 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
Today I was looking again in my Denon DA-500 DAC and was thinking: there must be a better output circuit possible than the one with the 4 used opamps. (I use 2xAD825 for I/V and buffer, OP275 for GIC filter for whatever that means :-) The sound is not *really* bad, but at least too bright.I saw the posting about using only 1K resistor at the output of the PCM1702. I wonder what happens with all the high frequency (>22K) information at the output of the DAC ? Is it smart to send everything directly to the amp. (I use a Harman Kardon HK690)
As a second option, I would like to make an I/V converter with the AD825, and also an output buffer. Can I safely keep out the filter? What surrounding components do I need? I do not want to spend too much money because I will probably build a TDA1543 TNT dac soon...
Thanks in advance for any reply...
Fedde Bouwman
The Netherlands
Follow Ups:
Hi Fedde,a couple of extra suggestions in addition to Dave's.
Someone posted here recently with good results after biasing his AD825s into class A - do a search for details. Simply connect a resitor of around 5K between the output and the negative rail. If you like the results, you could try using constant current diodes of say 3mA instead for more stable biasing. Class A working usually results in sweeter sound.
Also, try bypassing the filter - this should be very simple to do in the case of a GIC. Listen at low volume to start with. If it sounds bad, then you need a filter. Try then replacing the OP275s with OPA604s (2604 if you need duals). These op-amps are well know for their smooth, warm sound.
I would keep it to 100 ohms or less with the PCM1702/1704 DACs, and even that is still too high for best sound quality. But if you are gonna use an opamp anyway, why use resistor I/V conversion? I guess you could try it to see which way sounds best but there will always be a little softening of the sound with a resistor I/V in my experience so I would just use a good opamp as the integrator. Or use a l00 ohm (or preferably smaller) resistor followed by a good quality discrete amp stage. Or another very good opamp alternative is to use a transimpedance amp like the AD811 configured as the I/V convertor. They sound very good in this function and it works well for driving the cable as well to make a simple, single opamp output stage. Doesn't sound as good as the simple discrete current mirror approach that I talked about in an earlier post, but still very good and has been used in many high end type products and is very simple to implement since it can usually be used without any coupling cap. I tried the AD811 convertor without any filtering and also with a single pole filter without any problem, but I don't have your amp and speakers so can't really advise you there. I normally run the aforementioned discrete current mirror type I/V convertor without any filtering except for slight bandwidth limiting on the high end for RF.Dave
Thanks for your reaction !I am afraid that I have a lot more questions...
What is a transimpedance opamp ? Is there a way to make something good with the current opamps (AD825/OP275) ??? What would be the exact circuit then ?
Will the spectrum of the band till 22 Khz be mirrowed in higher frequency bands with much power (without filtering) ?
At this moment I do need feel much for buying the AD811's as they will cost me 10 dollar a piece, while I have some other good quality ones.Fedde
You can read all about transimpedance amps just by going to Analog Devices and reading the datasheet for the AD811. The circuit I used was similar to the one Walt Jung popularized years ago in Audio Amateur. I could probably dig it out if you are really serious but don't have a copy of the schematic handy right now.Yes, you can certainly make a good sounding DAC output stage using any number of modern opamps and I am not meaning to imply you can't. I was just stating my preference. Most integrator type circuits are very similar and probably not much different from what you have now. I'll leave it to others to answer your questions about the AD825 since I have not used it before.
If you are using a standard 8x oversampling design, the out of band components will be shifted to a much higher bandwidth and don't pose much of a problem to most components since they generally include input filtering, but it is always prudent to include a certain amount of filtering to avoid any potential for problems. You can usually accomplish a first order filter in the integration stage with an additinal capacitor and this may be enough for your needs. Many of the opamp datasheets show suitable I/V conversion circuits that could be adapted to your needs.
Dave
I just ripped out the OP275's. Sound is *much* better now :-)
The brightness is gone. There is still some distortion/noisiness. Jitter remains a big problem with the denon, probably caused by the yamaha chipset. I maybe have to try to upgrade the switching IC's for the channels. Also my digital interlink is not really good...The complete analog system is now:
- capacitor from dac to ground
- from dac also to the - of AD825-I
- AD825-I feedback of R/C parallel and + input on earth
- AD825-I output to voltage divider followed by RC filter
- after RC filter a AD825-II buffer
- then 15 ohm resistor to lineoutThe two AD825's are on a small PCB. After reading all the info on gaincards/clones I tried to decrease the feedback length. Wow! It really matters. I soldered the direct feedback of the buffer and the RC feedback of the first opamp near the SMD opamps. The DAC is *much* faster now, there used to be some sort of a echo.
