diyAB Amp - The "Honey Badger"

diyAudio Editor
Joined 2001
Paid Member
I have asked member ostripper to lay out a Class AB amp officially called the "diyAB amp" ( nicknamed "the Honey Badger" ) using many of the best characteristics of amps of this type that have appeared on our forums, so that we can feature it in the diyAudio Store. He has agreed to give it a go, is really interested in comments to help it be a paradigm of its type, and will be posting here very soon to describe where the design is at this point.

So pile on!

Mark
"Variac"

Build thread for diyAB Amp: http://www.diyaudio.com/forums/solid-state/211905-diyab-amp-build-thread.html
 
Last edited:
PCB

After corresponding with Variac on the present layout of the DIY case they will offer and the diyAudio heatsinks drilling pattern, I have ported the well known TMC LIN amp to a 75mm X 250mm PCB (below 1).

Some doubt the level of fidelity a circuit such as this can reproduce , but I aim to show them that simplicity can be good.

The design goals ( correct me if I'm wrong) are :

- 150W/8R - 250+W/4R , 3 pair easily obtained BJT output devices run in a Self EF2 configuration. (to-3p or to- 247)

- "All in one" construction - A pair of Apex Jr. 8200uf's , or just about any 10mm or 22mm 4 pin snap-in capacitor can be mounted on the main board. I looked up the panasonic line of 35mm caps and one could even grace this amp with 22,000uf X 2 @ 63V. :cool:

- To avoid some of the layout "issues" in common with some DIYA amps. Grounding will be fully starred with voltage stage/input stage "lifted" and brought to the star at center point. I could also implement separate or "boosted" supply points which could be used if desired. Output feedback take off point and speaker tap are also "dead center". All semi's will be thermally coupled for almost "military" reliability.

- Compensation and other aspects of the basic design will be user defined (miller - tmc - 2 pole) , I have tried all 3 with success. The basic design is quite documented already in both D. Self and B. Cordell's books. My oldest DIY amp is similar and has been used for 2 1/2 years.

-Cost , this would be up to DIYA - There are no "exotic parts" , all are readily sourced. Even as this is so , the amp generally will have -110db noise , 1-2mv offset , and near PPM distortions with 80V/us slew.


Any comments or suggestions , the base circuit is (below 2). Any valid idea's on either the circuit or layout welcome. I am at the voltage stage / input stage layout now - I think the output stage will be free from any errata - I've used this "style" before :) .
PS - the board will be double sided with the option of a single sided full DIY etch (no jumpers).
OS

please note that the latest BOM and schematic can be found at the diyAudio store here diyab-honey-badger-class-ab-power-amp-pair.html
 

Attachments

  • DIYA_AX.GIF
    DIYA_AX.GIF
    69.3 KB · Views: 44,487
  • DIYA_AB.gif
    DIYA_AB.gif
    29.3 KB · Views: 43,836
Last edited by a moderator:
I would have thought that over current protection would be mandatory. The only issue I've got is that the transistors in the schematic (mainly small signal) aren't readily available to those in the UK. I have used them all myself though, but that required an order to digikey/mouser. Although if you can show me where to buy 1845/992s in small quantities without mad high shipping costs, that would be great.
 
Hi OS, the only things I would suggest adding are onboard DC and overcurrent protection, make things nice and simple and bombproof for newbies. You could always allow the option of linking out if not required.

That would be nice. I am constrained by the case/heatsink configuration. A protection circuit needs it's own supply / relay / driver. After reading about member Bonsai's experience , a standard relay driven by a standard ta7317 WILL NOT always work with a high current amp. PMA's HEXFET solid state solution will handle 100's of amperes and disconnect in a couple uS's. Perhaps we can design a "deluxe" unit . A breakaway board with softstart/timer and ta7317/SS relay. DIYA could "corner the market". I agree this is a shortfall in many DIY amps.. but some audiophiles don't like a relay at the output regardless (OPT-IN is always better). I could integrate but only at the expense of the huge supply caps.


By tekko - Whats with the batteries in the schematic ?

The batteries are to manually DC bias the input stage , LT's capacitor models "cloud" the FFT data with artifacts. Keentoken enlightened me to this condition.

By bigun -Please make it easy to measure the bias voltage across the output device emitter resistors without blowing up the amp with 'slip' of the multimeter probes - you can guess why I want to see this feature !

