Have you guys tried these?
.MODEL 2SK1058 NMOS (VTO=403.969M KP=20U L=2U W=29.7482M GAMMA=0 PHI=600M
+ LAMBDA=184.988F RD=60.8251M CBD=2.56138N IS=10F CGSO=1.13517N CGDO=1.13517N
+ TOX=0 NSUB=0 TPG=1 UO=600 RG=50 RDS=1MEG )
.MODEL 2SJ162 PMOS (VTO=842.193M KP=20U L=2U W=21.3317M GAMMA=0 PHI=600M
+ LAMBDA=20.7067M RD=837.199M CBD=2.96862N IS=10F CGSO=1.13517N CGDO=1.13517N
+ TOX=0 NSUB=0 TPG=1 UO=600 RG=50 RDS=1MEG )
.MODEL ECX10N16 NMOS (LEVEL=1 VTO=-66.415M KP=20U L=2U W=76.3794M GAMMA=0
+ PHI=600M LAMBDA=3.44617M RD=244.181M RS=178.517M IS=10F CGSO=100P CGDO=100P
+ TOX=0 NSUB=0 TPG=1 UO=600 KAPPA=200M RG=367.902 RDS=16.0012K )
.MODEL ECX10P16 PMOS (LEVEL=1 VTO=56.1959M KP=20U L=2U W=49.9039M GAMMA=0
+ PHI=600M LAMBDA=16.6941M RD=458.28M RS=214.674M CBD=1.00674N IS=10F CGSO=100P
+ CGDO=100P TOX=0 NSUB=0 TPG=1 UO=600 KAPPA=200M RG=35.8793 RDS=1MEG )
.MODEL 2SK1058 NMOS (VTO=403.969M KP=20U L=2U W=29.7482M GAMMA=0 PHI=600M
+ LAMBDA=184.988F RD=60.8251M CBD=2.56138N IS=10F CGSO=1.13517N CGDO=1.13517N
+ TOX=0 NSUB=0 TPG=1 UO=600 RG=50 RDS=1MEG )
.MODEL 2SJ162 PMOS (VTO=842.193M KP=20U L=2U W=21.3317M GAMMA=0 PHI=600M
+ LAMBDA=20.7067M RD=837.199M CBD=2.96862N IS=10F CGSO=1.13517N CGDO=1.13517N
+ TOX=0 NSUB=0 TPG=1 UO=600 RG=50 RDS=1MEG )
.MODEL ECX10N16 NMOS (LEVEL=1 VTO=-66.415M KP=20U L=2U W=76.3794M GAMMA=0
+ PHI=600M LAMBDA=3.44617M RD=244.181M RS=178.517M IS=10F CGSO=100P CGDO=100P
+ TOX=0 NSUB=0 TPG=1 UO=600 KAPPA=200M RG=367.902 RDS=16.0012K )
.MODEL ECX10P16 PMOS (LEVEL=1 VTO=56.1959M KP=20U L=2U W=49.9039M GAMMA=0
+ PHI=600M LAMBDA=16.6941M RD=458.28M RS=214.674M CBD=1.00674N IS=10F CGSO=100P
+ CGDO=100P TOX=0 NSUB=0 TPG=1 UO=600 KAPPA=200M RG=35.8793 RDS=1MEG )
Haha I was smiling when I saw the situation. Those "numbers" do not represent the real sound quality. Of course people can build and listen to the circuits, but then the question is "how good are their ears?", otherwise all will sound the same.
I had no interest building the circuit because from the "structure" of the schematic I had a feeling that it will not sound good (I decided to wait for somebody to say "this is the best sounding amp in the world" then I will build it). But I can see from SPICE analysis (my own way) if it has potential or not at all. But I will do that only if there is asc file, PCB design, and I am interested enough
I don't know for sure whether OS has experience with designing MOSFET amp (I think this is the first time he designed a mosfet amp in DIYA). MOSFET is just different animal than a BJT!
Many of the models out there are okay with the exception that they don't model subthreshold conduction. That's what Hegglun's models do.
Sure. The basic problem is that the standard Spice model for a FET is based on a little FET in an IC because Spice was first intended to model ICs.
