DAC60 vs. Benchmark vs. PS Audio Link III

regal · « **Reply #160 on:** 14 May 2008, 02:28 pm »

The choke did clear things up, but I added the Jensen caps at the same time. Also I am noticing the slight treble break-up is smoothing out as the Jensens break in.

I think changing the plate and cathode resistors should be considered mandatory. An amperex 6DJ8 is only rated to 15 mA, the DAC-60 is set to 20 mA bias. A 6922 is borderline as they are rated 20 mA max. The 330 ohms gives a 6 mA bias.

The 330 ohms is probably being too conservative (I want my tubes to last a while.) You could possibly get better sound with 200 ohms, which is a 10 mA bias.

markC · « **Reply #161 on:** 14 May 2008, 09:26 pm »

Great, I'll just cheap out and reuse the 220r's that I pull from the I/V and save on parts and reuse at the same time. Parts Connexion is supposed to start carrying Elna caps, so once they do I'll order up the necessary parts and have another go at it.

regal · « **Reply #162 on:** 15 May 2008, 01:13 pm »

Quote from: markC on 14 May 2008, 09:26 pm

Great, I'll just cheap out and reuse the 220r's that I pull from the I/V and save on parts and reuse at the same time. Parts Connexion is supposed to start carrying Elna caps, so once they do I'll order up the necessary parts and have another go at it.

You need 4 total 220 ohm resistors. So order two.

regal · « **Reply #163 on:** 17 May 2008, 06:45 am »

Eureka! I couldn't understand why my DAC-60 had slighty harsher treble than my rich friend Spoiler DAC. The Spoiler is Emperical Audio's modified DAC-60 which sells for a few thousand $'s. I knew it wasn't the USB input because the SPDIF input sounded the same to us. On the spoiler the DF1704 digital OS filter is set to slow roll-off mode. On the stock DAC-60 it is brickwalled.

Caution this mod is not for the faint of heart. You need a good soldering iron with a .5 mm bevell tip. Set it for 600F or so. Then with a pair of sharp needle point tweezers you can lift pin 27 while heating the pad. You need bifocals and a LED headlamp to pull this off. Then solder a wire from the lifted pin to a +5V. There is an inductor right by the DF1704 that supplies +5V. Notice I don't like giving component #'s because I believe it is very possible Lite changes the silkscreen.

The difference in sound is amazing, this is really the best mod I have done, the treble is smooth now, like a NOS DAC without the distortion. This is also the most risky and difficult mod, if you fry the DF1704 it will be very difficult to replace.

markC · « **Reply #164 on:** 17 May 2008, 01:33 pm »

Yea, I've studied that chip and it is so damned small! I'm worried about frying the chip by flowing a couple pins together, 'cause it's so hard to see. Congrats on pulling it off.
The Spoiler has selectable slow or fast roll-off, I belive.

markC · « **Reply #165 on:** 21 May 2008, 02:53 am »

I run the filter/srpp circuit on spice simulation software, (with the help of my friend Tony), yesterday and it is apparrent,(if the simulation is accurate-which it always has been in the past), that Regal is indeed correct that the filter is a hack job. The response is down 1 db @ 10K and 3 db @ 20k. This could be the reason for the complaint of the Dac-60 being a "a little soft on the top end." Then it takes a huge spike up @ about 55k.The good news is that the draw on the tubes is just over 12 mA which is fine for 6922's but pushing it for 6dj8's. I played with the circuitry and it seems like eliminating the 2nd cap in the filter and replacing the first one with a .047u along with swapping the 220r for a 130r in the I/V goes a long way in giving a nice fr curve. This drops the output by about 4.5 db; should be O.K as when compared to several other dacs the output is about 3-4 db hotter.
I ordered the necessary parts today and hopefully will have the mods done on the weekend.

regal · « **Reply #166 on:** 21 May 2008, 12:57 pm »

Thanks for the update Mark. Don't trust the 6DJ8 spice models for the tube current, just measure the V drop across the cathode resistor.

I have been chewing on your impressions with the 1730+tube. The 1730 was a subpar chip, I really think the SRPP tube stage of the DAC-60 is holding it back. According to the articles on TubeCad journal, the SRPP is the wrong tube circuit for a linestage. Also the Spoiler has this changed to something different (looks CCS loaded.)

