Hafler DH-200/220 Mods

fab

Member
Joined 2004
Paid Member
As suggested in "hafler dh-220" thread to repair it, here is the starting point of mods that peoples have done on this amplifier.

So far in previous abovementioned thread, 2 mods were brought up:
#1) Replacing the load resistors in input stage with current mirror. I believe that this mod would increase the open-loop gain of the amp, thus would increase the amount of feedback. Is it good or bad? or can we re-adjust the gain of the input stage or play with the gain of the driver stage so the overall open loop gain is not modified?
#2) Regulating the power supply of the circuit (except the mosfet output stage!). This one I have personnaly done. See attached principle. The regulated voltage can be higher than indicated on the schematic so the maximum power loss is not much. Has someone else tried this mod #2 also? Even my girlfriend (who are not an audiophile) noticed the improvement on the sound (she was not trying to be kind with me...).

Note: the modifications I have suggested here on my part in this whole thread should be done and integrated only on a new pcb layout for final use. My goal here was to test the mods as prototypes only for evaluation purpose. I do not recommend someone else to do it the same way as shown.
Fab

My stuff USSA, FSSA: https://fabaudio.online
 

Attachments

  • dh-200_mod_power_supply.gif
    dh-200_mod_power_supply.gif
    22.7 KB · Views: 21,033
Last edited:
"Replacing the load resistors in input stage with current mirror. "

Who ever suggested that has never seen the schematic, won't work. The VAS devices are biased by the DC developed across the load resistors, its a complementary design.

"Regulating the power supply of the circuit (except the mosfet output stage!). "

Great idea. The ideal would be ±7V~10V above the nominal ±60V rails. A non-feedback regulator sounds the best, a pass transistor with a zener in the base.

Replace the main filter caps with 15,000µF~20,000µF and add a pair of 47µF bypass caps. Replace power switch and add 2 ohm inrush current limiter(thermistor). Replace the 16ga ground buss with a 3/8" copper strap.

Add a pair of Schottky diodes, one in series with the bridge + and - outputs going to the main filters.

Replace the 0.1µF cap in the output zobel with a 250V foil and film type.

Once the regulated supply for the front end is in place, remove the CCS for the diff inputs and replace with a pair of resistors.

Cascode the diff pairs and return the resistor tails to the opposite reference voltages(see the Leach V4.5).
 

fab

Member
Joined 2004
Paid Member
Yes, it is long overdue (you can check on Ebay to see the amount of dh-200/220 amps that are sold!)

#1)
""Replacing the load resistors in input stage with current mirror. "
Who ever suggested that has never seen the schematic, won't work. The VAS devices are biased by the DC developed across the load resistors, its a complementary design. ""

fab:I am not the one who suggested it but nothing is impossible. In Elect. & Wireless World issue of January 1985, it was described a symetric complementary discrete op-amp with current mirrors for negative and positive loads of input pairs. Question: Using the referenced Leach v 4.5, could Q8 with R21 imposes a fixed DC voltage at base of Q12 so a current mirror could be used (a course the same would apply for the negative complementary part of circuit)?

#2)
""Regulating the power supply of the circuit (except the mosfet output stage!). "
Great idea. The ideal would be ±7V~10V above the nominal ±60V rails. A non-feedback regulator sounds the best, a pass transistor with a zener in the base."

fab: Where do you get the additional 7 to 10 v (use of another transformer?, voltage doubler?) ?

#3)
"Replace the main filter caps with 15,000µF~20,000µF and add a pair of 47µF bypass caps."

fab: I used 2 x 30000uF in one amp and 4 x 14000uF in another. For the latter, I used 2 x 14000uF with one bridge for the positive supply and used 2 x 14000uF with one bridge for the negative supply. The 47uf along with 100nf bypass caps were installed directly between the mosfet drains and chassis dissipator.

#4)
"Replace power switch and add 2 ohm inrush current limiter(thermistor). Replace the 16ga ground buss with a 3/8" copper strap."

fab: of course for the power switch and the round buss. For the current limiter, have you experienced problem in the past without it or is it precaution measures?

#5)
"Add a pair of Schottky diodes, one in series with the bridge + and - outputs going to the main filters. "

fab: what is the advantage?

#6)
"Replace the 0.1µF cap in the output zobel with a 250V foil and film type."

fab: is it making a big difference?

#7)
"Once the regulated supply for the front end is in place, remove the CCS for the diff inputs and replace with a pair of resistors."

fab: I prefer keep the CCS to reject more common mode signal.

#8)
"Cascode the diff pairs and return the resistor tails to the opposite reference voltages(see the Leach V4.5)."

fab: cascode of course! for the resistors, see above comment on #7.