The variable capacitance that causes most of the distortion in a power amp is NOT well modelled, especially in the particular models you used where you have Tox=0, which turns off variability.
Don't have to take my word for this, it can be checked in the LTSpice help for "M. MOSFET" model.
The VDMOS model that Bob uses is specifically intended for power FETs and is much closer. It still does not model sub-threshold conduction and Bob's book explains how to combine the VDMOS model with yet another model to simulate distortion more accurately.
Bob is the real expert in this and has an actual FET amp that has been well tested and documented to back his claims.
My experience was to help Toni with his own real and well documented amp.
A model like the one you used produced wildly inaccurate simulations, for the reasons I explained.
After Toni moved to a VDMOS model, as I recommend to you also, the simulated distortion was fairly close to the measured data.
You can check this and his measurement system in his thread, both the thread and the measurement system are excellent.
Note that this is mainly about distortion simulation, the fact that a model works OK in some other areas does not mean it is accurate.
Best wishes
David
Last edited:
Any PMOS or NMOS model, like these, is probably worthless for distortion simulation.
Especially if Tox=0.
The first two look like the models Christophe used and his results are, unfortunately, quite unrealistic, use Bob's models instead.
The second two look even worse.
Best wishes
David
Last edited:
Sorry, I shouldn't have commented as I have not looked into the FET modeling situation much. Dave should know what he's doing.
AFAIK Cordell's models have okay capacitance modeling, but don't have subthreshold conduction. I'm looking closer at Hegglun's models, though it doesn't help that I don't have access to the article they're meant to supplement.
AFAIK Cordell's models have okay capacitance modeling, but don't have subthreshold conduction. I'm looking closer at Hegglun's models, though it doesn't help that I don't have access to the article they're meant to supplement.
The level 1 models, like those, that I have tested were bad but the level 12 PMOS or NMOS model is the EKV model that Bob recommends.
So this probably needs a FET model expert to tell you exactly which levels are useful.
Best wishes
David
Keantoken, Thanks for the link
Last edited:
I've built a few of these amps designs now and have also tried a VSSA. I've decided to wait to post my observations until they are mounted in proper chassis so I can eliminate noise due to poor layout. In my situation I am playing on 105 db/watt speakers with horn mids and tweeters (Klipsch Chorus). Terry's test was on Bose speakers that are less efficient so they will be playing at a different volume level than I do. Amplifiers will behave differently under these conditions.
He liked his VSSA amplifier. I found my VSSA amp lacked in the tweeter range at comfortable listening levels. If I turned it up it sounded excellent but my drink was moving around my coffee table and the dog was hiding under my bed.
I've found over the years that my horn tweeters need high voltage power supplies to bring them to life at lower volume levels. Lower efficiency speakers will get away with lower voltage supplies because the volume level the amplifier will be played at will be higher. Crossover distortion is much more prominent in my situation as well. I'm looking forward to having a chassis ready and trying different input/output combinations. I'm pretty sure different gain inputs are going to have an effect on the sound in my situation.
While I agree with you that simpler designs have less chance of colouring sound, it doesn't always work. To stop trying other designs because one design works good for one type of speaker really doesn't seem like a great idea for people with another type of speaker. Between that and the fact that I enjoy building things and learning more as I go, I hope OS keeps coming up with more.
Ok, so i simulated DVSSA with the power MOS Cordel models.
The catastrophic results in the same condition were not so...catastrophic:
0.000728% HD 1000Hz
0.004651% HD 4 harmonics at 10 000Hz
If no mistake, i'm still very happy.
Slew-rate the same (>1000µs) (please remove the filter inside the diamond to figure out, it should be a non sense to measure the slew-rate of the low pass input filter, witch is not in the gnfb, witch allows something like <200V/µs, giving a 5 time margin against TIM;-)
There is a lot of ideas to learn. Just i miss listening reports.
Oh, btw, i just got an idea.
It is a hard charge to drive multiple output MOS fets for a simple VAS, due to their gate capacitances.
Adding a driver in serial adds a pole in the loop, that i don't like.
Why not parallel multiple VAS, each with their output stage, instead ? I tried this in sims, and it seems to work.
