Is the 30+ multi tone test more stressful on D to A, than TIM ?
No, it's spectrum is strictly limited to 20 kHz. TIM test contains harmonics well beyond 100 kHz.
Is the 30+ multi tone test more stressful on D to A, than TIM ?
With all due respect to those on this forum, it seems that judging from past measurements there were a few flaws. And I suppose we could postulate on what is happening.
For instance on post 526, it shows high noise levels where the least significant bits lay for full Redbook resolution.
But, it would be interesting to get some concrete evidence of any changes made with Sunlight.
It could then substantiate what people are saying. After all, this is a science forum.
Uh, no. Any quantiser is by definition non-linear, the 1-bit variety being the most extreme. That's why it's so tricky to make them work well.A 1 bit modulator is theoretically linear,
Uh, no. Any quantiser is by definition non-linear, the 1-bit variety being the most extreme. That's why it's so tricky to make them work well.
You must be using the word linear differently from the rest of us.On playback if you do a pure 1-bit implementation, no noise shaping, it is theoretically "linear" in the audio pass band. It will be awful, with wicked susceptibility to jitter, massive ultrasonics, etc. but "linear". I probably should have cut and pasted more of what he wrote to give context w.r.t. what he was talking about, which was a "pure" 1-bit DAC. My use of wording obviously sucked too
On playback if you do a pure 1-bit implementation, no noise shaping, it is theoretically "linear" in the audio pass band. It will be awful, with wicked susceptibility to jitter, massive ultrasonics, etc. but "linear".
You must be using the word linear differently from the rest of us.
I think you will find that the Holo Audio is not a pure single bit modulator without filtering/averaging. They also dedicate a lot of effort to providing the purest clock signal.
Well, the data in that measurement is originated from my oversampling filters and modulator so I know exactly how it has been created. Spring 3 operating as bit-perfect D/A converter. It is not so different from my own DAC design or some other discrete DSD implementations I've been using and measuring (T+A, etc).
Why does the graph start at -90db?
Bit-perfect DA converter does not mean there is no averaging/filtering.
Hi, @amirm Does it make sense to replace the stock PCW150/150 Edcor output transformers with Edcor XS4400 transformer to improve the measurements and sound? The Directstream DAC is designed to take XS4400 transformer. Thanks, AlanThis is a review and detailed measurements of the PS Audio PerfectWave DirectStream DAC. It is on kind loan from a member. The unit that I have has the network "bridge" streaming option and costs US $6,899. Without that option it costs US $5,999.
The DS DAC is quite heft with a textured matt finish:
Other than a power button, you have the touchscreen to use for all other functions. I found the display graphics to be lackluster for a product at this price. It is hard to touch buttons even though I don't have large fingers (no jokes about the size of my mouth please). It actually looks better in the picture above because I used photoshop to increase its contrast. Otherwise it has a blue background which seems to suffer from lack of backlight uniformity. On the plus side, the normal display shows the details of the format which I appreciate. And changes in volume using the remote results in a large numerical values of the current level. So it works but misses an opportunity to impress aesthetically.
Back panel shows all the usual suspects plus dual I^2S ports:
View attachment 34133
For my testing, I mostly focused on USB input and XLR outs unless stated otherwise.
I call the PerfectWave DirectStream a boutique DAC meaning it implements the DAC functionality on its own rather than using an off-the-shelf DAC. In this case, it takes PCM and converts it to high-rate, 1-bit "DSD" and then uses a transformer for low pass filter and balanced output. The former is typical approach but the latter, was a head scratcher for me. We use transformers in audio when we have to, not because we choose to. More on this later.
Ted Smith is the designer behind this DAC (a software guy who used to work at Microsoft the same time I was, turned hardware designer) and has that great professorial look. Stories he and rest of the gang at PS Audio tell of the superiority of DSD format and simplicity of implementing it in hardware. Here though, an FPGA is used with all of its ancillary logic which throws out any arguments for simplicity. Certainly the cost shoots way high.
A rather clever scheme is used to extract the sample rate of the incoming stream and ignore its clock. For USB this doesn't matter anyway but for other ports, the claim is made that incoming jitter doesn't matter and hence all the inputs sound the same. I investigated this and report below.