A very easy to add convenience feature .... those 22mm ceramic .22R resistors do measure easily , but a set of extra pads with standoff pin's (tp1 - tp2) would be nice. :cool:

By 5th element - I would have thought that over current protection would be mandatory. The only issue I've got is that the transistors in the schematic (mainly small signal) aren't readily available to those in the UK. I have used them all myself though, but that required an order to digikey/mouser. Although if you can show me where to buy 1845/992s in small quantities without mad high shipping costs, that would be great.

Take a look at the Krell 300i below (no protection). I have mixed feelings about this , some from actually doing the "deed" ! This amp (the lin) typically just opens the fuses when shorted at even high- moderate volumes , I have only taken out the output stage with a slipped screwdriver or probe ( me too, - Bigun).

To address the sourcing , I'm sure we could find a way to include the fairchild semi's. Worst case , A970/C2240's can be used for the input stage , any 200+Hfe / 120Vce/ BCE device would give similar performance. MJE340/350 can be used for the VAS , but with a slight performance degradation. ANY to-3P or to-247 can be used for the output stage ( 60-120 Hfe/ 200v+ Vce). 2sc4793/2sa1837 can be used for drivers , with dedicated 8R use.
You could literally fill the BOM from your junkbox , and still have a pretty tight spec'ed amp.

OS
 

Attachments

  • krell300i.jpg
    krell300i.jpg
    54.7 KB · Views: 41,490
You may already be planning this - but if not, I'd suggest adding AKSA-style resistors in parallel with the rail fuses. They have proved helpful to me in the past.

On the schematic posted above you seem to have used the same ground symbol throughout (unless I've read it wrongly), I think you want to have the protective earth / ground lift go to the chasis instead - you probably intended this but I would normally have moved the ground symbol to signal input ground even in a simulation so thought I'd just mention it.

I defer to your better experience, but I usually plan to have the output choke off-board, wire wrapped resistor in line with speaker wiring - keeps the magnetic field lines away from sensitive parts of the circuit. I don't know if this is important or not, just something I do.

p.s. I like the long aspect ratio of this board.
 
Last edited:
Hi OS, the only things I would suggest adding are onboard DC and overcurrent protection, make things nice and simple and bombproof for newbies. You could always allow the option of linking out if not required.

Yes... about VI limiter? Just leave protection components out if desired.

What about fuses onboard with parallel resistors a la Aksa 55 and Baby-Aksa, to measure bias current?

:)
 
You may already be planning this - but if not, I'd suggest adding AKSA-style resistors in parallel with the rail fuses. They have proved helpful to me in the past.

On the schematic posted above you seem to have used the same ground symbol throughout (unless I've read it wrongly), I think you want to have the protective earth / ground lift go to the chasis instead - you probably intended this but I would normally have moved the ground symbol to signal input ground even in a simulation so thought I'd just mention it.

I defer to your better experience, but I usually plan to have the output choke off-board, wire wrapped resistor in line with speaker wiring - keeps the magnetic field lines away from sensitive parts of the circuit. I don't know if this is important or not, just something I do.

p.s. I like the long aspect ratio of this board.

OK - you are a bonanza of idea's, Bigun :)

If you look below , you will see the "TP1" and "TP2" , if you can't probe those - stay away from DIY :D

The fuse resistors are there , too. I was thinking , is it not the purpose of these resistors to keep the VAS/input stage "alive" in case of fuse failure ? If this is the case , would it not be better to have the "safety resistors" run directly from the unfused faston to the small signal section of the amp ?? On my modular creations , I use 22R fusible resistors on the boosted VAS/IPS board. These double as both the R/C filter and as a safety factor.


The grounds are denoted as "G1" and "G2". That is for my modular and "super" amps that use separate power supply PCB's. On this integrated supply board , G2 would be a separate trace that meets the center of the main star. From that trace , the 5-10R "lift" resistor + 2 diodes and the other input stage earth references are derived. I directly reference the DC cap and VAS ground to the earth BEFORE the lift resistor. If you reference them AFTER the resistor , you have modulation of your reference. Comments on this are welcome.

I really have thought about protection. On the typical OEM with just adequate capacitance and current limited trafo's , the classic 7317 IC/10A relay will protect (in most cases). Here we will have up to 20Kuf X 2 and most likely a hefty 400VA X 2 or 800VA+ toriod. Just ask Bonsai about his blown speakers and melted relay :( .