I have some more on my side....but let start with comments on the above mentioned so far. Doing a change is someting but appreciating it is something even more important. After all, the goal is to reduce the degradation of the power amp and make it more reliable.
 
"fab:I am not the one who suggested it but nothing is impossible. In Elect. & Wireless World issue of January 1985, it was described a symetric complementary discrete op-amp with current mirrors for negative and positive loads of input pairs. Question: Using the referenced Leach v 4.5, could Q8 with R21 imposes a fixed DC voltage at base of Q12 so a current mirror could be used (a course the same would apply for the negative complementary part of circuit)?"

Q8 is part of the short circuit protection and being engaged all the time would present such a heavy load as to negate the advantages of CCS loading. I don't like the sound either.

"Where do you get the additional 7 to 10 v (use of another transformer?, voltage doubler?) "

Easiest way to get it is to use a dual 5V transformer and add one winding to each end of the main transformer.

"I used 2 x 30000uF in one amp and 4 x 14000uF in another. For the latter, I used 2 x 14000uF with one bridge for the positive supply and used 2 x 14000uF with one bridge for the negative supply. The 47uf along with 100nf bypass caps were installed directly between the mosfet drains and chassis dissipator."

30,000µF is right at the edge of bridge failure. I also find with a regulated front end that it reduces the need for massive filters for the output stage. For 60hz the point of diminishing returns for a single 8 ohm load seems to be 3,300µF or 6,800µF for a stereo amplifier, and double that for 4 ohms. For 50hz I go with 4,000µF. The bypss caps should be attached to where the speaker pulls power from. For the ±V that would be at the outputs, but the ground is at the transformer center tap. I find it works best to put the 47µF right on the main filters and the 0.1µF~1µF at the outputs. The smaller value can break into oscillation due to stray inductance from the wiring. Sometimes a small resistor in series helps, sometimes I have to remove the smaller value.

"For the current limiter, have you experienced problem in the past without it or is it precaution measures?"

Power switch failures, bridge failures (above 30,000µF), thermostat failures (mainly DH500 and Adcom GFA555).

"Schottky diodes, one in series with the bridge + and - outputs going to the main filters. "

fab: what is the advantage?"

Reduce rectifier noise. I find it much easier than trying to replace the whole bridge and make a heatsink for same. On any Crown bridge design (Power Base, Micro Tech, Macro Tech, etc) you only need one diode.

"Replace the 0.1µF cap in the output zobel with a 250V foil and film type."

fab: is it making a big difference?"

Think about it. The charge absorbed by the dielectric has to discharge when the waveform crosses zero in quiet passages. With no signal to mask it this discharge sounds like noise and hash. The currents through here can be huge. A 250V film cap may only be rated at 10V above 10Khz. A 250V foil and film type with soldered leads is best. Most film types have paste and mash lead termination, not so hot. The resistor in series must not burn up.

"fab: cascode of course! for the resistors, see above comment on #7."

Leach agrees with me, CCS sound bad (even though they measure better).

Actually, the best thing to do with a Hafler is lose the FETs and convert it to a Leach.
 
djk said:
"Replacing the load resistors in input stage with current mirror. "

Who ever suggested that has never seen the schematic, won't work. The VAS devices are biased by the DC developed across the load resistors, its a complementary design.

......

Actually, yes, current mirrors can safely replace the load resistors in this amp. The technique is commonplace in IC design, I certainly did not invent it. The mirror is a differential Darlington structure, which acts as both an excellent mirror and a buffer to the next stage and has well defined output levels to bias the VAS.

As for the other changes suggested, feel free. I don't see a return on investment for most of them, however. The sonic improvement for the current mirror loads is high and it is not difficult or expensive, so I place it at the top of my list. Regulating the front end supply is expensive, and the payback is possibly not so dramatic, but again, feel free...

If you want a Leach amp, build one, don't modify the Hafler to be one. Certainly any design that has lasted as long as the Leach amp must be pretty good. I'm impressed that it is on revision 4.4, and that it has been critiqued by everyone and continuously improved since 1977. The Hafler 200 was designed in 1979, and also continuously improved in later products in the Hafler product line, so they must be doing something right there too.
 
fab said:
.....#1) Replacing the load resistors in input stage with current mirror. I believe that this mod would increase the open-loop gain of the amp, thus would increase the amount of feedback. Is it good or bad? or can we re-adjust the gain of the input stage or play with the gain of the driver stage so the overall open loop gain is not modified?

It will increase the open loop gain, about a doubling. So either the compensation should be re-analyzed at the same time, or the input pair degenerated further by increasing the emitter resistors to return to the same gain and stability. I have not done this mod to this amp, so I will not recommend either choice. I have done the current mirror mod to the Citation MF12, with very excellent results.