The catastrophic results in the same condition were not so...catastrophic:
0.000728% HD 1000Hz
0.004651% HD 4 harmonics at 10 000Hz
If no mistake, i'm still very happy.
Slew-rate the same (>1000µs) (please remove the filter inside the diamond to figure out, it should be a non sense to measure the slew-rate of the low pass input filter, witch is not in the gnfb, witch allows something like <200V/µs, giving a 5 time margin against TIM;-)
So do I.
There is a lot of ideas to learn. Just i miss listening reports.
Oh, btw, i just got an idea.
It is a hard charge to drive multiple output MOS fets for a simple VAS, due to their gate capacitances.
Adding a driver in serial adds a pole in the loop, that i don't like.
Why not parallel multiple VAS, each with their output stage, instead ? I tried this in sims, and it seems to work.
Last edited:
Well, I'm confused now. As sometimes happens, I encountered a glaring contradiction in the datasheet. In the ECW20N20 datasheet, the Id/Vgs chart shows no more than 2.5S transconductance - but the transconductance chart shows up to 5.5S transconductance! Furthermore, the transconductance doesn't droop at high currents like it should.
I could model both charts, which would essentially give two Lfets, one with twice the transconductance.
So now it makes perfect sense why Ian's models don't match the datasheet - the datasheet is nonsense.
I could model both charts, which would essentially give two Lfets, one with twice the transconductance.
So now it makes perfect sense why Ian's models don't match the datasheet - the datasheet is nonsense.
Hi Jeff,
Great review. I actually do most of my listening on JBL speakers. They are less than 100db but fairly efficient. I wonder if your Klipsch might be bass heavy. The only ones I've heard were in a showroom and had huge folded horn cabinets. The bass was impressive to say the least but of course we weren't listening at low levels either.
I must admit, I don't usually do A/B testing at really low levels. I will try that next time. I generally listen at low levels throughout the day while I work, and switch between amps as the day progresses. Looking forward to more of your impressions.
Blessings, Terry
Great review. I actually do most of my listening on JBL speakers. They are less than 100db but fairly efficient. I wonder if your Klipsch might be bass heavy. The only ones I've heard were in a showroom and had huge folded horn cabinets. The bass was impressive to say the least but of course we weren't listening at low levels either.
I must admit, I don't usually do A/B testing at really low levels. I will try that next time. I generally listen at low levels throughout the day while I work, and switch between amps as the day progresses. Looking forward to more of your impressions.
Blessings, Terry
With a VAS and driver stage, the gain is multiplied. With a paralleled VAS, the gain only adds. You would need many more transistors to have the same effect.
However if adding drivers, despite the extra gain, forces you to chop the gain back off again by adding additional stability compensation, then it may be more effective to add parallel VASes, and add only as much extra gain as you can without needing to increase the compensation.
How about a thumbnail visual of what you are saying, a picture is worth a 1000 quotes
The transistor(s) who's collector goes to the OPS, try using 4 in parallel instead of just 1. That's the basic idea. At best, distortion will be divided by the number of paralleled transistors.
Although I think that Early effect and Hfe will prevent this from working very well. It should still work for load-dependent distortions, and if cascoded, it may work for voltage-dependent distortions as well.
Although I think that Early effect and Hfe will prevent this from working very well. It should still work for load-dependent distortions, and if cascoded, it may work for voltage-dependent distortions as well.
Yes I remember you said JBL now. 2 pairs.Hi Jeff,
Great review. I actually do most of my listening on JBL speakers. They are less than 100db but fairly efficient. I wonder if your Klipsch might be bass heavy. The only ones I've heard were in a showroom and had huge folded horn cabinets. The bass was impressive to say the least but of course we weren't listening at low levels either.
I must admit, I don't usually do A/B testing at really low levels. I will try that next time. I generally listen at low levels throughout the day while I work, and switch between amps as the day progresses. Looking forward to more of your impressions.
Blessings, Terry
Klipsch' are a little bass heavy with a lower powered amp. They have a 15" driver in them. They are meant for big open areas. With around a 200 watt amp they really come to life. The extra headroom really shines at lower volumes.