advertisement


New Regulator

I have now reduced the incursions to the bottom ground plane to just three instances. In yellow below
Not really worried about the long horizontal one as it does not obstruct line of sight down the length of the ground plane.
But the other two are across line of sight for the op-amp.

bottom2_zpspdfypy7d.png



The one below chows C7 the op-amp supply pin decoupling below the board.
I chose this position as it keeps it closest to the supply pins.

Considering bringing C7 to the top layer to tidy the ground plane under the op-amp. Probably place it above the op-amp close to supply pin 7.
Any thoughts?

bottom1_zpsrqvtxism.png
 
Tony,

Are you using the "frying pan" method for soldering the SMT's using solder paste and a stencil or doing them by hand? If the former, then the more SMT's, the better - saves hand soldering. Otherwise, the SMT transistors are more difficult to solder by hand than the SMT resistors (IMHO) - they are so light that they blow away more easily! Also, nothing really wrong with TO-92 and they tend to have higher power dissipation limits if that's important in the circuit. Your choice.

Kit
 
Not sure what the "frying pan" method is :D
I use a Metcal SP200 with a 1\4" chisel tip, 0.38mm leaded solder wire and a flux pen. The existing board was done with this method including the SMD AD797 and a couple of 0603 size.

I have previously done ESS Sabre 9018 chips the same way although very tricky and I now have a hot air station and paste for the Sabre but can easily put down 0603 parts with a conventional iron so don't bother with the hot air gun on Mite.

Don't think dissipation is a problem in the small signal transistors so are happy to suck it and see. Swapping them to SMD looks neater, and tidies the ground plane?

Assuming I have the right SMD transistor pads they look the same size as 0603 resistor pads and the device is twice the size of a 0603 resistor.
I have BC850, BC860 and the 2SK209-Y jfet.

Tony
 
Tony,

Sorry, the frying pan method is taken from the SparkFun site where you actually put the board with all the SMD components on it (with solder paste applied through a stencil) and cook the board for 3-4 minutes on high. All the solder melts and the parts are soldered all at once. Nice for production. Not easy for one-offs because of the cost of the stencil.

Sounds like you are very experienced at this so no more advice on soldering. Hot air stations are great!

I hadn't noticed the 2SK209 FET and have been using the 2SK880 instead. They have exactly the same specs from Toshiba but your 209 is slightly larger and easier to handle. I should have noticed those before I started using the smaller ones. Thanks for using them(!).

Good luck,

Kit
 
I thought this shot may be interesting.
Just prior to finding the 680pF fix I was looking at the phase shifts across the output transistor

Larger lower sinusoidal waveform is output transistor base at 100mV / division.
Sawtooth form is output transistor emitter at 5mV / division.
Time base is 0.5uS

I have two images, one is rotated through 90 but they both show incorrectly in my browser. :confused::mad:

a65110c5-48ec-42bf-b13b-9b37ae48f4d8_zpsi7vf1lrc.jpg



20160107_124931_zpskrmgvbzg.jpg
 
RyanSoundLab said:
Tony,

Sorry, the frying pan method is taken from the SparkFun site where you actually put the board with all the SMD components on it (with solder paste applied through a stencil) and cook the board for 3-4 minutes on high. All the solder melts and the parts are soldered all at once. Nice for production. Not easy for one-offs because of the cost of the stencil.
Click to expand...

Thanks, learned something again. Might try it one day. With my hot air I apply the paste with a small syringe by far the hardest part to get right for me.
When I hot air the Sabre I usually end up blowing the adjacent bypass caps away and refit with the iron.

RyanSoundLab said:
Sounds like you are very experienced at this so no more advice on soldering. Hot air stations are great!
Click to expand...

Just broke a lot of DACs so had some practice out of necessity :D
Its only a circa £150 hot air station but gets by, also have a cheap rework vacuum gun but my SP200 is by far the best tool I have so tend to use it for most everything.

RyanSoundLab said:
I hadn't noticed the 2SK209 FET and have been using the 2SK880 instead. They have exactly the same specs from Toshiba but your 209 is slightly larger and easier to handle. I should have noticed those before I started using the smaller ones. Thanks for using them(!).
Click to expand...

Got lucky with google on that one.
 
Tony,

The scope pictures make some sense. The base drive is going to be very good looking since it's the output of the opamp (that is, "good looking" for a sine wave oscillation that shouldn't be there!). The emitter trace shows that the emitter follows the base up (charging the output cap) and then 10mv down until (most likely) the output 22uf capacitor holds the charge and keeps the emitter from going any lower. That trace then more slowly discharges (ramps down to the right) until the base drive picks it up again. It appears to discharge 10mv in 1 us, so with a 22uf capacitor, I calculate a load of around 220ma (I = VC/t)? Seems too high for your circuit but may be just capacitors charging and discharging at this high frequency. Glad you fixed this oscillation.

Kit
 
Well Kit,
You sure did a neat job of picking the bones out off that scope shot.
Interesting for me to learn of I = VC/I, Ill add it to me spread sheet as a further weapon!

Load was originally circa 85mA but by this point I had reduced it to 65mA so around 155mA (<200%) current transient.

Out of my depth here but maybe that transient whacked back down the regulator (had previous shots showing this) and hit the reference voltage on pin 3. A nice closed loop perpetual oscillator?