I want to do a new driver board for this amp that incorporates the most effective improvements and uses modern available plastic packaged lateral MOSFETs (and possibly vertical MOSFETs, with appropriate biasing). It has become expensive to replace the old TO-3 packaged Hitachi's, and I think all the hand wiring to the outputs can be eliminated too. So I will be breadboarding this on an existing amp soon, hopefully prior to going to new PC boards. Stay tuned.
 

fab

Member
Joined 2004
Paid Member
thanks for the reply. Your comments are greatly appreciated.

Maybe this thread is going a little bit too technical, isn't it? It would be fun to get also other peoples (Variac, Larryg where are you?) involved as well.

For the speaker ground, I have connected it directly at main filter caps instead of passing through the pc board as suggested by Musical-concept (see attached file because it looks like their site has been hacked!)

djk said:"Actually, the best thing to do with a Hafler is lose the FETs and convert it to a Leach."

It looks like you have a big preference for the Leach amp! I respect that but the goal of this thread is to improve the actual DH amp with easy low cost mods. I like too much the "mosfet" sound (maybe because I learned to work with them from the beginning). Mr Pass, Borbely, Halcro seem to agree with me on that! I may repeat myself but all the mods I have done made quite a difference over the stock DH-200 amp sound! I believe that with a stock DH-200/220 and a few bucks for some mods can make this amplifier sounds like one at several thousand bucks...My speakers are not hi-fi but only "mid"- fi (Boston Acoustics VR-30).

Getting back to one specific mod (regulated front end power supply), I used a 78xx series regulator with Zener diode to increase the voltage output. The ripple obtained was extremely low, which preserves the stability of operating points of the pc board transistors. I have built this regulator circuit at very low cost on a Radio-Shack IC pc-board 276-150 (good quality proto board). With all the components, for 2 channels, the cost is less than 15$(can or us) and it fits easily inside the DH-200/220 amp. If someone is interested, I can provide the circuit schematics and layout.
 

Attachments

  • dh-200_ground_mod.gif
    dh-200_ground_mod.gif
    56.8 KB · Views: 18,896

fab

Member
Joined 2004
Paid Member
Re: Re: Hafler DH-200/220 Mods

slowhands said:


Actually, yes, current mirrors can safely replace the load resistors in this amp. The technique is commonplace in IC design, I certainly did not invent it. The mirror is a differential Darlington structure, which acts as both an excellent mirror and a buffer to the next stage and has well defined output levels to bias the VAS.

I want to do a new driver board for this amp that incorporates the most effective improvements and uses modern available plastic packaged lateral MOSFETs (and possibly vertical MOSFETs, with appropriate biasing). It has become expensive to replace the old TO-3 packaged Hitachi's, and I think all the hand wiring to the outputs can be eliminated too. So I will be breadboarding this on an existing amp soon, hopefully prior to going to new PC boards. Stay tuned.

Can you draw your current mirror for our knowledge (remember that the VAS does not have a current source)?

I am workinfg also on a new driver board for the DH-200/220 (I have one ready layout with traxmaker but I am always looking for possible improvements). Maybe we can share some hints and may end with a SUPER driver board? Imagine the power of design of many music addicted peoples all together!!
 
Originally posted by fab

#1) Replacing the load resistors in input stage with current mirror. I believe that this mod would increase the open-loop gain of the amp, thus would increase the amount of feedback. Is it good or bad? or can we re-adjust the gain of the input stage or play with the gain of the driver stage so the overall open loop gain is not modified?

It is good, provided it can be implemented.

In a discrete symmetrical topology, where the VAS consists of Darlingtons (like the DH-220), it is extraordinarily difficult to use active loads due to the lack of precision on the input stage.

There are quite a few areas where the DH-200 can be improved upon. The driver stage, for one, is worth examining. Since Power Mosfets can be driven directly from the VAS, the use of drivers in the DH-220 implies that the VAS could be biased slightly on the lean side. With these drivers removed and the VAS re-biased to about 15mA, I would imagine the character of the DH-200 to change significantly.
 
fab said:
#4)
"Replace power switch and add 2 ohm inrush current limiter(thermistor). Replace the 16ga ground buss with a 3/8" copper strap."

fab: of course for the power switch and the round buss. For the current limiter, have you experienced problem in the past without it or is it precaution measures?


After I beefed up the wiring and made other mods the in-rush current draws a huge spark across the power switch.
 

fab

Member
Joined 2004
Paid Member
dH-200/220 driver stage

mcp said:


There are quite a few areas where the DH-200 can be improved upon. The driver stage, for one, is worth examining. Since Power Mosfets can be driven directly from the VAS, the use of drivers in the DH-220 implies that the VAS could be biased slightly on the lean side. With these drivers removed and the VAS re-biased to about 15mA, I would imagine the character of the DH-200 to change significantly.