Here is a work up of Broskie's recommendation:

This is a work in progress and not tested. I would appreciate any feedback.

MarkM · « **Reply #167 on:** 21 May 2008, 01:21 pm »

I have to say thanks to Regal and MarkC for really digging into the Dac60. A couple of years ago, I did some minor mods to my dac60, Cree diodes throughout, choke, Blackgates and Jensen 4 pole filter caps for HV. I thought it was a great deal with the minor mods I did then...

Looks like this thing can really be taken to another level, thanks.

markC · « **Reply #168 on:** 21 May 2008, 07:35 pm »

Here's a crappy scan of the FR of the dac-60, (as simulated on spice software).
The blue is original, (mine has .22mH inductors in the filter although the silkscreen says .39mH), the red is with a 131 ohm resistor instead of 220 in the I/V, the second cap in the filter removed and the first cap changed to a .047uf.

markC · « **Reply #169 on:** 27 May 2008, 02:21 am »

I tagged on a 340 ohm resistor to the 220r in the I/V circuit. This made a very noticeable change. I'm sure that we all understand that this is my system, my room. blah, blah, blah.
Prior to this, I had noted that even with all the previous mods that there was still the same annoyance..the ring I could detect on dynamic piano notes and the glare I could also detect on dynamic vocals.
These annoyances disappeared. Everything was a lot smoother and less edgy. Still, a bit of harshness in the treble existed. I then removed the 2nd cap in the filter circuit and replaced the remaining one with a .047uf.
It's sounding very nice now and if there's any truth to component burn-in, it may get even better.

Many thanks to Regal for bringing this to my attention.

regal · « **Reply #170 on:** 27 May 2008, 12:45 pm »

Marc, I am glad you like the improvements but hate to tell you that there are more gains in SQ by going down to 50 ohms like the filter I posted. Over 100 ohms and the internal diodes can still clip in fast transients, some designers on DIYHIFI even believe 10 ohms is optimal but this would require an expensive step-up transformer. If you don't want a bypass capacitor you could use an LED to get the gain increase needed to go to 50 ohms.

markC · « **Reply #171 on:** 28 May 2008, 02:15 am »

I am quite sure that you are correct, Regal. At this time, though, I'm going to sit back a while and live with what I have accomplished. I will drop the value of the I/V resistor as far as feasible with my current layout. There's still room to go down to about 110 ohms, ( according to spice), so I will do that soon. I did measure the voltage drop across the cathode resistor after the last mod and by my calculations the mA draw is 13.8.
All in all, the current state of this Dac is leaps and bounds beyond stock IMO.

rollypolly · « **Reply #172 on:** 28 May 2008, 06:39 pm »

Might be worth exploring Gordon's tubey version of the I/V using a grounded grid scheme aa. Even Hagerman was impressed.

Rolly

capo · « **Reply #173 on:** 28 May 2008, 07:35 pm »

I am just checking into this thread again after my last post under the "Cappy" moniker. I unfortunately created an extra account a few years ago. Most of my posts on AudioCircle are using "Capo".

Anyway, wow, I see some great stuff has been going on.

Fortunately, my Lite Dac 50 is hard wired to slow rolloff. I would be nervous trying the Regal technique!

Ayre has a CD player that can choose between sharp and slow rolloff in the pcm1704 - the slow rolloff switch says "Listen" and the sharp rolloff says "Measure". Charles Hansen of Ayre mentions that while the sharp rolloff has slightly better frequency performance, it isn't worth it because of the time domain problems it causes -- smearing, ringing, and such.

MarkC - thanks, those are very interesting filter results. So the filter is aggressive. And there is actually a third pole too because of the Miller Capacitance in the tube. So it is more rolled off than even your simulation, probably. I'm still going to try to pull the filter out completely, and stick in a 220 ohm Allen Bradley carbon composite grid stopper instead. Combined with the Miller Capacitance in the vacuum tube, this should give a much less aggressive RC first order filter. I'm thinking resolution will be improved by getting rid of the two inductors in the signal path as well as restoring high frequencies from getting rid of the aggressive filter - we will see. Another idea if that doesn't pan out is putting a pF range silver mica cap across the I/V resistor for a milder filter.