To monitor TP1 and 2 with a SOTA DC/current detection IC and output this to a MUCH faster solid state relay would be better. If TP1 or 2 reached a preset voltage for X milliseconds (Rod elliots PIC detector or similar)Project 111 - PIC Based Speaker Protection
, it could fire the relay within a couple uS.

On board VI limiters will always affect sonics , as well. MUCH more defined and scalable VI limiting can be done with an external circuit as I described.

Variac discussed 2 PCB's originally , The amp itself just needs one , PS and all. A comprehensive VI/DC/soft start would be a good solution to all the above factors. It could also be ported to any amp.

Sound buster , the safety resistors can be used to measure OP current + the
current demands of the input circuitry. To measure the output Re's directly is more accurate and you can see if any OP devices are failing to "share" (gross Hfe mismatches). Most ON semi output devices from the same order will typically read within a mV across one of the .22R resistors. For example , these amps are typically set at 10-12mV.

OS
 

Attachments

  • Diya_ax2.gif
    Diya_ax2.gif
    35 KB · Views: 39,315
The safety resistors can be used to measure OP current + the current demands of the input circuitry. To measure the output Re's directly is more accurate and you can see if any OP devices are failing to "share" (gross Hfe mismatches). Most ON semi output devices from the same order will typically read within a mV across one of the .22R resistors. For example , these amps are typically set at 10-12mV.

So, let's TP on every Re... :wrench:
 
Just a few ideas/options.

So, let's TP on every Re... :wrench:

Most builders will run through their pairs the first time . After that, one would just measure 22mv across the TP's , run the amp , maybe readjust. The emitter resistor leads are 20mm away from any rail.

A few thoughts to ponder .... I have built this particular amp with cascoded input stages , one pair with (cascoded)SS9014 low noise BJT's and one pair with
cascoded 2n4416 Jfets. The BJT's were quieter (only measurable with a CRO) , the Jfet's subjectively sounded better (I am aware of the psychology).In Either version, the differences were minimal.

Benefit's of the cascode are the wider choice of input devices , less thermal drift as the input pair Vce is lower =cooler. Of course you need a minimum of 2 resistors or resistor/zener to bias the cascode , increasing the BOM. The Jfet's will have an effect on the loop gain of the amp , reducing the effectiveness of TMC. With this amp you can choose standard miller compensation , so a FET input pair amp is still possible. The "tradeoff" is more distortion vs. A "slightly different" sound. I have no AP , but in simulation , a slight shift in harmonic content can be seen with the Jfets.

I would appreciate opinions as to whether I should include the cascode.

The layout features would be :
1. Jfet or BJT - I know both work and have the values already worked out.

2. Total "opt out" - jumper the cascode , use the 2 high Vce input pair (992/1845).

3. Cascode voltage reference can be just a simple resistive divider or zener referenced (or omitted).

I would like feedback as to whether this would be a good addition or would it overly complicate the design. It would be good if this design could satisfy both the "new" builder and someone with higher expectations (picky people :D).

It was mentioned to me that the power supply capacity might be a bit low. Is up to 22,000uf X 2 (panasonic) ECE-T1JP223FA Panasonic Electronic Components Aluminum Electrolytic Capacitors - Snap In Insufficient for a 150W amp ?

I think this would be quite sufficient The class A design usually has much poorer PSRR and lower or no negative feedback , thus needing 100Kuf + CLC/CRC to have acceptable performance. An AB amp (this one) is 70+ db PSRR "right out of the box" . Our popular goldmund clone only uses 2 Apex jr. 8200uf's , the AKSA's use 4 X 10K onboard (good deal). All these far exceed any OEM I've repaired - most will get away with 10K X 2 for a STEREO 120W setup ..... pure economics.

Below 1 / JFET - TMC is really not effective (68pF could be used w/ standard miller comp.). Distortion is still 10-12 PPM @ 10K , H2 is more dominant. Whether this can be heard might depend on the "golden ear factor" :D

OS
 

Attachments

  • jfet.gif
    jfet.gif
    24.9 KB · Views: 10,277
diyAudio Editor
Joined 2001
Paid Member
I like the idea of some options, if it were clearly spelled out how to implement them or not. I found notes on the board itself to be the least threatening. Like a note saying something like:

For no cascode eliminate all asterisked components & jumper "A" to "B"

Of course room is needed to write this, but it came be done more succinctly than my example..

Hoever I'm no expert on this type of amp, so you guys should decide this..

If OS could write up a short article after the board is designed in which all the options are explained clearly, it would always be available to builders.. Or diyAudio could publish it..
 
