stability issues output triple and high freq extension ?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I would like to go with output triple with ksa1381 pre drivers only for a reason not to load the VAS stage as much as the direct driver connection. But I have a question that if the frequency extension is done to 300KHz or 500KHz does the amplifier gets unstable?
Any sonic degradation with triple?
 
Absolutely none. The trick with the triple is to decouple either the pre or
driver/pre from the output rails. A simple 10R/47uf will do here.

I've seen member thimios test 200khz square waves , the only burn out
was by using "slow" output devices (21193/4's).

SQ wise , a triple will sound better .... as the VAS never has to negotiate
the impedance swings like it has to with a EF2. EF3 "loves" to drive high
current / high output.

OS
 
I think triples sound better. You will see a lot of differences of opinion here, though. Whether it's better or worse is topology dependent - what works for one circuit doesn't for another. For my pet front end toplologies it does.

Stability is technically a little worse. You're more prone to need base stoppers compared to a double, and there is additional phase shift through the extra stage which degrades loop gain margin. I deal with that by taking the lead compensation from the predriver emitter rather than the output. In some circles you'll be hanged, beheaded or burnt at the stake for that but....
 
Absolutely none. The trick with the triple is to decouple either the pre or
driver/pre from the output rails. A simple 10R/47uf will do here.

I've seen member thimios test 200khz square waves , the only burn out
was by using "slow" output devices (21193/4's).

SQ wise , a triple will sound better .... as the VAS never has to negotiate
the impedance swings like it has to with a EF2. EF3 "loves" to drive high
current / high output.

OS
but the pre driver works in class AB though. With quiescent current as much as 20ma through the resistor which connects its compliment. How do we bias that pre driver in class A?
 
but the pre driver works in class AB though. With quiescent current as much as 20ma through the resistor which connects its compliment. How do we bias that pre driver in class A?

No, class A is right up to the outputs. (below)
Only the last plot switches off (AB plot).

PS - I have mine set 2.5ma pre /8-10ma driver. At the AB xover point , even
huge currents with a 1R load keep pre/driver in A , never passing less than 8ma (main driver I).
Edit - Drivers will pass as much as 150ma to the output pairs , but R41 will always pass
>8ma - still class A.
OS
 

Attachments

  • EF3-a-ab.jpg
    EF3-a-ab.jpg
    146.3 KB · Views: 399
Last edited:
I get around the triple by using super high gain drivers and outputs. The reflected impedance to the Vas/transconductance stage gets divided by the current gain of the following stages so a triple is better that way. The triple has a fair amount more propagation delay increasing loop stability issues and like the smart guy above, take the feed back from earlier in the amp for extreme frequencies and the audio band at the output. This has been done forever and is a fair solution. I always use fast high gain transistors for any output stage. Like now, the drivers and outputs are both 160MHz Hfe>150 devices.

In the above circuit R62 R63 I consider pure evil. All those base resistors do is slow propagation and lower current gain while leaving the bandwidth of the two transistors hooked together by the 150 ohm able to yet behave as a single port oscillator. I always take base resistors out and fix the issue some other way. Just sayin... Also that 150 ohm.... should be a diode or something and a much smaller resistor.
 
In the above circuit R62 R63 I consider pure evil. All those base resistors do is slow propagation and lower current gain while leaving the bandwidth of the two transistors hooked together by the 150 ohm able to yet behave as a single port oscillator. I always take base resistors out and fix the issue some other way. Just sayin... Also that 150 ohm.... should be a diode or something and a much smaller resistor.

The 150R is standard in two of the best triples around - the H/K680 and 990.

R62 and 63 are for the huge Cob of the drivers .... you see they are also
outputs. None of the 500+ circuits (oPS's) like this has ever oscillated unless
driven by a "bad" IPS with insufficient margin.

OS
 
Hi OS,

Since you have investigated/simulated/built more amp variations than most, do you have any comments on the attached circuit?
It is a decent sounding Fostex Lab series amp circa late 1970s.
Would you reckon the large number of small value caps scattered throughout are there to try and keep a problem EF3 stable?
Would that be due to slow devices? Bad layout? Other?