I have already done it in the past (removed the output stage emitter follower drivers and re-biased the VAS from 10 to 15 ma (changed the 100 ohms to 68 ohms in DH-200) for restore original slew rate. This had the advantage to have a more DC stable value at output mosfet gates (because of voltage multiplier compared to the voltage across the 220 ohms resistor which is varying with Vbe of drivers transistors). In fact this was done by the original designer of the DH-200 amp (Erno Borbely) in 1981 and also in Hafler 9505 for example. I confirm that it changed a lot the sound of the amplifier and was probably better at low and medium listening levels but it seemed worst to me at high levels than with the drivers. After reading some recent audio patents, and also noticing that Mr Borbely himself came back with the driver stage in his latest designs, I have been convinced that the driver stage was necessary to de-sensibilize the load seen by the VAS against the AC load at high volume level. So , I re-installed the driver stage and even installed a darlington as the driver stage for the latter reason (as an experiment)!!
 

fab

Member
Joined 2004
Paid Member
djk said:
"fab:I am not the one who suggested it but nothing is impossible. In Elect. & Wireless World issue of January 1985, it was described a symetric complementary discrete op-amp with current mirrors for negative and positive loads of input pairs.

I found the article. See by yourself the scanned circuit. I believe it is John Linsley Hood design.
 

Attachments

  • symetrical_jfet_input_opamp.jpg
    symetrical_jfet_input_opamp.jpg
    28.4 KB · Views: 9,086
There's also an article in Randy Slone's "High Power Audio Amplifier Construction Manual" that is related.

A symmetrical amp with current mirrors at the diff inputs coupled to a darlington VAS is featured in Fig 7.6a.

He then followed up with cascodes in the VAS and lastly the drivers and the lateral mosfets in the current stage are in CF fashion.
 
Re: Re: pc19-c board

fab said:



There is no "stock" boards (DH-200/220/280/500) from Hafler that incorporates the mods discussed in this thread (to my knowledge).

The Hafler XL-280 and XL-600 incorporate the higher front-end voltage in their PC-40 driver board (circa 1987). This board uses complementary monolithic JFET duals in the "J. Curl" configuration on input, and the Lender VAS, but no current mirrors.

The Self book really awakened audio amp designers to the crucial importance of input stage balance for minimal distortion, and mirrors are one tool to attain it. So far this thread has pointed out two satisfactory means to implement them in complementary symmetry amps; neither is the method I alluded to earlier. Apparently it is not merely possible, but also old news.
 

fab

Member
Joined 2004
Paid Member
mcp said:
There's also an article in Randy Slone's "High Power Audio Amplifier Construction Manual" that is related.

A symmetrical amp with current mirrors at the diff inputs coupled to a darlington VAS is featured in Fig 7.6a.

He then followed up with cascodes in the VAS and lastly the drivers and the lateral mosfets in the current stage are in CF fashion.

I have checked the diy forum titled "Unstable VAS current in amp from Slone book" and I saw the Stochino-based amplifier which incorporates also current mirrors in symetrical input stage (I already had this schematic but I don't know why I have fogotten about it...) I do not have the book on Slone circuit (does anybody can scan it and post it?)

slowhands said:


The Hafler XL-280 and XL-600 incorporate the higher front-end voltage in their PC-40 driver board (circa 1987). This board uses complementary monolithic JFET duals in the "J. Curl" configuration on input, and the Lender VAS, but no current mirrors.

The Self book really awakened audio amp designers to the crucial importance of input stage balance for minimal distortion, and mirrors are one tool to attain it. So far this thread has pointed out two satisfactory means to implement them in complementary symmetry amps; neither is the method I alluded to earlier. Apparently it is not merely possible, but also old news.

The XL-280 includes the higher front-end voltage but not a regulated one as I have suggested in this thread. The difference between both options is very different. If we can combine both of them, then it is even better (higher clipping voltage before degradation of performance and more stable DC operating point of transistors. Regarding J-fet at the input pairs, I know Mr Borbely uses J-fet a lot but what are the pros and cons?

What is the method you have "alluded to" earlier for current mirror use in sym input amp?
 

fab

Member
Joined 2004
Paid Member
mcp said:
He then followed up with cascodes in the VAS and lastly the drivers and the lateral mosfets in the current stage are in CF fashion.

The cascode in the VAS is one of the mods I have done in my DH-200 amps and I am pleased with the results. What is the best base voltage for the cascode transistor?
"...current stage are in CF fashion": Slone's circuit required please.