And thanks for giving me the courage to remove the GIC filter ;)
Yesterday I was modding my DA-500 again. For a long time, I was worried about the differences between the two channels. Right was sounding great, but left had a little distortion and sounded somewhat uncomfortable. Due to the differences imaging was foggy. I found out that one of the caps around the PCM1702 was not connected on one side!!! I *really* wonder how is that possible. Probably has never been connected. Earlier I found out that one cap (!) was missing from the same DAC. The last owner must have fun with it or the assembly department has been sleeping!But anyway: the sound :-)))
I use it directly in the power amp of the Harmon Kardon HK690 with DIY Flatline TL speakers. So the volume has to be turned down digital for this moment, not ideal. But the sound already rocks!!! So natural, relaxed and the soundstage is fantastic. Only the bass is powerless, with preamp the bass is better. How's that possible? In the end of the DA-500 circuit I have a AD825 buffer with 15 ohm resistor to lineout. Is the resistor the problem ? Is it safe to omit it? I think the AD825 should be able to deliver enough power to drive an outputstage???Thanks in advance for any comments...
Hi Dave,I remember your saying that sound quality improved down to around 10 ohms - I take it you used a discrete stage afterwards for gain. Did your current mirror stage improve on this setting too? I'm not a fan of resistive I/V conversion, but this preference is based on using 100 ohms. I'm tempted to try transformer, too, if only to provide ground isolation at the DAC.
The other post seems to have some kind of problem that won't let it display the image and won't let me delete or respond to it so this is a repost.I did use a fairly simple discrete stage after the I/V resistor and was actually quite happy with the sound. It was a no feedback type design with a gain of about 30 or so and easily bettered opamp type equivalents. It was only when I couldn't get an answer from BB about voltage compliance that I decided to conduct my own tests and that led to my converting over to the current mirror design which sounded much better.
Above is a simplified schematic of my present output stage. I also use an offset control circuit which acts on the current source to avoid any coupling capacitors but would prefer not to show that part since it required quite a bit of development to implement one that didn't significantly degrade the sound, especially in a no feedback design. I also have left the values off of the filter components since they are somewhat dependent on other circuit values and I don't generally use them on my own DAC. I directly drive the amp via a low resistance precision wirewound pot so it is biased very high, almost like a small power amp.
The transistor pair Q1/Q3 is just a classic current mirror. Since the bases are common, the emitters must stay close to the same potential as well. The emitter of Q3 is connected to ground and hence Q1 must follow which keeps the DAC output at a virtual ground. This is a no feedback design (other than the local resistor degeneration in the current source) and sounds very good. In my tests it sounded better than a resistor I/V and discrete gain stage, but remember with a 10 ohm resistor I/V you would need to follow it with a voltage gain of about 300 so it would actually be more complex than this circuit and would still not be ideal. When testing different values for the I/V resistor, I kept the gain stage constant and adjusted volume with a pot on the amplifier input to match levels. The values in this simplified circuit are for a PCM1704 DAC with a full scale current output of 1.2mA. The current from the DAC output will be reflected through R7 (773 ohm) which you could call the I/V conversion resistor. The voltage across R7 will then be amplified a bit further by Q6. The gain of the cascoded, single-ended output stage is set at about 3 by the source and drain loads. One could always skip the output stage and increase the value of R7 and take the output at the junction of R7 and Q1 through a large value capacitior (and I believe Jan Didden presented a design that did just that many years ago in The Audio Amateur) but I prefer to avoid capacitors whenever possible and I like having a high current stage to drive the volume control, cable and power amp.
One should note that there are many other different current mirror configurations that will work well and many will deliver a lower dynamic impedance because of negative feedback, but this no feedback design was the best sounding of the large number of designs I tried.
Dave
Hi Dave,thanks for the useful information. You are an oracle for all thigs digital.
I've had a circuit using an LM394 for the mirror lying around ofr a while - the designer seemed to think that matching was important (shrug). This one didn't use a buffer on the output though, so was very simple. I'm guessing the output impedance would be roughly equal to the loading resistor, which shouldn't be a problem driving my 100K pot.
Maybe I'll botch together a hybrid, just so I can call it mine :o)
... would the value of the current mirror's loading resistor that determine Vout affect linearity at all? I'm wondering, as the concept of going to a lot of trouble to get our 2V rms out of a DAC, only to throw much of it away at the attenuator seems wasteful - maybe the loading resistor could be made variable.
in this case it wouldn't work because it is also used to set the DC bias of the output stage. I don't quite follow what is gained by changing the voltage at this point instead of later in the chain since the you will get better signal to noise ratio the farther downstream it is attenuated and you still would need the capability of full voltage output from the DAC so it's not really wasteful. I have my attenuator incorporated into the amplifier input circuit so it's pretty transparent anyway.Dave
Well, with a traditional setup you get added noise from the resistive attenuator, plus it usually needs a drive afterwards. It seems philosophically pleasing to me to get exactly the voltage you need from the I/V stage, then feed it staight to the power amp. Not practical of course. Philosophy may not translate very well into practice, either.
Hi Dave and other DAC expects,The discrete I/V you'd shown is really impressive. Is C1 used as one of the pole in the low pass filter or just for stability?
Btw, I am working on another zero feedback I/V stage, using OPA660 diamond transistor. I would like to know if there is anyone having the same idea. Anybody every use this device in whatever applications?
