, I might be able to point you at some people who are better qualified to help: have you asked around on the Aussie Guitar Gearheads (AGGH) forum? Several members there build and sell valve guitar amps. Perhaps they could give you some useful information on getting started?
The AGGH forum is here: Aussie Guitar Gear Heads - Index
-Gnobuddy
The key is finding a reasonable sized gap in the market and servicing it well.
Firstly, don't expect to make money by being the cheapest.
Secondly, identify and manage/accept the risks to your business.
Finally, follow GnoBuddy's advice, have a read through AGGH. There's a number of people having a good crack at the amp business.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.