The 680pF just clamped the bandwidth below the frequency of oscillation and was a band-aid fix and not root cause?

Some perhaps undeclared information here,
This regulator feeds my Sabre 9018 plus an additional load of

A pair of 1K resistors forming a 1/2 voltage divider with 220uF across one resistor.

264.jpg


Actually two of the above which feed the Vref in each of the two single ended fet circuits below.

Untitled-3.jpg
[/IMG]
 
Tony,

I do remember your FET DAC circuits as the impetus for building your supplies. Very complex and elaborate circuit concept. You say the sound is great so that makes it worthwhile.

The 680pf adds a pole to the response which brings the gain down below unity before the phase shift around the whole regulator gets close to 180 degrees. It's not a Band-Aid per se but a well known technique for ensuring stability. Studying poles and zeros was never one of my strong points but I'm sure Martin can explain what's happening better than me.

Just one question - why did you feel the need to use bi-polar electrolytics for the Vref circuits? Or was it just convenient for some other reason?

Kit
 
Thanks Kit,
No particular reason for the bipolars other than I had them and they are also on the end of the fet I/V circuit.
Infact the cap across the V/Ref divider has no apparent impact on the spice model or SQ of the I/V stage. I seem to remove and then add it back on an annual basis.
The I/V was originally powered from battery banks and a choke filter with no regulation, I developed an itch to run regulated and the sixty or seventy odd batteries always gave me the willies!

You know the I/V stage is not my design of course, came from the Linear audio article and a guy called Patrick.

Regarding poles and zero's, if its not your strong point I'd probably best not try to learn too much advanced electronics in a diy room and settle for what is within my stride. I am rather good at making electronic circuits run hot though .D

Tony
 
Latest below,
Schematic unchanged other than more SMD with C2 and C3.
C2 now Nichicon UQ series and C3 7343 Tantalum.

Bottom ground plane (blue) now has only two incursions from the top layer and both are "in line if sight" rather than across.

Q6 now closer bypassed by C10 and room for a large heat sink.
Heat sink is floating off the board and 36mm wide 13mm thick 50mm high.
Not sure of the thermal characteristics but it is not even warm dissipating 1 watt.

X2 is a hair brained option to feed one Mite into the Vref circuit of a second Mite. Something I tried with no great successes on the augmented Fleas but I think John Lukins had success doing this with ALWSRs

LK1 & LK2 is another hair brained idea to provide an option to feed the output transistor from the LM7812.
Current link is LK1 at 2R2 which provide a convenient series shunt for output current measurement, adds a little series R between C4 & C10 and coupled with running LK2 provides a means of reducing dissipation in Q6.

Untitledschem_zpshq23lumt.png


Untitledtop_zps6jrzk0q8.png


Untitledbot_zpsssnph6ce.png
 
RyanSoundLab said:
The 680pf adds a pole to the response which brings the gain down below unity before the phase shift around the whole regulator gets close to 180 degrees. It's not a Band-Aid per se but a well known technique for ensuring stability. Studying poles and zeros was never one of my strong points but I'm sure Martin can explain what's happening better than me.
Click to expand...

Aha, I had one of my moments, When I first read this I thought you had a typo with
180 degrees
Click to expand...
thinking you ment 90°, shame on me, but the penny has just dropped!

At 90° phase shift we still have negative feedback and stability, with a further 90° IE 180° phase shift the feedback turns from negative to positive and the system oscillates.
 
C12 - 22nF now removed with no negative effect.

Scope shot is with, without and regulator powered down.
Same shot all three so time to implement PDs suggestions, BNC connector ordered.
5mV / 0.05uS

20160111_150016_zpszr4aiotk.jpg
 
Tony,

Now it looks like real noise - just like you wanted. I often get similar inherent noise when putting the scope probe on an unpowered grounded circuit. Doesn't make a lot of sense but it's also harmless. Sometimes I will put a simple R-C filter between the circuit and the scope probe to filter out RF-type noise so I can better see what's happening at audio frequencies. Some scopes have bandwidth limiting built into a switch but this also works.

Regarding the heat sink, I use this style and it works very well. It does have two pins on the bottom designed to be soldered to the board itself which gives it more rigidity. However, the position you have it in currently will block one of the mounting holes/screws for the board so I suggest either moving the heat sink away from the corner or just leaving it unsoldered. Note also that the heat sink will usually sit right on top of the board and could interfere with even SMD components there. I'm sure you've already spotted this.

Also, as a follow up to my comment about the "frying pan" soldering technique, here's a link that explains it quite well. Even with hand paste application, if you put all the dabs and components on first, then do the soldering it will look much better than hand-soldering each piece (or having it blow away with the heat gun- as has also happened to me).

https://www.sparkfun.com/tutorials/59

Best regards,

Kit
 
Scope is a Tektronjx 465, more knobs than I know how to twiddle :D
It has HF reject if I switch it on.

Yes the heat sink has to float off of the board and be placed after the top right board pin is fitted. It's not ideal but does work. I plan to pull the heat sink fixing lugs off.
I'll remove the heat sink profile before the gerbers are sent off.

Love your link to diy reflowing but will stick with me poker for now.
Does look fun though.

Tony
 
Excellent quality scope. I have at least 10 scopes (lost count), almost all Tek, and they're the best.

Looking forward to seeing your completed new board sometime.

Kit
 
You must log in or register to reply here.


advertisement


Back
Top