Hi,
I bought some boards which I plan to use as small mostly class A amplifiers,
I also plan to use them for mid/treble with a separate AB amplifier for bass.
But further research revealed them to be the DX amplifier from this forum.
The major variation with the schematic is a Ccomp of 100pF compared
to the the DX amplifiers 22/27pF used on a BD139 VAS transistor.
This is a change of two octaves and around 12dB of open loop gain
in the mid treble regions. It also implies an extra 12dB of gain margin.
But is this needed ? Or is it being overly slugged with a baseball bat
for stability ? How stable is the 22/27pF variant ? Don't know myself.
Any informed opinions would be appreciated.
rgds, sreten.
(somebody please fix my crap title spelling and change Dx->DX)
I bought some boards which I plan to use as small mostly class A amplifiers,
I also plan to use them for mid/treble with a separate AB amplifier for bass.
But further research revealed them to be the DX amplifier from this forum.
The major variation with the schematic is a Ccomp of 100pF compared
to the the DX amplifiers 22/27pF used on a BD139 VAS transistor.
This is a change of two octaves and around 12dB of open loop gain
in the mid treble regions. It also implies an extra 12dB of gain margin.
But is this needed ? Or is it being overly slugged with a baseball bat
for stability ? How stable is the 22/27pF variant ? Don't know myself.
Any informed opinions would be appreciated.
rgds, sreten.
(somebody please fix my crap title spelling and change Dx->DX)
Last edited:
This is sadly how it works nowdays, they take a design, change a component value or two and call it theirs.
THe rule is to use as small value compensation cap as possible.
To keep the amplifier stable.
This can be determined exactly only through an oscilloscope.
Using 100p when such value is not needed
is called overcompensation.
To keep the amplifier stable.
This can be determined exactly only through an oscilloscope.
Using 100p when such value is not needed
is called overcompensation.
Yup. They say 'your idea no good, can't be productive' and the next thing you see the exact product appear out of China, or somewhere else in the world.....and people wonder why I don't disclose all of my IP here.🙄
I suppose imitation could be looked upon as flattery.🙂
I suppose imitation could be looked upon as flattery.🙂
Last edited:
Dodgy compensation capacitors
It could also have been a necessary step, now we know that BD139s ain't BD139s any more. I've also had to replace specified 47pF and 68pF comp. capacitors in some cheap amps with typically 82 pF. Usually, there had been substitutions in the small signal transistors too. The amps. could only just have been marginally stable when sold new. 😡
DX himself claimed that marginally stable amps were "the good ones" so it may not be all bad to have an occasional fried amp in some people's quest for "sonics".😱
With kit supplies though, any risk of oscillation is probably much worse than overcompensation. Caveat emptor!
It could also have been a necessary step, now we know that BD139s ain't BD139s any more. I've also had to replace specified 47pF and 68pF comp. capacitors in some cheap amps with typically 82 pF. Usually, there had been substitutions in the small signal transistors too. The amps. could only just have been marginally stable when sold new. 😡
DX himself claimed that marginally stable amps were "the good ones" so it may not be all bad to have an occasional fried amp in some people's quest for "sonics".😱
With kit supplies though, any risk of oscillation is probably much worse than overcompensation. Caveat emptor!
Hi,
What do you mean ? Hfe is not up to specification ? So the miller effect
is less ? And therefore the capacitor needs to go up ? But I thought
the miller effect in the VAS compensated automatically Hfe variation .....
Guess I might need to get a proper circuit emulator out ....
rgds, sreten.
Member
Joined 2009
Paid Member
I have used two methods to pick Cdom, one using my ears, the other using a 'scope. The use of a 'scope I have mixed feelings about - I used the square wave test signal from the 'scope (used for tuning up scope probes) which has extremely steep fall-rise times. I'm not convinced that this is a good way to do it unless you are really just looking for a bomb proof system - the steep edges are extreme for an audio amp.
I was much happier choosing Cdom by ear. My amp wasn't that different from the DX so my experience may be relevant. I found a value of Cdom that was too low and a value that was too high was quite audible. The too-low value (22pF) was thin sounding and in one of my amps I also had some sibiliance with low values. The too-high value (139pF) was dead sounding. The middle value (in the range of 47pF), the Goldilocks value, sounded just right.
I used BD139 for the VAS and I've read some people who find a better sound can be achieved with a device that has higher performance (in particular, lower Cob).
I was much happier choosing Cdom by ear. My amp wasn't that different from the DX so my experience may be relevant. I found a value of Cdom that was too low and a value that was too high was quite audible. The too-low value (22pF) was thin sounding and in one of my amps I also had some sibiliance with low values. The too-high value (139pF) was dead sounding. The middle value (in the range of 47pF), the Goldilocks value, sounded just right.