Someone asked me about version and firmware numbers. I could not make sense of PS Audio support page as to what software package went with what. Guessing, I downloaded one and it seemed to be lower version number than the one I tested. So I did not mess with it.
Edit: here is the version info:
I am told 3.0.6 is the latest version.
EDIT: right after this review was completed, the new "Windom" update was made available. It made no difference in performance. See here: https://www.audiosciencereview.com/...perfectwave-directstream-dac.9100/post-239292
And oh, as you see the cling plastic is still on top of the unit. The unit was just purchased and the owner expressed dissatisfaction with its sound relative to current DAC. He offered to unit to me to test to see if we can find any clues regarding that.
For testing, I downloaded PS Audio driver package which had an ASIO interface which I used for my testing. Before that, I used the stock windows driver and it worked the same.
DAC Audio Measurements
View attachment 34134
PS Audio provides the following specs:
View attachment 34135
We see that we are right on the money with output voltage. Unfortunately that is much lower than nominal 4 volt that should come out of the XLR output. For such a beefy product, it is strange to see such a shortfall. Same issue exists on RCA outputs as they note with 1.4 volt output instead of 2 volts (I did not verify this).
Our measured THD+N of 0.016% is much better than 0.03% they spec. But before celebrating, we need to consider that this awful performance. When converted to SINAD, we get only 75.5 dB which firmly places the PerfectWave DirectStream DAC at the bottom of the rankings of all DACs tested so far:
View attachment 34136
From my rough count, there are 171 DACs in the above graph and DirectStream DAC ranks #10 from the bottom.
High-end dacs always advertise ability to present detail so let's see how it does on dynamic range:
View attachment 34137
Gosh, we can't even clear 16 bits of audio hear let alone have lower noise floor than other DACs.
Poor low level linearity (caused by noise) confirms the same assessment:
View attachment 34138
Here is the spectrum of said noise:
View attachment 34139
Despite my use of balanced XLR output, we have mains leakage. And on top of that, power supply related spurious tones. And some odd tones at 80 Hz and such. Perhaps coming from the microprocessor controlling the front panel?
Filter is analog and created out of the combination of a capacitor and inductor on the primary side of the output transformer:
View attachment 34140
I am fine with this as it nicely cuts right off at 22.05 kHz but sure goes against the trends of "high-end audio dac" these days with gentle slopes and such. Same is confirmed in square wave output:
View attachment 34141
You get your "pre-ringing" so audiophiles hating that should run away. Not!
Here is the jitter and noise spectrum at 48 kHz sampling:
View attachment 34142
We have the low frequency spikes that we saw in the noise floor analysis. Outside of that, the spectrum is very clean and just compromises of noise. That noise floor is high due to poor dynamic range of the unit. I worried that it was hiding something. I increased FFT depth and averaging but nothing came out of that.
Next, I hooked up Toslink from my APx555 analyzer. It too showed identical output to above. I then turned on jitter on Toslink and no matter what I did with sine wave jitter, nothing would disturb the DAC's output. Or its ability to lock on it. So this claim regarding input independence and ignoring input jitter is correct. Nicely done.
High noise floor and general distortion of course hurts in intermodulation test versus level:
View attachment 34143
On top of that, we have output saturation starting at around -8 dB.
Multitone test starts to reveal new issues of interest:
View attachment 34144
This is something we have not seen before. Intermodulation distortion between the tones is much higher at low frequencies than high. That pushes the distortion floor to just 60 dB. That gives the stunningly low distortion-free dynamic range of just 10 bits! Yes, this is bad. Really bad. Let's keep going with THD+N versus frequency:
View attachment 34145
Let's ignore for now that the entire curve is way, way up there (i.e. is bad). We have the same rise in low frequency distortions we saw in multitone test. What explains that? Use of output transformer! Ted Smith said he built his prototype using Jensen transformers. Jensen is the king of high-quality audio transformers. But high-quality does not mean high performance in vocabulary of audio DACs.
Bill Whitlock is the foremost expert in transformers (and grounding) at Jensen so let's go to his bible on transformers here: https://jensen-transformers.com/wp-content/uploads/2014/09/Audio-Transformers-Chapter.pdf
And focus on this distortion versus frequency:
View attachment 34146
Distortion starts to rise from 200 to 300 Hz as we go lower in frequencies -- precisely what we see in our measurements.