I think the cascode is a good idea, it gives people more room to play around if they want, especially if the BJT and Jfets share the same pin outs too. This also gives greater flexibility on the input pair choices, easing the selection of a high beta pair as now the Vce restraints have been lifted.

I'd say that any economical advantage when removing the cascode + biasing would be largely irrelevant as the additional cost would be peanuts when compared to the rest of the expensive items. As Variac says, having the option to 'jumper this', if you want X to happen etc is a nice idea, it does add complexity to the overall PCB design and increases the BOM again, but if anything DIYers prefer to have options.

I think having an amplifier where you can try out the various compensation methods + different styles of input stage would be quite interesting and would probably pique quite a few peoples interests.

I also think that having three output pairs is a good choice too as it allows the amplifier to deliver a decent amount of current when set at the 150 watt @ 8 ohm target.

Another thing perhaps is that the cascode would give more flexibility on the power rails. With three output pairs, if you had a benign load to drive, you could easily up the rail voltages to give more into 8 ohms. Of course if you've got a high bias input stage this could push the input transistors beyond their comfortable limits, some small signal stuff will only tolerate 300mW and that's at 25 degrees C. 50 volt rails and a tail current of say 2mA would push them to their limits. If someone wanted to up the rails to 60 volts they'd be in danger territory.
 
Most builders will run through their pairs the first time . After that, one would just measure 22mv across the TP's , run the amp , maybe readjust. The emitter resistor leads are 20mm away from any rail.

A few thoughts to ponder .... I have built this particular amp with cascoded input stages , one pair with (cascoded)SS9014 low noise BJT's and one pair with
cascoded 2n4416 Jfets. The BJT's were quieter (only measurable with a CRO) , the Jfet's subjectively sounded better (I am aware of the psychology).In Either version, the differences were minimal.

Benefit's of the cascode are the wider choice of input devices , less thermal drift as the input pair Vce is lower =cooler. Of course you need a minimum of 2 resistors or resistor/zener to bias the cascode , increasing the BOM. The Jfet's will have an effect on the loop gain of the amp , reducing the effectiveness of TMC. With this amp you can choose standard miller compensation , so a FET input pair amp is still possible. The "tradeoff" is more distortion vs. A "slightly different" sound. I have no AP , but in simulation , a slight shift in harmonic content can be seen with the Jfets.

I would appreciate opinions as to whether I should include the cascode.

The layout features would be :
1. Jfet or BJT - I know both work and have the values already worked out.

2. Total "opt out" - jumper the cascode , use the 2 high Vce input pair (992/1845).

3. Cascode voltage reference can be just a simple resistive divider or zener referenced (or omitted).

I would like feedback as to whether this would be a good addition or would it overly complicate the design. It would be good if this design could satisfy both the "new" builder and someone with higher expectations (picky people :D).

It was mentioned to me that the power supply capacity might be a bit low. Is up to 22,000uf X 2 (panasonic) ECE-T1JP223FA Panasonic Electronic Components Aluminum Electrolytic Capacitors - Snap In Insufficient for a 150W amp ?

I think this would be quite sufficient The class A design usually has much poorer PSRR and lower or no negative feedback , thus needing 100Kuf + CLC/CRC to have acceptable performance. An AB amp (this one) is 70+ db PSRR "right out of the box" . Our popular goldmund clone only uses 2 Apex jr. 8200uf's , the AKSA's use 4 X 10K onboard (good deal). All these far exceed any OEM I've repaired - most will get away with 10K X 2 for a STEREO 120W setup ..... pure economics.

Below 1 / JFET - TMC is really not effective (68pF could be used w/ standard miller comp.). Distortion is still 10-12 PPM @ 10K , H2 is more dominant. Whether this can be heard might depend on the "golden ear factor" :D

OS
I suggest this cascoded LTP. Cascode is floating with LTP emitters.
dado
 

Attachments

  • DADO-TT-TMC-3-cascVAS-015-tripl-ltp.jpg
    DADO-TT-TMC-3-cascVAS-015-tripl-ltp.jpg
    49.8 KB · Views: 8,180
5'th .... You know , I am going to allow for boosted/separate rail capability. Standard on all my amps. In this case, even know it will be added/documented/labled , you would just leave the last 2 resistors out of the last R/C filter for the front end and hook your boosted rails there.

In general , you can use higher global rails , too. My Nikko version of this amp uses nearly 70V with a current limited trafo and 2 pair outputs. It's headroom at 8R is impressive (quite scary , even).