Thanks in advance.
http://www-s.ti.com/sc/ds/opa660.pdf
C1 is used as one of the filter poles (in combination with R7) but isn't required for stability in my case although it might be helpful with a different layout or design. C2 can be any high grade capacitor (I use a bidirectional TVS) for a bit of RF rejection.I haven't seen the OPA660 before. How do you propose to configure it?
I've just receviced some samples. They are in SMD package, I'm building prototype PCB to have a test.I'll try to connect them in common "B" config, a current control current source with low input imp. With 1Kohm "C" loading, buffered by an OPA637 preamp.
I believe Wadia is using something very similar in their "Swift currrent technology". OPA660 is the nearest I can find, but definitely not identical.
I'm still struggling with the prelimary schematic, and a long way to go......
I did use a fairly simple discrete stage after the I/V resistor and was actually quite happy with the sound. It was a no feedback type design with a gain of about 30 or so and easily bettered opamp type equivalents. It was only when I couldn't get an answer from BB about voltage compliance that I decided to conduct my own tests and that led to my converting over to the current mirror design which sounded much better.Below is a simplified schematic of my present output stage. I also use an offset control circuit which acts on the current source to avoid any coupling capacitors but would prefer not to show that part since it required quite a bit of development to implement one that didn't significantly degrade the sound, especially in a no feedback design. I also have left the values off of the filter components since they are somewhat dependent on other circuit values and I don't generally use them on my own DAC. I directly drive the amp via a low resistance precision wirewound pot so it is biased very high, almost like a small power amp.
The transistor pair Q1/Q3 is just a classic current mirror. Since the bases are common, the emitters must stay close to the same potential as well. The emitter of Q3 is connected to ground and hence Q1 must follow which keeps the DAC output at a virtual ground. This is a no feedback design (other than the local resistor degeneration in the current source) and sounds very good. In my tests it sounded better than a resistor I/V and discrete gain stage, but remember with a 10 ohm resistor I/V you would need to follow it with a voltage gain of about 300 so it would actually be more complex than this circuit and would still not be ideal. When testing different values for the I/V resistor, I kept the gain stage constant and adjusted volume with a pot on the amplifier input to match levels. The values in this simplified circuit are for a PCM1704 DAC with a full scale current output of +/-1.2mA. The current from the DAC output will be reflected through R7 (773 ohm) which you could call the I/V conversion resistor. The voltage across R7 will then be amplified a bit further by Q6. The gain of the cascoded, single-ended output stage is set at about 3 by the source and drain loads. One could always skip the output stage and increase the value of R7 and take the output at the junction of R7 and Q1 through a large value capacitior (and I believe Jan Didden presented a design that did just that many years ago in The Audio Amateur) but I prefer to avoid capacitors whenever possible and I like having a high current stage to drive the volume control, cable and power amp.
One should note that there are many other different current mirror configurations that will work well and many will deliver a lower dynamic impedance because of negative feedback, but this no feedback design was the best sounding of the large number of designs I tried.
Dave
Great looking I/V convertor, Dave!A couple of questions....
I take it, the input is connected to the DAC's I/V out and ground. Is the "4 ohm imput impedance" you list a resistor to ground?
Also, how tightly matched (and at what parameters) do the 2N5087's have to be?
Thanks,
Steve
Yes, the input shown in the schematic is connected to the analog output pins of the DAC. The 4 ohms just refers to the measured dynamic input impedance of the I/V convertor which is what the output of the DAC will see (no added resistor is used). In other words, if you connect a scope to the output of the DAC you will measure less than 4.8mV peak at the full scale output of 1.2mA. It helps to have closely matched transistors for current gain in the mirror but isn't really necessary if you have some type of offset control. R5 can be adjusted to control the bias current in Q3 to initially set the DAC output (P7) to zero volts after warmup. Just about any type of transistor can be used in the current mirror but the 2N5087 is cheap and quiet and has good gain. I first used a mosfet pair since I generally prefer the sound of mosfets but couldn't get much below about 15 ohms dynamic impedance so switched to bipolar. You can also change the polarity of the transistors, using a NPN pair for the current mirror and current source and P-channel mosfets for the output but I prefer to use the N-channel mosfets in the output for a variety of reasons which dictated the choice of PNP for the input section. You could also use bipolar for the outputs too but you need to use care in the drive circuit and can't just substitute in the cicuit as it is shown. The power supply rails can be increased too with normal dissipation precautions. You can't really go much lower though unless changes are made to the circuit because of the cascode voltages and other factors. The output stage is run at very high current so must be cascoded with a power transistor as shown (Q7) to keep the voltage across the gain transistor (Q6) minimized since it is just a small signal mosfet in a TO92 type plastic package. Using the small signal mosfet minimizes the capacitive loading of the input stage so keeps internal bandwidth high. The circuit is built on a plugin module and hence the "P" designations used on the schematic which are the PCB connections to the motherboard.Dave
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