I investigated the high frequency distortions and they are clearly harmonic distortions, not some ultrasonic junk:
View attachment 34147
For this test (and with THD+N vs Frequency) sampling rate is set to 192 kHz so we don't have to deal with aliasing components. As you see in red, with a 10 kHz tone, the distortion products are much taller than at 1 kHz (in blue). They line up almost perfectly on multiples of 10 kHz.
In that sense then, our dashboard test at 1 kHz was very beneficial to DirectStream DAC. It provides its best performance there. Above and below, SINAD drops to 60 would would firmly put it in the category of worst DACs we have ever tested.
Frequency response was flat:
View attachment 34148
Listening Tests
For subjective testing, I chose to use the recently reviewed and superb Monoprice Monolith THX 887 Balance Headphone Amplifier. This Monolity has vanishingly low distortion and hence is completely transparent to DACs being tested. For the alternative DAC, I used my everyday Topping DX3 Pro 's line out RCA to Monolith. I then used the XLR input to connected the DirectStream DAC. Once there, I played a 1 kHz tone and used my Audio Precision analyzer to match levels using PS Audio's volume control. PS Audio claims perfection there ("bit perfect") so I figured they can't complain about that. The final matching was 0.3 dB difference between the two.
For headphone I used DROP + MRSPEAKERS ETHER CX with its XLR connection to THX 887 amp.
I started the testing with my audiophile, audio-show, test tracks. You know, the very well recorded track with lucious detail and "black backgrounds." I immediately noticed lack of detail in PerfectWave DS DAC. It was as if someone just put a barrier between you and the source. Mind you, it was subtle but it was there. I repeated this a few times and while it was not always there with all music, I could spot it on some tracks.
Next I played some of my bass heaving tracks i use for headphone testing. Here, it was easy to notice that bass impact was softented. But also, highs were exaggerated due to higher distortion. Despite loss of high frequency hearing, I found that accentuation unpleasant. WIth tracks that had lisping issues with female vocals for example, the DS DAC made that a lot worse.
Conclusions
Many of you would have guessed that PerfectWave DirectStream DAC would not do well on the bench. That is of course true. The real issue in my opinion though is not the internal design of the DAC functionality. After all, there are other companies that use the same approach and achieve superb measured results. The problem here is simple: the output transformer. Transformers are just not linear to the levels we expect to have in a DAC (as opposed to say, in a tube amp).
The sonic effects are there in my semi-formal tests. Perhaps the older audiophiles including the designer Ted Smith, have lost so much high frequency hearing that the harmonic distortion this DAC adds makes up for some of that and they think they are hearing more. If that is the case, just buy a Jensen transformer and put it on the output of your DAC and you would get the same effect!
The fidelity loss here is apparent in both objective and subjective testing. There is nothing to hang your hat on hence the decapitated Pink Panther which leaves no room for such attire.
Needless to say, I cannot in any shape or form recommend the PerfectWave DirectStream DAC.
These companies need to hire trained listeners and perform a simple level matched test before spending so much money on engineering and producing subpar products. Our hobby deserves better. Much better.
------------
As always, questions, comments, recommendations, etc. are welcome.
Our weather is getting colder but I think I am going to jump in the bay and see if that can uplift my spirits after this test. To enable the uplifting, some more money in my bank account would surely help. So please consider donating using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/
No, just give up on this product, it's a lost case.Hi, @amirm Does it make sense to replace the stock PCW150/150 Edcor output transformers with Edcor XS4400 transformer to improve the measurements and sound? The Directstream DAC is designed to take XS4400 transformer. Thanks, Alan
I don't think so, since the main distortions found are not transformer related, so lowering any potential transformer issue won't help much if anyHi, @amirm Does it make sense to replace the stock PCW150/150 Edcor output transformers with Edcor XS4400 transformer to improve the measurements and sound? The Directstream DAC is designed to take XS4400 transformer. Thanks, Alan
Hi,This is a review and detailed measurements of the PS Audio PerfectWave DirectStream DAC. It is on kind loan from a member. The unit that I have has the network "bridge" streaming option and costs US $6,899. Without that option it costs US $5,999.
