Leolabs said:
in many cases - yes
I could give you 10 really good hifi amplifiers
using only one of these output pairs:
2SK1058 - 2SJ162
2SK1530 - 2SJ201
outputs
... and driven by 'collectors' of VAS.
MATCHED PAIRS Lateral K1058 J162 - fair price from diyAudio.com member
Data and INFO of 2SK1058 2SJ162 - Renesas.com
Naturally we have to have not Too Low current to for drive ANY POWER MOSFET.
Does not matter if they are laterals.
If you use only < 5mA, you are not looking for hifi performance ..
Ask John Curl,
one of our most experienced members, when comes to amplifiers.
He will tell you this, like lineup tells you now:
=================
The Cure
for Capacitance
is Current
=================
Regards
to you Leolabs
..... lineup admires such members having personal courage to ask
..... lineup think such members will learn more than others
..... lineup feels sorry for those knowing all
..... lineup feels sorry for them afraid to ask what they don't now
..... lineup pity them with foolish pride, pretending they already know
..... lineup pity them to afraid to ask
..... whether they call themselves Nelson Pass, Christer or whatever
..... at our forum of www.diyaudio.com
Happy New Year, Leolabs
lineup
does not know it all but sure is on way getting there a couple of questions left ... only Hi,
the VAS is run in ClassA and is usually fed from the LTP.
At high frequencies, the load the LTP sees is the capacitance of the VAS//collector load.
At these same high frequencies the load the VAS sees is the capacitance of the output FET//CCS effective load.
The problem is that the LTP wants to see a small capacitance at the VAS and the VAS wants to see a small capacitance at the FETs.
This is a conflict and the usual solution is impedance conversion. That's another way of saying, provide more current.
Use an emitter follower and the problem is divided by the current gain through the EF.
the VAS is run in ClassA and is usually fed from the LTP.
At high frequencies, the load the LTP sees is the capacitance of the VAS//collector load.
At these same high frequencies the load the VAS sees is the capacitance of the output FET//CCS effective load.
The problem is that the LTP wants to see a small capacitance at the VAS and the VAS wants to see a small capacitance at the FETs.
This is a conflict and the usual solution is impedance conversion. That's another way of saying, provide more current.
Use an emitter follower and the problem is divided by the current gain through the EF.
I think, there is not necessary to use EF stage to drive MOSFETs. I made power amplifier with 4pairs of IRFP240/9240, which has no EF, but there is 100mA biased VAS stage, using MJE15032/33. Of course, You need darlington for the Vgs multiplier to handle this current, and EF between the input stage and the VAS, but this is not so critical.
sajti
sajti
One must remember that the input capacitance of a lateral FET is much lower than that of a vertical FET. This makes the drive requirements more relaxed.
Of course you still need a relatively low impedance driver, but there are plenty of topolologies that can do this without burning huge amounts of current and without using emitter followers.
A topology that I've had a lot of success with is that developed by David Tilbrook in the '80's, for the AEM6000 power amp. He uses a symmetrical differential VAS. In my take on this topology I burn just 6mA on each side of the differential VAS, and achieve 0.0048% THD (measured) at 10KHz (100W into 8 Ohms).
Cheers,
Suzy
Of course you still need a relatively low impedance driver, but there are plenty of topolologies that can do this without burning huge amounts of current and without using emitter followers.
A topology that I've had a lot of success with is that developed by David Tilbrook in the '80's, for the AEM6000 power amp. He uses a symmetrical differential VAS. In my take on this topology I burn just 6mA on each side of the differential VAS, and achieve 0.0048% THD (measured) at 10KHz (100W into 8 Ohms).
Cheers,
Suzy
i noticed that little peak in the open loop response where the phase crosses the 180 degree mark, as well as your phase plot with feedback. rule of thumb is if you cross 180 with gain, it may oscillate, which is why it did when built (you still had gain at 1.1mhz, even after adding the bypass cap to the input). it would be interesting to see the gain/phase plot now that you added compensation to the VAS.
nice design...... i might try it myself.......
nice design...... i might try it myself.......
Driving lateral MOSFETs
I agree with Suzyj's comment.
As long as you are driving a relatively small number of output devices, you can use the input capacitance of the output stage, in conjunction with the high output impedance of the voltage gain stage to set the dominant pole of the open-loop transfer function. You have to set it somewhere
Provided that you are not trying to use too much overall feedback, it works very well. In fact, I regard this as one of the advantages of MOSFETs over BJTs in the output.
Best wishes
Dave
I agree with Suzyj's comment.
As long as you are driving a relatively small number of output devices, you can use the input capacitance of the output stage, in conjunction with the high output impedance of the voltage gain stage to set the dominant pole of the open-loop transfer function. You have to set it somewhere
Provided that you are not trying to use too much overall feedback, it works very well. In fact, I regard this as one of the advantages of MOSFETs over BJTs in the output.
Best wishes
Dave
David Tilbrook ?
Now that's a name I haven't seen around for a fair while !
Thanks for those great designs of yours like the Ultra Fidelity preamp,5000 and 6000 etc. that so many of us constructed years ago . Actually, I know of a member who is still using his 6000 to drive the low end of his biamped Infinity speakers.