Variac , even as the physical is quite important , the documentation is equally important. The goal is to not scare the new constructor away or shock the class A crowd (these amps for bi-amp bass , maybe ?) , but allow for refinement if wanted - broad appeal. My house amps are already like this , 1 pair is cascoded Jfet with no TMC (just one miller cap/1 jumper/omitted resistor) , the other (2 miller caps/resistor/no jfets).

I'm not throwing unproven design's out to the crowd , BTW. I specifically chose this one for the quality/simplicity/support aspects. Also , I have abused these to the max - they keep on truckin'. :cool:

OS
 
I suggest this cascoded LTP. Cascode is floating with LTP emitters.
dado

Yes , that actually is the way I have my jfet one , but with just 2 resistor's and a small bypass cap.

If you do a poll of the OEM cascoded amps , you will see them divided into 3 camps :

1. floating emitter referenced with bypass between cascode and LTP emitters.

2. Rail to ground with bypass between cascode/ground (some use a 12v zener).

3. rail to rail floating - bypass to ground (rare).

Comments on best choice welcome (I've used 1/2 sucessfully).

PS - KISS = keep it simple !!! (just the 2 semi's + 2-4 passives)
OS
 
Last edited:
I like the cascode idea.
It adds a tiny bit of board area.
It can be omitted using zero cost jumpers.
The jFETs in the LTP can be replaced with BJTs if desired.

I love the test points idea. I'd make them mandatory.

Fused rails with 100r bypasses permanently in place.
Once the circuit is proved the empty fuseholders get 10r or 1r resistors soldered inside old fuses. This allows quiescent current measurement with very little voltage loss. Output bias current and driver bias current are measured across the test points fitted to the emitter resistors.
Finally the intended fuses replace the 1r0 resistors.

I hate the EF2. Why not EF3. Again easy to link out the pre-driver stage, if desired.
All capacitors must have at least dual pin pitch options.
All drivers and output to have provision for base stoppers.
Power input via a twisted triplet to three, very closely spaced, power terminals.
All or most signal transistors to have 4 pin pads to allow the different pin outs. bc, 2n, 2sc etc.
 
I hate the EF2. Why not EF3.

You "hit the nail on the head". I hate the standard EF3 , harder to thermally and electrically stabilize with modern High Ft semi's.The Cordell DBT triple is quite superior , but too complicated. The EF2 can be good if the VAS Ic is high and the driver/output beta's are high. 3 OP pairs also work very well with an EF2.

The "tradeoffs" here are:
-Larger onboard capacitance vs. design sophistication.
-Project had to be ported to an existing Class A chassis with preset parameters.
-A real world assessment of what projects members are actually building , and what skill level these projects are at.
-Just how "bulletproof" to make the project. I really feel the protection should be offboard , PIC or ta7317 based - with PMA's HEXFET ultrafast SS relay. Onboard VI sucks , except for a "road amp" , relays don't always work - unless you get the 100A/ 20$ platinum edition. :D


The Leach amp was good for first time constructors as it used slow devices .My present triple would not tolerate the stray inductances (wires) that you see on the typical leach build. I do know how to mitigate this , but with extra local bypass circuitry , L/R coupling. In my opinion , the leach is too hard for the first time constructor. For every leach build , 10 buy an ebay kit and 10 build a DX.

After I perfect the luxman/DBT (finish any R/D , give it the torture test) , it might be "round 2" as a boardstore kit.

OS
 
Last edited:
Yes , that actually is the way I have my jfet one , but with just 2 resistor's and a small bypass cap.

If you do a poll of the OEM cascoded amps , you will see them divided into 3 camps :

1. floating emitter referenced with bypass between cascode and LTP emitters.

2. Rail to ground with bypass between cascode/ground (some use a 12v zener).

3. rail to rail floating - bypass to ground (rare).

Comments on best choice welcome (I've used 1/2 sucessfully).

PS - KISS = keep it simple !!! (just the 2 semi's + 2-4 passives)
OS

The cascode I suggested is actually proposed to me by Cordell in his book thread.

"I normally recommend caution in regard to the type of circuit you are using to bootstrap the input cascodes; I'm not crazy about injecting extra current into the LTP tail circuit. I might have used an emitter follower into whose emitter I would have flowed that current. I prefer the driven cascode that is driven by a replica of the feedback signal, which, under normal conditions, is pretty much the same as the common mode signal."

dado
 
  • Like
Reactions: 1 user