The DS DAC is quite heft with a textured matt finish:
Other than a power button, you have the touchscreen to use for all other functions. I found the display graphics to be lackluster for a product at this price. It is hard to touch buttons even though I don't have large fingers (no jokes about the size of my mouth please). It actually looks better in the picture above because I used photoshop to increase its contrast. Otherwise it has a blue background which seems to suffer from lack of backlight uniformity. On the plus side, the normal display shows the details of the format which I appreciate. And changes in volume using the remote results in a large numerical values of the current level. So it works but misses an opportunity to impress aesthetically.
Back panel shows all the usual suspects plus dual I^2S ports:
View attachment 34133
For my testing, I mostly focused on USB input and XLR outs unless stated otherwise.
I call the PerfectWave DirectStream a boutique DAC meaning it implements the DAC functionality on its own rather than using an off-the-shelf DAC. In this case, it takes PCM and converts it to high-rate, 1-bit "DSD" and then uses a transformer for low pass filter and balanced output. The former is typical approach but the latter, was a head scratcher for me. We use transformers in audio when we have to, not because we choose to. More on this later.
Ted Smith is the designer behind this DAC (a software guy who used to work at Microsoft the same time I was, turned hardware designer) and has that great professorial look. Stories he and rest of the gang at PS Audio tell of the superiority of DSD format and simplicity of implementing it in hardware. Here though, an FPGA is used with all of its ancillary logic which throws out any arguments for simplicity. Certainly the cost shoots way high.
A rather clever scheme is used to extract the sample rate of the incoming stream and ignore its clock. For USB this doesn't matter anyway but for other ports, the claim is made that incoming jitter doesn't matter and hence all the inputs sound the same. I investigated this and report below.
Someone asked me about version and firmware numbers. I could not make sense of PS Audio support page as to what software package went with what. Guessing, I downloaded one and it seemed to be lower version number than the one I tested. So I did not mess with it.
Edit: here is the version info:
I am told 3.0.6 is the latest version.
EDIT: right after this review was completed, the new "Windom" update was made available. It made no difference in performance. See here: https://www.audiosciencereview.com/...perfectwave-directstream-dac.9100/post-239292
And oh, as you see the cling plastic is still on top of the unit. The unit was just purchased and the owner expressed dissatisfaction with its sound relative to current DAC. He offered to unit to me to test to see if we can find any clues regarding that.
For testing, I downloaded PS Audio driver package which had an ASIO interface which I used for my testing. Before that, I used the stock windows driver and it worked the same.
DAC Audio Measurements
View attachment 34134
PS Audio provides the following specs:
View attachment 34135
We see that we are right on the money with output voltage. Unfortunately that is much lower than nominal 4 volt that should come out of the XLR output. For such a beefy product, it is strange to see such a shortfall. Same issue exists on RCA outputs as they note with 1.4 volt output instead of 2 volts (I did not verify this).
Our measured THD+N of 0.016% is much better than 0.03% they spec. But before celebrating, we need to consider that this awful performance. When converted to SINAD, we get only 75.5 dB which firmly places the PerfectWave DirectStream DAC at the bottom of the rankings of all DACs tested so far:
View attachment 34136
From my rough count, there are 171 DACs in the above graph and DirectStream DAC ranks #10 from the bottom.
High-end dacs always advertise ability to present detail so let's see how it does on dynamic range:
View attachment 34137
Gosh, we can't even clear 16 bits of audio hear let alone have lower noise floor than other DACs.
Poor low level linearity (caused by noise) confirms the same assessment:
View attachment 34138
Here is the spectrum of said noise:
View attachment 34139
Despite my use of balanced XLR output, we have mains leakage. And on top of that, power supply related spurious tones. And some odd tones at 80 Hz and such. Perhaps coming from the microprocessor controlling the front panel?
Filter is analog and created out of the combination of a capacitor and inductor on the primary side of the output transformer:
View attachment 34140
I am fine with this as it nicely cuts right off at 22.05 kHz but sure goes against the trends of "high-end audio dac" these days with gentle slopes and such. Same is confirmed in square wave output:
View attachment 34141
You get your "pre-ringing" so audiophiles hating that should run away. Not!