Also, Mark, you might try sticking in a Mills resistor for I/V - very smooth but balanced sounding.

markC · « **Reply #174 on:** 29 May 2008, 02:55 am »

When I get enough parts together for a minimum order, I will try the Mills. I use them in my speaker crossovers and am pleased with what they do and don't do.
I have a ton of listening ahead of me b4 I do any more surgery to the board. I like the current sound but need more hours of listening to decide whether or not I am done yet.
One thing I haven't yet mentioned is that during the last mod procedure, one of the wires to an RCA broke. Only been flexed a few times. I replaced them all with silver coated copper, teflon insulated.
I'll run a few more filter configs. on B2 Spice and see what it shows.
I have confidence in this software as this is what Tony and I have used for the design of our speakers, amps and dacs.

regal · « **Reply #175 on:** 29 May 2008, 05:10 am »

Quote from: capo on 28 May 2008, 07:35 pm

Fortunately, my Lite Dac 50 is hard wired to slow rolloff. I would be nervous trying the Regal technique!

I have seen other reports that the DAC50 SRO pin was at +5V (slow roll-off) even though the schematic shows it grounded. Lifting the pin on the DF1704 was actually more nerve racking than it was difficult, it was the biggest sonic improvement of all the mods I have done.

MarcC , I have been considering buying Spice to model this Compound mod for the tube stage. Do you know where I can get the Spice program ?

markC · « **Reply #176 on:** 29 May 2008, 09:14 pm »

Regal; B2 Spice v5 is what I use. It's by Beige Bag. I'm sure if you Google it, you'll find it. There's a free trial version but it is really limited in the parts count that you can model. I think it allows 23 components all told including grounds.

capo · « **Reply #177 on:** 31 May 2008, 05:57 pm »

I took out the two pole analog post dac filter a few days ago. I replaced it with a single 180 ohm Allen Bradley grid stopper resistor, acting as a -6 db filter combined with the Miller capacitance in the tubes.

The good news is that resolution improved across the board, from the bass on up. The bad news is that a lot of music is "harder" sounding, with a glare around 3-5k. I'm thinking the aggressive filter hid the flaws in the SRPP output stage and the power supplies. More thoughts later...

regal · « **Reply #178 on:** 1 Jun 2008, 07:01 am »

I agree that the SRPP stage is the limiting factor after fixing the filters and adding the choke. An SRPP is not meant for a large imepedance load like an amplifier. A few posts back I posted Broskie's improvement to the SRPP which is a compound version. I am going to try the modification next week.

Also I discovered that Hagerman is using this design on his Chime design (minus one CCS.) I updated the schematic, but the 500R resistor should be a trimpot to adjust the voltages. Or you can set it using a 9V baterry then determin the required resistor.

capo · « **Reply #179 on:** 1 Jun 2008, 05:04 pm »

Regal,

Thanks for the schematics. They are interesting.

Regarding the B+ power supply, I think it still has some problems, even with the choke. It is still full of electrolytic caps, even after I swapped in the Obbligato and Solen as second filter caps. There is the 120 uF first cap, two 47 uF electrolytics in the regulator, and a 2 uF in the regulator. The remaining electrolytics could definitely be contributing to the now sometimes weedy lower treble. Speaking of the regulator, that is a suspect too. Probably it is too much trouble to change any of those things though, and no more room, except maybe for the 2 uF regulator cap, particularly if like you suspect it is the SRPP stage that is the ultimate culprit.

The low voltage supplies are also not state of the art and are probably part of the problem. They use LM317s which many audio designers despise. The analog and digital supplies to the pcm63 are also shared by a single regulator. I've read that the DF1704 also wants the cleanest +5v supply possible. Better shunt regulators swapped in here might make sense.

I'll be able to compare the SRPP stage against a better tube output stage when my Aikido is completed in a few weeks. That will be interesting and may narrow down where the problem is coming from.

At this point there isn't any going back to the old filter. There is just too much of an improvement in resolution. It also got rid of the last of the "chocolate" sounding tone coloration in the mids. And the highs 8k+ are a lot nicer. This change was just as big as any of the others. It does feel like three steps forward, one step back though.