I used BD139 for the VAS and I've read some people who find a better sound can be achieved with a device that has higher performance (in particular, lower Cob).
Good old jims audio! They don't sell these kits anymore...
I find it funny that people slag ebay shops like this off. Okay it's not their design. But I have no way of making PCBs that are that good myself. Jims audio PCBs are double sided thick PCB thick copper, solder masked and silk screened for essentially very little price! they are fantastic quality as well. Where can I buy a DX amp PCB right now and get it in less than a weeks time for a tenner? I personally think they fantastic it's just a shame that their change the schematics, if they just sold the PCBs and did not call them their own which jim's audio doesn't what's the problem? I think it's a great resource to the community.
I find it funny that people slag ebay shops like this off. Okay it's not their design. But I have no way of making PCBs that are that good myself. Jims audio PCBs are double sided thick PCB thick copper, solder masked and silk screened for essentially very little price! they are fantastic quality as well. Where can I buy a DX amp PCB right now and get it in less than a weeks time for a tenner? I personally think they fantastic it's just a shame that their change the schematics, if they just sold the PCBs and did not call them their own which jim's audio doesn't what's the problem? I think it's a great resource to the community.
Last edited:
Hi
I disagree that borderline stability improves sonics, as in specifically intending to operate near instability. If the GBW of the amplifier is high enough, you can cap it well below that instability point and still have plenty to work with. This way you have a high 'sonics' amplifier and an unconditionally stable one to boot.🙂 Design/PCB layout becomes much more important and specific tho.
Some think to bring a amp to the "edge" (undercompensate) will improve sonics.
Wrong , and it will not always improve slew - which WILL have a impact on sonics.
I seen a slow ft amp with impressive slew and a "fast amp" with mediocre slew. More to it than just changing a cap. Devices and topology are bigger factors.
OS
Wrong , and it will not always improve slew - which WILL have a impact on sonics.
I seen a slow ft amp with impressive slew and a "fast amp" with mediocre slew. More to it than just changing a cap. Devices and topology are bigger factors.
OS
'No argument with your views on undercompensation guys, The point is that right or wrong, many "modders" do it.
There are several threads here, discussing the change in specs. over more than 40 years production of BD139/140. Check out even the data sheets of second sources like On-semi, to realise they are different in the gain/bandwidth curves and Ft. Cob etc. while all are said to be different again to the original Philps data sheets/products.
If builders use similar parts to the relevant prototypes, all may still be fine but in reviving old designs and with so many manufacturers supplying these parts, there's surely going to be problems.
Regarding the original DX values, you may have to dig early in the original thread to find relevant comments from Carlos. Greg Erskine may recall such details as he also drew a PCB version for it. I also have a pair of Jim's very different kits, unopened, with the tiny blue boards JSDX2008. The supplied caps also appear to be 100pF ceramic, hopefully NPO. The drivers are TIP41/2 for some reason, perhaps unsinked diissipation.
There are several threads here, discussing the change in specs. over more than 40 years production of BD139/140. Check out even the data sheets of second sources like On-semi, to realise they are different in the gain/bandwidth curves and Ft. Cob etc. while all are said to be different again to the original Philps data sheets/products.
If builders use similar parts to the relevant prototypes, all may still be fine but in reviving old designs and with so many manufacturers supplying these parts, there's surely going to be problems.
Regarding the original DX values, you may have to dig early in the original thread to find relevant comments from Carlos. Greg Erskine may recall such details as he also drew a PCB version for it. I also have a pair of Jim's very different kits, unopened, with the tiny blue boards JSDX2008. The supplied caps also appear to be 100pF ceramic, hopefully NPO. The drivers are TIP41/2 for some reason, perhaps unsinked diissipation.
can you show the scope pics for a 5kHz or 10kHz square before after passing through the complete amplifier with it's passive input filters in place?
Did you carry out this test?
Peaking on medium rise time transients shows up clearly and gives rise to exaggeration of transients in the audio signal. JLH actually describes how to use the scope and the feedback to trim the squarewave to best sounding.
And this is why I don't disclose any IP on this site (or any other for that matter).... It's hard enough to try and make a living from audio electronics without giving away all the tricks of the trade 😱
Hi
Do you know where I can find a page with a full description this JLH scope test ?
Thanx
Paul
Last edited:
Typical LIN amp , Cdom can be 47p - 100p , BD or fairchild. Biggest factor is input degeneration. In the fastest "flavor" of this amp (VERY high Ft devices) , instability can only be obtained with insufficient degeneration (<47R).
With mje340/50 (20Mhz Ft-20 pf Cob) VAS with 33Re , I use 47p
KSA1381/3503(150Mhz Ft/2pf Cob ) - 68pf and I increase the input pair degeneration to 100R.