Here is the jitter and noise spectrum at 48 kHz sampling:
View attachment 34142
We have the low frequency spikes that we saw in the noise floor analysis. Outside of that, the spectrum is very clean and just compromises of noise. That noise floor is high due to poor dynamic range of the unit. I worried that it was hiding something. I increased FFT depth and averaging but nothing came out of that.
Next, I hooked up Toslink from my APx555 analyzer. It too showed identical output to above. I then turned on jitter on Toslink and no matter what I did with sine wave jitter, nothing would disturb the DAC's output. Or its ability to lock on it. So this claim regarding input independence and ignoring input jitter is correct. Nicely done.
High noise floor and general distortion of course hurts in intermodulation test versus level:
View attachment 34143
On top of that, we have output saturation starting at around -8 dB.
Multitone test starts to reveal new issues of interest:
View attachment 34144
This is something we have not seen before. Intermodulation distortion between the tones is much higher at low frequencies than high. That pushes the distortion floor to just 60 dB. That gives the stunningly low distortion-free dynamic range of just 10 bits! Yes, this is bad. Really bad. Let's keep going with THD+N versus frequency:
View attachment 34145
Let's ignore for now that the entire curve is way, way up there (i.e. is bad). We have the same rise in low frequency distortions we saw in multitone test. What explains that? Use of output transformer! Ted Smith said he built his prototype using Jensen transformers. Jensen is the king of high-quality audio transformers. But high-quality does not mean high performance in vocabulary of audio DACs.
Bill Whitlock is the foremost expert in transformers (and grounding) at Jensen so let's go to his bible on transformers here: https://jensen-transformers.com/wp-content/uploads/2014/09/Audio-Transformers-Chapter.pdf
And focus on this distortion versus frequency:
View attachment 34146
Distortion starts to rise from 200 to 300 Hz as we go lower in frequencies -- precisely what we see in our measurements.
I investigated the high frequency distortions and they are clearly harmonic distortions, not some ultrasonic junk:
View attachment 34147
For this test (and with THD+N vs Frequency) sampling rate is set to 192 kHz so we don't have to deal with aliasing components. As you see in red, with a 10 kHz tone, the distortion products are much taller than at 1 kHz (in blue). They line up almost perfectly on multiples of 10 kHz.
In that sense then, our dashboard test at 1 kHz was very beneficial to DirectStream DAC. It provides its best performance there. Above and below, SINAD drops to 60 would would firmly put it in the category of worst DACs we have ever tested.
Frequency response was flat:
View attachment 34148
Listening Tests
For subjective testing, I chose to use the recently reviewed and superb Monoprice Monolith THX 887 Balance Headphone Amplifier. This Monolity has vanishingly low distortion and hence is completely transparent to DACs being tested. For the alternative DAC, I used my everyday Topping DX3 Pro 's line out RCA to Monolith. I then used the XLR input to connected the DirectStream DAC. Once there, I played a 1 kHz tone and used my Audio Precision analyzer to match levels using PS Audio's volume control. PS Audio claims perfection there ("bit perfect") so I figured they can't complain about that. The final matching was 0.3 dB difference between the two.
For headphone I used DROP + MRSPEAKERS ETHER CX with its XLR connection to THX 887 amp.
I started the testing with my audiophile, audio-show, test tracks. You know, the very well recorded track with lucious detail and "black backgrounds." I immediately noticed lack of detail in PerfectWave DS DAC. It was as if someone just put a barrier between you and the source. Mind you, it was subtle but it was there. I repeated this a few times and while it was not always there with all music, I could spot it on some tracks.
Next I played some of my bass heaving tracks i use for headphone testing. Here, it was easy to notice that bass impact was softented. But also, highs were exaggerated due to higher distortion. Despite loss of high frequency hearing, I found that accentuation unpleasant. WIth tracks that had lisping issues with female vocals for example, the DS DAC made that a lot worse.