John
 

Attachments

  • scancrop3.jpg
    scancrop3.jpg
    264 KB · Views: 349
Hi OS,

Since you have investigated/simulated/built more amp variations than most, do you have any comments on the attached circuit?
It is a decent sounding Fostex Lab series amp circa late 1970s.
Would you reckon the large number of small value caps scattered throughout are there to try and keep a problem EF3 stable?
Would that be due to slow devices? Bad layout? Other?

John

No , it looks like a Z3900 sansui from the 70's. 3 gain stage IPS.

C605,609, (and others) are the lag, miller. All the R/C's paralleled
in the op stage are to keep that CFP driver stable.

This type IPS was like my kypton V - the hardest to get stable by itself.
Also the hardest to make "play nice" with the EF3.

This amp does not just have one dominate pole at the VAS , like most
typical amps. Add this fact to the CFP OPS - kind of "shaky" (hard to
compensate).

OS
.
 
Kypton V (below) was the only IPS that required removing the driver
base-stoppers. Sumaudioguy is correct in this case.

Having multiple poles allowed the triple's high gain to "misbehave".
This IPS also has 120db open loop gain , requiring less gain at the (first) input
pair and/or healthy degeneration of all the stages.\

In the end , stability and overload were stellar (D3-4). VERY hard circuit
to tame !!
OS
 

Attachments

  • KVcomp.jpg
    KVcomp.jpg
    162.3 KB · Views: 325
Thanks for the insight. You mean the 3900Z receiver? I'll have a look.

Reason I ask is I've had one of these Lab 600's for years and love it. Just got my hands on a few more but they all have little "issues".

One had a DC offset on the output and stayed in protect. Turns out the dual input transistor was out of whack - replaced with matched ksc1815's and all is well.

Another has been "worked on" and has an assortment of mixed, incorrect transistors which led me to thinking this design probably isn't too tolerant of random substituting without reworking the compensation... haven't even switched that one on yet!

Oh, another one has an approx. 1khz oscillation in one channel!

So it looks like I'll have my work cut out for me....

I see the Fostex does use a 33R+Diode rather than the 150R as mentioned by Sumaudioguy but still has the bypassed base stoppers.

Sorry for the digression...
 

Attachments

  • WP_20151009_22_19_37_Pro.jpg
    WP_20151009_22_19_37_Pro.jpg
    173.6 KB · Views: 281
Last edited:
Is the mg6333 simply two dies of the mg6330 in one package? Have seen two dies in one package with double the rating several times.

From post #7, are you saying replacement of the 150 ohm with a smaller resistor and diode is absolutely not an improvement or simply commenting on history? My testing shows the smaller resistor with diode in series works better.
 
Kypton V (below) was the only IPS that required removing the driver
base-stoppers. Sumaudioguy is correct in this case.

Having multiple poles allowed the triple's high gain to "misbehave".
This IPS also has 120db open loop gain , requiring less gain at the (first) input
pair and/or healthy degeneration of all the stages.\

In the end , stability and overload were stellar (D3-4). VERY hard circuit
to tame !!
OS
can you consider emitter follower for Q9 and Q10? or Q9 and Q10 are just acting like buffers?
 
Hey ostripper, got a question. When you indicate "stability and overload were stellar" does that mean you can short the output and the amp will blow the line fuse to protect? That is how stable all my amps are with the line fuse being the only protection needed. Shorted output blows the line fuse. If you have that then I applaud the design. If if smokes on a short that would be loss of stability and not very stellar.
 
I've done both , without "21'st century protection" ... fuses blow quick
and all survives.
21'st involved ..... I can do anything , speaker might move a mm ... red
led flashes , I know I screwed up.

That how good Vzaichenko's "21'st century" is. DC and overcurrent can't
even exceed SOA.

This is a truthful assessment of what screwups I have done.

PS - these test's of the protection were of my stupidity - I hail to
Vzaichenko's creation.
PS - can't smoke a semi - that's a global condition with these designs.
OS
'
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.