For a subwoofer amp voltage stage where it does not matter - 100p
I test with a large 10R/10uF load. You can see the leading edge of a square wave go from slight overshoot to damped ringing. If Cdom and/or degeneration is too low your Zobel network's resistor will begin to smoke 😀 ...
I did this by accident with 10R input pair resistors (misread them - oops!)
OS
With mje340/50 (20Mhz Ft-20 pf Cob) VAS with 33Re , I use 47p
KSA1381/3503(150Mhz Ft/2pf Cob ) - 68pf and I increase the input pair degeneration to 100R.
For a subwoofer amp voltage stage where it does not matter - 100p
I test with a large 10R/10uF load. You can see the leading edge of a square wave go from slight overshoot to damped ringing. If Cdom and/or degeneration is too low your Zobel network's resistor will begin to smoke 😀 ...
I did this by accident with 10R input pair resistors (misread them - oops!)
OS
Hi,
Input pair (2n5401) is 100R degenerated. bd139 VAS to tip41/42 drivers
to 2sc5200/2sa1943 outputs. Zobel but no output inductor. no protection.
Looking at the Baby AKSA, (68pF no degeneration), and the real DXamp,
(22/27pf with 100R degeneration), swayed towards using only 33pf ????
rgds, sreten.
Input pair (2n5401) is 100R degenerated. bd139 VAS to tip41/42 drivers
to 2sc5200/2sa1943 outputs. Zobel but no output inductor. no protection.
Looking at the Baby AKSA, (68pF no degeneration), and the real DXamp,
(22/27pf with 100R degeneration), swayed towards using only 33pf ????
rgds, sreten.
DX started out with..... 5.6 pF 😱 !!
After feedback and mods, it was raised to the final figure by about post 90 so your estimate might be risky but still within an established range.
Do you have access to a scope and generator to verify stability with even a 1kHz square wave response?
Feeling for a hot Zobel resistor is not my idea of stability testing, more just being certain you indeed have a disaster!
After feedback and mods, it was raised to the final figure by about post 90 so your estimate might be risky but still within an established range.
Do you have access to a scope and generator to verify stability with even a 1kHz square wave response?
Feeling for a hot Zobel resistor is not my idea of stability testing, more just being certain you indeed have a disaster!
Hi,
Presumably this was not due to terminal stability issues for the initial value,
otherwise how could it start at the 5.6pF point ? but was that with bd139 ?
My concern is the phase compensation in the feedback network (22pF) as
in the DXamp might be completely out of kilter with using a 100pF Ccomp.
Probably go with 33pF or maximum 47pF as I understand it currently.
rgds, sreten.
The thread is over 1,600 posts long and I can't check it all out so I can't say and I doubt we will know without asking the questions directly. However, the usual quick and dirty method is to go in with a 100pF cap and verify stability. From that point, reduce capacitance as Lineup suggests until sound is noticeably thin and go back one or so values.
This may be enough for experienced tweakers but to do it professionally is with a scope, verifying that square waves are as close to the input form, (inserted after the input filter) as possible.
Due to stray and semiconductor capacitances, the theoretical calculations we make of bandwidth, gain etc. are usually wide of the mark and smaller values may well be better. DX's layout for the 5.6 pF value was only changed after it was transferred to an Eagle produced PCB by Greg Erskine, so there is variation by the construction method, I would conclude.
Again, your choice will no doubt be fine at 33 or 47pF, even with the slowest BD139s around. The highest Cob I saw was 20 pF, Stinius has elsewhere quoted a lowest of 7pF. so there is considerable difference possible. Only reduce the initial Comp. value if indications support it. There is no other way to be certain without experience of the particular PCB and the specific part types.
Hi,
The PCB layout is identical to G.E.'s layout for the DXamp.
Consequently I think going with 33pF is the most logical option.
(I don't have a scope, which is why I want informed opinions.)
100pF seems too much and IMO will slug slewrate too much,
nevermind what seems the pointless reduction of hf feedback.
I also plan to swap C10 and C12 for my mid/treble application.
I will run it in high bias class AB, not full class A, say a 1/4 A,
meaning at "normal" volume levels as opposed to full, which is
unfortunately not possible in a block of flats, it should remain
in class A all the time. (1/4A = biased to half full A current.)
rgds, sreten.
The PCB layout is identical to G.E.'s layout for the DXamp.
Consequently I think going with 33pF is the most logical option.
(I don't have a scope, which is why I want informed opinions.)
100pF seems too much and IMO will slug slewrate too much,
nevermind what seems the pointless reduction of hf feedback.
I also plan to swap C10 and C12 for my mid/treble application.
I will run it in high bias class AB, not full class A, say a 1/4 A,
meaning at "normal" volume levels as opposed to full, which is
unfortunately not possible in a block of flats, it should remain
in class A all the time. (1/4A = biased to half full A current.)
rgds, sreten.
Last edited:
- Status
- Not open for further replies.