Conclusions
Many of you would have guessed that PerfectWave DirectStream DAC would not do well on the bench. That is of course true. The real issue in my opinion though is not the internal design of the DAC functionality. After all, there are other companies that use the same approach and achieve superb measured results. The problem here is simple: the output transformer. Transformers are just not linear to the levels we expect to have in a DAC (as opposed to say, in a tube amp).
The sonic effects are there in my semi-formal tests. Perhaps the older audiophiles including the designer Ted Smith, have lost so much high frequency hearing that the harmonic distortion this DAC adds makes up for some of that and they think they are hearing more. If that is the case, just buy a Jensen transformer and put it on the output of your DAC and you would get the same effect!
The fidelity loss here is apparent in both objective and subjective testing. There is nothing to hang your hat on hence the decapitated Pink Panther which leaves no room for such attire.
Needless to say, I cannot in any shape or form recommend the PerfectWave DirectStream DAC.
These companies need to hire trained listeners and perform a simple level matched test before spending so much money on engineering and producing subpar products. Our hobby deserves better. Much better.
------------
As always, questions, comments, recommendations, etc. are welcome.
Our weather is getting colder but I think I am going to jump in the bay and see if that can uplift my spirits after this test. To enable the uplifting, some more money in my bank account would surely help. So please consider donating using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/
Why throw good money after bad?Hi,
@amirm I have a Directstream DAC. I decided to play around with modifications after reading your review. The three I am trying are:
- Replace the stock Edcor PCW 150/150 transforms with highest Edcore model, the XS4400.
- The dac was made to take the xs4400 so it is compliant with the dac's specs per the PS Audio.
- One of the main conclusions of your review is: "The problem here is simple: the output transformer."
- Based on the conclusion, will the upgrade to the XS4400 help cure the distortion and other problems?
- Use an outboard DC 12V power supply to power the analogue board in the DAC.
- I am thinking about the Farad super 3, but it is expensive
- The dac is split into two boards, a digital board and an analogue board.
- The external power supply bypasses the internal AC to DC power supply to give clean DC power directly to the analogue board.
- It is an easy mod, unplug the power to the analogue board and plug in a specially terminated Farad cable.
- I read the article about using AC power regenerators saying that clean AC power into a component does not change the signal coming out. This case seems different in that the internal AC to DC converter is bypassed.
- Maybe this will solve the problem stated in the review about the Mains power supply "Despite my use of balanced XLR output, we have mains leakage. And on top of that, power supply related spurious tone"
- Put RF/EMI shielding/absorption paper on the display, around the ethernet input board and on big chips in the dac.
- I am going to use 3M AB6005S. It is sticky on the back so it will be easy to apply.
- Maybe this will rectify the issue noted in the Mains Power Supply about "And some odd tones at 80 Hz and such. Perhaps coming from the microprocessor controlling the front panel?"
I don't have an electronics background. I read several of your articles. I want to see if using your articles in conjunction with posts about modifying the PS Audio DAC will give better sound. The challenging part is knowing if anything changed for the positive without measurements.
Your comments are greatly appreciated.
Kind regards,
Alan
Its too easy to buy a new DAC. It's more fun to experiment, take cards out, and talk to my local electronics guy. However, at some point I may try a Topping.Why throw good money after bad?
Sell the PS Audio. Buy any one of the superbly measuring DACs that cost a small fraction of what you’ll get for the PS.
Fair enough. The journey can be fun tooIts too easy to buy a new DAC. It's more fun to experiment, take cards out, and talk to my local electronics guy. However, at some point I may try a Topping.
if you do some measurements I bet you'd not be happy, but at the end of the day it's you who enjoys the music, I would say if you already pulled the trigger, just enjoy music and not the gear. and TBH, while not to say it's measuring horrible, it's measured ok-ish without comparison, a lot of stereo amp can't produce that SINAD anyway so in the end of the day you might just ended up indifferent.Well, I made the upgrades to my PS Audio PerfectWave DirectStream DAC. Added the Edcor XS4400 transformers, added Farad Super 3 Power Supply, and 3M AB6005S EMI absorber. Sounds great, but I don't know if it is psychoacoustics. Wish I could get some measurements.... Have to see if I start using it more than my CD player and turntable. In the past I could never figure out why people started streaming, and got rid of their other source components. Maybe this will change.