Regards
Alex
P.S.
A couple of years ago SuzyJ tested my old 5000 that had been rescued from a friend's garage. The channel where the electros hadn't dried out still tested .0062%
Now that's a name I haven't seen around for a fair while !
Thanks for those great designs of yours like the Ultra Fidelity preamp,5000 and 6000 etc. that so many of us constructed years ago . Actually, I know of a member who is still using his 6000 to drive the low end of his biamped Infinity speakers.
Regards
Alex
P.S.
A couple of years ago SuzyJ tested my old 5000 that had been rescued from a friend's garage. The channel where the electros hadn't dried out still tested .0062%
Last edited:
Hi Dave T and welcome to DIY.
I do still have an original AEM6000 that's fully operational and sounding as good as the day it was built - which is now > 22 years ago. A truly great design if I may say so.
And yes, the input C Vs VGS is reasonably linear on the old laterals - above 2V. Most amps these days tend to use miller compensation schemes which is really good for THD but at the expense of slew rate.
RRR Sandy K.
I also met Phil Wait on Sydney Harbour last week on NYE. I asked him about the ETI 470 design along with a few other projects that he was involved with back in the 70's. His advice today is to use one of the new LME 498xx chips - to drive lateral mosfets.
Hi
I agree for the most part with this but remember that lateral fets have much lower Gm than verticals and require a larger variance in Vgs for a given conductance. Since most of CISS is Cgs, laterals require a larger 'Vgs signal'. Laterals may have less overall capacitance charging required than verticals but not by as much as it appears; refering to source follower output.
As long as you are driving a relatively small number of output devices, you can use the input capacitance of the output stage, in conjunction with the high output impedance of the voltage gain stage to set the dominant pole of the open-loop transfer function. You have to set it somewhere
When the input capacitance as seen by the driving stage, that being CISS and CRSS, changes due to changing Vds, how does this affect the location of the pole and is any change significant?
Last edited:
There is no need to use an EF
Many circuits built show driving the LFET off the collector give great results. Its is not even nessasary to use a special device (ie 2sc2911) as a mje340 performs well.
Well the 5000 has copped some criticism recently. I think, unjustifiably. It was designed as a affordable and accessible diy project that would still give much better than average performance for it's day, and I think it achieved that.
We also had some pretty strong (but sensible) restrictions on us at the time. Our brief was that we had to have very good reasons to use a component that was not a standard stock item in the most popular electronics components stores, but we did break the rules quite often
The Hitachi power MOSFETs were a case in point. I stumbled across them by accident when I was searching for devices to use in a switched-mode power supply, and could immediately see their potential for use in audio power amps. As far as I am aware there were no other serious power amp designs using MOSFETs at the time I started working on it, so it was a bit of a gamble.The 6000 is an order of magnitude better, and if carefully built will still compare well by todays standards, imho. I have only just seen SuzyJ's take on the 6000 and she has done a great job. I no longer have a 6000, so I may build a pair based on her take, just out of curiosity
VHF man said:
I have been nervous about ICs ever since I designed the 5000 preamp
CBS240 said:
There are several reasons that it is not such a problem - within reason. It also depends on amplifier topology.
Your are right that in these devices Ciss is dominated by Cgs, but this works in your favour, because Cgs is very weak function of Vds. In the vast majority of topologies, it can be regarded as independent of Vds, although it is rather large.
The physics of these devices is very complicated so without getting into the messy details I prefer to compensate the gate-source impedance with a small series RC, while the p-channel devices often need a larger capacitance. This then then gives a relatively balanced load to the output of the VAS of around 2nF for every pair of typical output devices, and it is virtually independent of Vds. Cgd, does vary with Vds, roughly proportional to 1-K*Vds^1/2 for most devices (where K is an constant), but Cdg is roughly two orders of magnitude smaller than Cgs.
Samuel Jayaraj said:
I
Sorry Samuel, I cut off my reply on the previous post - I need to hone my forum sklls
I have some designs that may be of interest, but they would need modernising - a lot of the components are no longer available.
Also, in replying to CBS240 I forgot to say, that it is necessary to use a fairly large current in the VAS. In the 6000 I think I used around 20mA in both sides of a symmetrical differential VAS, but I will have to check the circuit to be sure.
best wishes
Dave
Sorry Samuel, I cut off my reply on the previous post - I need to hone my forum sklls
I have some designs that may be of interest, but they would need modernising - a lot of the components are no longer available.
Also, in replying to CBS240 I forgot to say, that it is necessary to use a fairly large current in the VAS. In the 6000 I think I used around 20mA in both sides of a symmetrical differential VAS, but I will have to check the circuit to be sure.
best wishes
Dave
Oh , so it is your design ?? I am going to use (some of) it for running my EF2 (below 1). I was very impressed by the full "monty" (single ended IPS/full comp 2nd/3rd stage) ... except I had TOO much loop gain using modern fairchild devices(NO MJE340/50's). It does manage PPM distortion @20K with the modern devices.
Since I will be starting off with either EF driven laterals or a standard BJT OPS , I'll prototype my simpler version and then do the full circuit (littlefish with new fairchilds) If I like the first one. Awesome circuit , Dave ... even using it for "building blocks" is exciting.
OS
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.